Change Log

[2026-10-30] — Weekly Red/Amber refresh — Google P4 → GREEN (Spend Cap budgets GA); Databricks A4 → AMBER (Supervisor Agent moved to legacy)

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-10-30].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-10-30].html

Cloud Assessment

On-Prem Assessment

No Change (verified accurate)

• Four parallel research agents swept the 23–30 Oct 2026 window (hyperscalers, Snowflake/Databricks, on-prem vendors, Teradata + partnership Track A/B), with key findings re-verified against primary sources.
• A1/A2/A7 databricks (cloud): stay G — Genie Agents (GA, formerly Genie Spaces) replace the legacy Knowledge Assistant/Supervisor Agent for no-code agents and ground on GA Metric Views; Lakebase remains the GA memory store (managed memory and sessions are Beta). Text rewritten.
• A2/google (cloud): stays A — Knowledge Catalog semantic-model ingestion (LookML, BigQuery Graph, dbt MetricFlow, Snowflake, Databricks, SAP CSN) went GA on 30 Oct and BigQuery data agents publish to Gemini Enterprise, but the Looker → Gemini Enterprise path and the Gemini Enterprise Knowledge Catalog connector remain Pre-GA. Flagged for manual review next run.
• T2/T4/P2/P3 databricks (cloud): Genie Ontology still Public Preview; Ossie import/export only visible as flag-gated error classes; new service policies, MCP jailbreak screening and Agent Services still Beta.
• Snowflake (cloud): K6 data-quality dashboard and T4 Ossie procedures still Preview; CoWork user memory (A7) has no Preview banner but no dated GA announcement — stays A; Snowflake on OCI announced (P6, no on-prem). EXPEDITION (3–6 Nov) to re-check.
• Microsoft (cloud): Fabric IQ Ontology still preview (T2 A); Foundry memory preview (A7 A); Foundry Local GA and Sovereign Control Panel announced for Ignite (P6 A, re-check after Ignite); agent guardrails all preview (P2 A).
• AWS (cloud): Glue Data Catalog business context and SearchAssets semantic search Preview (K3/K7 A); AgentCore CLI 1.0 GA is pro-code only (A1 A); OpenSearch intelligent ingestion is a separate service (K1 A); no governed metric layer (T1/T3/T6 A).
• Teradata (both): Tera Context Engine / Harness still “Request early access”; WisdomAI-in-Tera not yet GA — T1/T6/K4/A1/A2/A4 stay A. Expanded Teradata Factory agentic stack timing corrected to Q1 2027 in cloud narrative. Q3 earnings 5 Nov.
• SAP (on-prem): Agent Gateway declared GA at TechEd (27 Oct), but n8n non-SAP workflows remain early-adopter only and Joule Studio GA is phased — A4 stays A (text updated). BDC Ontology Manager is beta (T2 A); Enterprise AI Gateway announced (P2/P3/P7 A).
• Oracle (on-prem): Private Agent Factory 26.10 adds Business Knowledge Graph + Deep NL2SQL grounding — T6 stays A (only Deep NL2SQL consumes it; manual review); strengthens GREEN A6/A7.
• IBM (on-prem): Agent Identity GA is SaaS-only (P3 A); Orchestrate per-agent cost tracking not confirmed on Software Hub (P4 A).
• Cloudera (on-prem): AI Studios still Technical Preview; Cloudera Blueprints with NVIDIA Agent Toolkit “coming soon” — A1/A4 stay A. Neo4j joined the Cloudera AI ecosystem (K4 stays R — announcement only).
• SAS / Dell (on-prem): no Viya agent-framework GA (A7 R); Dell semantic layer / knowledge graph / knowledge agents remain 1H 2027 roadmap; Mistral Studio validated solution is an ISV blueprint (A max).
• Partnerships: Track B re-confirmed 16 rows (Informatica→Oracle Select AI and AtScale→Snowflake XMLA strengthened); Cloudera × VAST Data flagged needs-research (scope drift). New rows: Cloudera × Neo4j (needs-research), Microsoft × IntuigenceAI, Snowflake × NVIDIA Kumo, Snowflake × Oracle OCI (documented). No rating impact.

Narrative Updates

• VLN.microsoft.agentic.why / excels (cloud) — corrected pre-existing stale claims (“AMBER”, “knowledge-grounded agents remain AMBER”, “Teradata have since joined” GREEN on A1/A4); Microsoft, Snowflake and Google are now the three vendors GREEN on both A1 and A4.
• VLN.snowflake.agentic.excels (cloud) — A4 GREEN count five → four (Databricks removed).
• VLN.google (cloud) — vision, policy.why/capabilities/excels/gaps rewritten (Policy six-of-seven GREEN); knowledge gaps and vendorDiffs corrected (BigQuery Graph is GA since 31 Aug 2026, not Preview).
• VLN.databricks (cloud) — vision and agentic why/capabilities/excels/gaps rewritten around the 12 Oct legacy reshuffle (Genie Agents, Agent Bricks CLI Beta); stale “managed MCP servers (Preview)” replaced with Genie One MCP Service GA.
• STATIC_NARRATIVES.execSummaryTakeaways[2], strategicFindings[0] and [2] (cloud) — A4 count 5 → 4, Databricks A4 AMBER, scores updated (Google 94%, Microsoft 92%, Databricks 91%).
• STATIC_NARRATIVES.competitiveLandscape (cloud) — Databricks (strongest Agentic / Translation, weakest Policy) and Google (strongest Policy / Knowledge / Agentic, weakest Translation) re-derived; mustAccelerate A1/A4 benchmarks now Microsoft Copilot Studio and Snowflake Cortex Agents + CoWork.
• STATIC_NARRATIVES.fieldGuidance vs. Databricks and vs. Google (cloud) — updated for A4 AMBER and P4 GREEN / A2 AMBER.
• T4/microsoft and T4/google details + rationale + TECH (cloud) — Apache Ossie Microsoft Converter (26 Oct, open-source offline tool); Knowledge Catalog one-way semantic-model import (GA 30 Oct).
• A4/sap details + rationale + TECH (onprem) — Agent Gateway GA (TechEd 27 Oct) replaces “not yet generally available / bidirectional targeted Q4 2026”.
• STATIC_NARRATIVES.editionDate (cloud + onprem) → “30 October 2026, 09:29 CET”.

References Added

Cloud:

• Google Cloud Billing Spend Cap Budgets GA — Google, 2026: GA spend-cap enforcement for Gemini API / Gemini Enterprise Agent Platform and per-user quotas; informs the Google P4 upgrade.
• Databricks Legacy Agent Offerings — Databricks, 2026: Supervisor Agent / Knowledge Assistant moved to legacy with Genie Agents and the Agent Bricks CLI (Beta) as replacements; informs the Databricks A4 downgrade and A1/A2/A7 refreshes.

[2026-10-23] — Weekly Red/Amber refresh — no rating changes; Microsoft/Google join Apache Ossie; Teradata Factory Tera timing → Q1 2027

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-10-23].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-10-23].html

Cloud Assessment

On-Prem Assessment

No Change (verified accurate)

• Four parallel research agents swept the 9–23 Oct 2026 window (hyperscalers, Snowflake/Databricks, on-prem vendors, Teradata + partnership discovery). No AMBER/RED capability reached GA; no downgrades.
• T4/microsoft and T4/google (cloud): both vendors backed Apache Ossie (Open Semantic Interchange) ~30 Sep–2 Oct 2026; Microsoft announced a bidirectional Power BI ↔ Ossie converter that has not shipped, and no Ossie interchange exists in Looker/BigLake — stays A (wording corrected, see below).
• Microsoft (cloud): Fabric IQ Ontology still Preview (T2 A); Foundry Memory Store still experimental (A7 A); Foundry Local GA and Sovereign Control Panel announced for Ignite (Nov 2026) — P6 A, re-check next run.
• Google (cloud): Knowledge Catalog self-learning context enrichment and Knowledge Studio launched 22 Oct in Preview (T2 A); Gemini agent industry specializations still Preview (A2 A).
• AWS (cloud): real-time ACL enforcement for Amazon Quick + Bedrock Knowledge Bases (7 Oct) hardens grounding but no governed metric layer — K1/T1/T3/T6/A1/A2 stay A.
• Databricks (cloud): Unity Gateway service policies still Beta; new MCP tool-call jailbreak screening (21 Oct) also Beta — P2 stays A. Lakeflow system tables and Genie Code tasks GA (outside tracked scope).
• Snowflake (cloud): data-quality dashboard still Preview (K6 A); Snowflake Postgres data mirroring GA 15 Oct (outside tracked scope).
• Teradata (both): Tera Context Engine / Harness / Agent Skills still Q4 2026 roadmap on Teradata Cloud, now Q1 2027 on Teradata Factory (20 Oct release) — T1/T6/K4/A1/A4/A2 stay A. 20 Oct data-protection release (Anonomatic, format-preserving encryption, post-quantum TLS 1.3, immutable backups) strengthens already-GREEN P1/P5; Teradata Sovereign Cloud on Google Distributed Cloud air-gapped is Public Preview (P6 already G).
• Oracle (on-prem): RU 23.26.4 (20 Oct) adds DBMS_AI_GRAPHRAG and Select AI NL-to-graph — strengthens already-GREEN K4; K3 stays A (glossary / auto-classification gap unaffected).
• IBM (on-prem): Software Hub 5.4.0 Patch 8 (20 Oct) ships Agent Runtime v2 and AgentOps GA on-prem — strengthens already-GREEN agentic ratings; P3/P4 stay A (dashboards' on-prem scope contradictory).
• SAP (on-prem): Joule Studio still “soon to be generally available” (SAP Connect, 15 Oct); AI Agent Hub GA in LeanIX is SaaS-hosted — A2/A4/P2/P3/P7 stay A. SAP TechEd (27 Oct) to re-check.
• Cloudera: Agent Studio still Technical Preview (3.1.0); no 1.5.6. SAS: Agentic AI Accelerator v2.2–2.3.1 still unsupported GitHub framework (A1/A4 A, A7 R). Dell: no new ISV blueprint GA; semantic layer / knowledge graph / knowledge agents remain 1H 2027.
• Partnerships: 4 needs-research rows resolved — IBM × Copilot Studio → not-found; IBM × Langfuse, SAS × Phoenix/Langfuse, SAS × Hugging Face TEI → confirmed-documented. New rows: Teradata × Anonomatic (certified, P1/P5), Microsoft × Apache Ossie and Google × Apache Ossie (incidental, T4). Track B re-confirmed: Microsoft × Neo4j, Teradata × Immuta, Cloudera × Mistral AI. No rating impact.
• Manual review: Snowflake CoWork user memory (A7) and Databricks Genie Ontology (T2) docs no longer show a Preview banner, but no dated GA announcement exists — unchanged pending confirmation.

Narrative Updates

• T4/microsoft details + rationale (cloud) — replaced “not a member of / absent from Open Semantic Interchange” with Apache Ossie backing (Oct 2026) and the unshipped Power BI ↔ Ossie converter; TECH → “XMLA endpoint + Apache Ossie (member)”.
• T4/google details + rationale (cloud) — added Apache Ossie backing (Oct 2026) with no shipped Looker/BigLake interchange; TECH → “Looker API + BigLake; Ossie member”.
• STATIC_NARRATIVES.strategicFindings and execSummaryTakeaways (onprem) — Tera Context Engine timing corrected from Q4 2026 to Q1 2027 on Teradata Factory.
• STATIC_NARRATIVES.editionDate (cloud + onprem) → “23 October 2026, 17:26 CEST”.

References Added

Cloud:

• Open Semantic Interchange (OSI) — Snowflake, 2025-2026 (updated): now notes Apache Incubator entry as Apache Ossie and Microsoft/Google backing (Oct 2026); informs Microsoft and Google T4 rationale.

[2026-10-16] — Weekly Red/Amber refresh (no rating changes)

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-10-16].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-10-16].html

Cloud Assessment

On-Prem Assessment

No Change (verified accurate)

• Three parallel research agents swept the 9–16 Oct 2026 window (cloud GA, on-prem GA, Teradata + partnership discovery). No AMBER/RED capability reached GA; no new in-gap partnerships.
• Snowflake: Horizon Context (11 Oct) Preview; Cortex Agent SDK (12 Oct) Public Preview — A7 stays AMBER (Threads/session-oriented, no persistent-memory GA).
• Databricks: Unity Gateway service-policy items (9 Oct) still Beta — P2 stays AMBER; managed MCP/Skills API still Beta/Preview.
• Microsoft: Foundry Agent Memory still preview — A7 stays AMBER.
• AWS: no unified K1 layer GA; QuickSight/Quick semantic Topics still pre-GA — K1/T1 stay AMBER; AgentCore remains BYO-framework runtime (A1 AMBER).
• SAP: Joule Studio 2.0 managed GA still pending (Q3 slipped); A2A bidirectional Q4 2026; BDC Connect for Fabric slipped to Q1 2027.
• IBM Software Hub frozen at 5.4.0 (patch-only); Oct watsonx GA items are IBM Cloud SaaS, not the on-prem train.
• Cloudera AI Studios still Technical Preview (no 1.5.6/SP4). SAS Agentic AI Accelerator remains a GitHub framework (A1/A4 AMBER). Oracle: no on-prem RU this window.
• Dell: 6 Oct Unified Semantic Layer + Enterprise Knowledge Graph + Knowledge Agents all slated 1H 2027 (not GA); no new ISV blueprint this week.
• Teradata: Elastic Python Compute in AI Studio GA 14 Oct (data-science compute — outside the tracked requirement set, no rating impact); Tera Context Engine / Tera Harness / Agent Skills remain Q4 2026 roadmap (cloud + on-prem A1/A4 AMBER).

Narrative Updates

• editionDate (both assessments) → "16 October 2026, 07:55 CEST".

[2026-10-09] — Weekly Red/Amber refresh — no rating changes

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-10-09].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-10-09].html

Cloud Assessment

On-Prem Assessment

No Change (verified accurate)

• Snowflake agentic (cloud): Cortex Agent code-execution tool reached GA 5 Oct 2026 — strengthens an already-GREEN capability; no rating movement (STRENGTHEN enrichment is reserved for the monthly green-refresh). Native agent memory still Threads-only (A7 stays A).
• AWS (cloud): no first-party no-code agent builder GA (A1 A); QuickSight multi-dataset Topics semantic layer still preview (T1 A); no unified storage-layer GA (K1 A).
• Microsoft (cloud): Foundry Agent Memory still preview (A7 A); Fabric IQ Ontology still preview.
• Databricks (cloud): Unity Catalog service policies still Beta (P2 A); Governance Hub not GA (P3 A); managed MCP not GA.
• Google (cloud): no new BigQuery Graph / knowledge-graph GA; "Gemini agent" (8 Oct) launched largely in preview.
• Dell (on-prem): 6 Oct announcement of Unified Semantic Layer, Enterprise Knowledge Graph and Knowledge Agents — all explicitly slated for 1H 2027, not GA; T1/K4/T6/A2 unchanged (A). No new ISV blueprint GA; Data Analytics Engine AI Assistant not shipped.
• SAP (on-prem): Joule Studio 2.0 managed GA still pending (Q3 target slipped); Agent Gateway bidirectional A2A targeted Q4 2026; BDC Connect for Microsoft Fabric slipped to Q1 2027. A1/A4/A7/K1 unchanged (A).
• Teradata (on-prem): Factory broad on-prem GA already reflected (15 Jul 2026); Tera Context Engine / Harness / Agent Skills remain Q4 2026 roadmap — A1/A4 stay A.
• IBM (on-prem): Software Hub frozen at 5.4.0 (June 2026); recent activity is patches only — no new on-prem feature GA.
• Cloudera (on-prem): AI Studios / Agent Studio still Technical Preview (no 1.5.6 / SP4 feature GA).
• Oracle / SAS (on-prem): no new agentic or semantic-layer GA in-window.
• Partnerships: no new in-window gap-area partnerships; Dell×Accenture (8 Oct) is a GSI co-sell, not an ISV blueprint.

Narrative Updates

• STATIC_NARRATIVES.editionDate (cloud + onprem) — stamped to this run (09 October 2026, 13:00 CEST).
• STATIC_NARRATIVES.refreshScope (cloud + onprem) — reset RAG → RA (weekly Red/Amber run; the monthly full Red/Amber/Green refresh ran 02 Oct 2026).

[2026-10-02] — Monthly full Red/Amber/Green refresh: no rating changes; 19 GREEN capabilities strengthened; SAP Q3-2026 roadmap items slipped

🔴🟡🟢 Full Red/Amber/Green update

Files: Knowledge Platform Comparison - Cloud [2026-10-02].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-10-02].html

First run of the month, so the Phase 2g GREEN-Refresh deep sweep ran in addition to the regular Red/Amber pass. All 174 GREEN requirement+vendor pairs (100 cloud + 74 on-prem) were re-researched via parallel per-vendor agents for net-new GA developments in the Aug–Oct 2026 window, deduplicated against the 2026-09-11 green-refresh. No rating changed in either assessment: every AMBER→GREEN candidate re-verified as still preview or slipped (Snowflake K6 data-quality dashboard still Public Preview; Microsoft Foundry Agent Memory still preview; SAP Joule Studio 2.0 managed GA slipped past its Q3 2026 target). The sweep captured 19 details-only STRENGTHEN findings inside already-GREEN capabilities.

Cloud Assessment

On-Prem Assessment

GREEN Strengthened (details-only)

• T1/databricks (cloud): Metric Views now shareable across metastores/accounts via OpenSharing (GA 24 Sep 2026) — stays G
• T3/databricks (cloud): governed scalar + batch Unity Catalog Python UDFs GA (23 Sep 2026) — stays G
• P7/databricks (cloud): tag-driven ABAC GRANT policies GA for models, model services and MCP services (31 Aug 2026) — stays G
• K4/google (cloud): GA natural-language chat-with-graphs in Conversational Analytics + Data Engineering Agent graph building (Oct 2026) — stays G
• K7 & P7/google (cloud): Knowledge Catalog Data Products expose Apache Iceberg REST assets (GA 17 Aug) + dbt metadata import (GA 24 Sep); stale "BigQuery Graph (Preview)" corrected to GA — stays G
• T5/google (cloud): Conversational Analytics market-basket + AI.KEY_DRIVERS GA (Sep) and chat-with-graphs (Oct) — stays G
• T6 & T1/microsoft (cloud): Fabric IQ grounds Microsoft Copilot in governed Power BI semantic models, on by default (GA FabCon Barcelona, 28 Sep 2026) — stays G
• K1/microsoft (cloud): OneLake open-mirroring extended with GA zero-copy mirroring for Google BigQuery + SharePoint Lists — stays G
• A1 & A2/snowflake (cloud): Cortex Agents Coding Agent sandbox GA (26 Aug) + session-scoped temporary agents GA (16 Sep); agent skills reference Cortex Extensions GA — stays G
• A4/snowflake (cloud): CoWork Automations GA (11 Sep 2026) — stays G
• T3/snowflake (cloud): Declarative Change Management DEFINE SEMANTIC VIEW GA (30 Sep 2026) — stays G
• A4/aws (cloud): next-generation AgentCore Runtime GA (18 Sep 2026) — serverless microVM + snapshot cold starts — stays G
• A1 & A4 & A7/oracle (onprem): Private Agent Factory 26.7 GA (17 Aug 2026) — reusable agent Templates + expanded model providers (A1), A2A communication + new workflow nodes (A4), agent memory on by default (A7) — stays G
• K4/sap (onprem): HANA Cloud Knowledge Graph Engine added GeoSPARQL (QRC3, Sep 2026) — stays G
• K1/dell (onprem): ObjectScale 4.4 brought native S3 Tables (Apache Iceberg) to GA + S3-over-RDMA across all-flash + X7700 query-in-place (31 Aug 2026) — stays G
• T5/dell (onprem): Starburst Enterprise 482-e LTS extended AIDA with GA custom personas, per-data-product rules, audit logging and Gemini provider (11 Sep 2026) — stays G
• P6/dell (onprem): ObjectScale 4.4 hardened sovereign posture (allow-list OS, automated CA rotation, offline/air-gapped compliance reporting) — stays G

No Change (verified accurate)

• Teradata (both): Aug–Oct announcements — Tera Context Engine, Tera Harness, Agent Skills — are Q4 2026 forward-looking, not GA; no strengthen.
• IBM (onprem): on-prem train frozen at IBM Software Hub 5.4; the Aug–Oct GA items (Trace Inspector, custom LLM-as-Judge, AgentOps Agent) are IBM Cloud SaaS-only — no on-prem strengthen.
• SAS (onprem): Viya 2026.08–10 are incremental monthly stable releases; nothing crossed a GA bar.
• Cloudera (onprem): frozen at Data Services 1.5.5 SP3; AI Studios / Agent Studio remain Technical Preview.
• SAP (onprem): Joule Studio 2.0 managed runtime, embedded n8n, and BDC Connect for Fabric all missed their Q3 2026 GA target (BDC Connect for Fabric slipped to Q1 2027); A4/A7 hold AMBER.

Narrative Updates

• editionDate (both) — 02 October 2026, 12:16 CEST. refreshScope (both) — RAG (Phase 2g green-refresh ran). methodologyPara2 (both) — "as of October 2026".
• SAP (on-prem) Q3-2026 roadmap text corrected now that the quarter closed without GA: A7 details, A4 rationale and A7 rationale now describe Joule Studio 2.0 managed memory and embedded n8n as pre-GA as of early October 2026 (ratings hold AMBER).
• Google (cloud) K7 details and the Google knowledge VLN corrected the stale "BigQuery Graph (Preview)" claim to GA (31 Aug 2026).
• Partnership registry: Track B updates recording that the Q3-2026-targeted partnerships slipped — Snowflake↔Collibra, SAP↔n8n and SAP↔Microsoft (BDC Connect for Fabric), all re-dated 2026-10-02; const PARTNERSHIPS rebuilt (47 cloud / 62 on-prem rows).

[2026-09-25] — Weekly Red/Amber update: 5 upgrades — Databricks MCP GA, Microsoft knowledge-grounded agents GA, AWS ontology off RED, Teradata Factory GA lifts on-prem agent memory, Dell AIDA GA

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-09-25].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-09-25].html

Regular weekly Red/Amber pass over the full 181-cell AMBER/RED gap matrix across both assessments, plus the mandatory partnership sweep. Twelve parallel vendor sweeps produced five rating changes. Every rating-change driver was re-verified against primary vendor documentation before being applied, and that verification rejected three further upgrades the research proposed — Google A2, Oracle K1 and Snowflake P7 — each for a stated reason recorded below.

The largest single finding is not a rating at all: Teradata Factory is generally available on-premises, confirmed against Teradata's own 15 July 2026 announcement. That contradicts the Q2 FY26 earnings remark the 2026-09-18 edition relied on, and required rewriting roughly 39 passages of on-premise narrative that described Factory as being in controlled availability pending an H2 2026 GA. Only one rating actually moved as a result (A7), because the other Teradata on-premise AMBERs rest on capability gaps rather than delivery — a distinction this edition now makes explicit.

Cloud Assessment

On-Prem Assessment

No Change (verified accurate)

• Google (cloud) — A2 upgrade rejected. Publishing Looker data agents to Gemini Enterprise is still Preview: the docs carry the Pre-GA Offerings Terms verbatim and state "the Looker – Gemini Enterprise integration is in preview with limited support". What reached GA were sub-features of published agents (visualizations, verified queries, semantic search GA 4 Sep 2026), not the publishing path — so the exact blocker named in the existing rationale still stands.
• Oracle (on-prem) — K1 upgrade rejected. The existing AMBER rationale already separates the closed query gap (hybrid vector search GA on-prem Jan 2026) from the open storage gap; binary artefacts still live in Object Storage and image/video understanding still calls the separate OCI Vision service, and K1 requires a single storage layer "without requiring separate systems". Oracle AI Database 26ai on-prem GA is also Linux x86-64 only. T6 held for a related reason: Select AI grounds on schema + RAG, not on a governed semantic model.
• Snowflake (cloud) — P7 upgrade rejected. Genuine GA movement on the AI/agent half (Trust Center AI Security 3 Aug, agent lineage 2 Sep, external lineage and restricted session scope 3 Sep, AI Agent Identity Jun 2026), but data products remain listings and shares rather than a first-class governed object with the ownership, versioning, quality SLAs and approval workflows P7 names. Evidence folded into details without moving the rating.
• P4 held at AMBER for Microsoft, AWS, Google, Cloudera and Teradata under a single consistent bar: GREEN requires AI-specific per-agent budget enforcement (as Snowflake and Databricks ship), not general capacity FinOps. Microsoft's Fabric Chargeback and surge protection GA, Cloudera's Quota Management GA (1.5.5 SP3) and IBM's Software Hub 5.4 multitenancy cost attribution are all infrastructure-level; Google Spend Caps and AWS hard token budgets remain pre-GA or absent.
• Cloudera (on-prem) — AI Studios / Agent Studio remain Technical Preview on both cloud and on-premises 1.5.5, despite "Agent Studio 3.0" marketing that reads like GA; there is no 1.5.6/1.6/2.0 on-prem release, only service packs. All agentic ratings hold.
• SAP (on-prem) — managed Joule Studio 2.0 is still in Early Adopter Care with GA targeted Q3 2026 and unmet as of 25 Sep (5 days from quarter end); SAP Domain Models and AI Agent Hub observability/identity carry the same unmet Q3 target; bidirectional A2A remains Q4 2026. All 18 AMBER requirements hold.
• IBM (on-prem) — the current on-prem train is still IBM Software Hub 5.4.0 (June 2026); the Jul–Sep shipments in these areas are IBM Cloud SaaS-only. T3/T6 "pending agent-reach" confirmed still pending: watsonx BI 3.5.0 on Software Hub ships no metrics API, no MCP server and no Orchestrate integration.
• SAS (on-prem) — Viya 2026.06–2026.09 were maintenance releases; no requirement crossed a GA bar. SAS AI Navigator missed its Q3 2026 GA target and is Azure-Marketplace SaaS only, so it cannot lift on-prem P2/P3/P4 regardless.
• Dell (on-prem) — ObjectScale S3 Tables remain Tech Preview and AI-Optimized Search is still "coming 2H26", so K2 holds; T1 holds because AIDA is a conversational NL-to-SQL assistant, not a declarative metrics layer.

Narrative Updates

• editionDate (both) — 25 September 2026, 08:15 CEST. refreshScope (both) — RA (green-refresh did not run). methodologyPara2 confirmed current (September 2026).
• Teradata Factory GA correction (on-prem, ~39 passages): every "controlled/early availability … broad on-prem GA targeted H2 2026" clause replaced with the GA fact across requirement details, rationales, TECH labels, the Teradata vision and agentic VLN blocks, vendorDiffs, competitiveLandscape, mustAccelerate and two strategic findings. The A1/A4/A6/P4/T1 rationales were rewritten to rest on the capability gap alone, since the delivery argument no longer exists. The Teradata vision records the earnings-call conflict explicitly rather than hiding it.
• Teradata A7 ripple (on-prem): winsToday gained an agent-memory row; the Teradata agentic GREEN count moved from two of seven to three of seven; the Teradata GREEN list became K1, K2, K5, A3, A5, A7, P1, P5, P6.
• Oracle "only vendor with managed in-database agent memory" corrected — an inverted-trigger claim in strategicFindings broken by Teradata's change, now "a distinction it now shares with Teradata".
• Dell T5 ripple (on-prem): five passages claiming the Data Analytics Engine AI Assistant "is not evidenced as shipped"/"slipped its 1H CY26 target" corrected — it shipped as Starburst AIDA. Updated the Dell translation VLN block, vendorDiffs (new strengths bullet; the "all six T requirements" gap bullet now correctly says five), competitiveLandscape and the "Against Dell" field guidance.
• Cloud A3 became the assessment's first universally GREEN requirement (6/6). The winsToday A3 row was corrected from "5 of 6" to all six — caught by the 4c-8 validator as an inverted-trigger failure — plus the agentic takeaway, strategic finding 3 and the "vs. Databricks" field guidance.
• Cloud A2 count updated from two GREEN vendors to three across the executive summary takeaways, strategic findings, the Microsoft agentic VLN gaps, vendorDiffs and the "vs. Microsoft" field guidance.
• Cloud score claim rewritten: five of six vendors now exceed 80% (AWS crossed at 80.5%), with Databricks and Microsoft tied at 91.8%.
• Downgrade signals recorded (no rating movement): AWS maintenance-mode wave of 30 Jun 2026 closing Bedrock Agents Classic, Kendra, SageMaker Clarify, Model Monitor and Role Manager to new customers from 30 Jul (K6, P7 details); Microsoft Healthcare data solutions moving to a downloadable package with the managed solution closed to new customers 1 Oct 2026 and support ending 31 Dec 2027 (K5 details); SAP Data Intelligence on-premises end of maintenance 31 Dec 2026 (K1 details).
• Product-name corrections: SAS Information Catalog → SAS Data Governance (renamed in Viya 2026.06); IBM P3/P4 rationales refreshed to record the first-party agent-identity preview runtime and Software Hub 5.4 multitenancy cost attribution.
• Partnership registry: 17 rows added or updated. New — Teradata↔Dell (certified, Factory hardware) and Teradata↔Microsoft OneLake; SAP↔Vercel, Pydantic AI/LlamaIndex, Mistral AI/Cohere; Cloudera↔Mistral AI, Protegrity; Databricks↔seven Unity Gateway guardrail ISVs; Google↔four graph-visualization partners; AWS↔Stardog; Snowflake↔Microsoft Power BI. Upgraded — Teradata↔WisdomAI to certified/embedded (still not GA, Q4 2026). Flagged needs-research — SAS↔Arize Phoenix/Langfuse and SAS↔Hugging Face TEI (not corroborated in SAS first-party repos), IBM↔Microsoft Copilot Studio and IBM↔Langfuse (absent from current Orchestrate integration docs; possible native supersession).

References Added

Cloud:

• Databricks Managed MCP & Genie One MCP Service GA — Databricks, 2026: existing card retitled and extended with the Genie One MCP GA (22 Sep 2026), the GA SaaS connectors and Unity Gateway MCP API, and the Beta endpoint deprecation — the source for the Databricks A3 upgrade.
• Microsoft Fabric Data Agent GA — Microsoft Learn, 2026 (new): documents the GA statement, the governed grounding sources including Power BI semantic models via NL2DAX, the credential and DLP enforcement model, and the M365 Copilot / Copilot Studio GA milestones — the source for the Microsoft A2 upgrade.
• AWS Context Ontology Accelerator GA — Amazon Web Services, 2026 (new): documents the Apache-2.0 accelerator, its W3C-standard knowledge-graph output and MCP server, and that the managed AWS Context successor is still "coming soon" — the source for the AWS T2 upgrade.

On-Prem:

• Teradata Autonomous Knowledge Platform GA — Teradata, 2026: card corrected and extended with the verbatim 15 July 2026 availability bullets, and now records the 4 Aug 2026 earnings-remark conflict and why the written announcement is treated as authoritative — the source for the Teradata A7 upgrade.
• Dell AI Factory, Dell Lakehouse and Starburst Enterprise Documentation — Dell Technologies, 2024–2026: extended with the Starburst Enterprise 480-e LTS AIDA GA release note, the Guardrails GA, and Dell's own characterisation of AIDA — the source for the Dell T5 upgrade.

[2026-09-18] — Weekly Red/Amber update: IBM T1 upgraded to GREEN (on-prem semantic layer); all other fast-moving items still preview

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-09-18].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-09-18].html

Regular weekly Red/Amber pass over the full AMBER/RED gap matrix plus the always-on partnership discovery sweep, one week after the 2026-09-11 green-refresh. Nearly every fast-moving item resolved to still preview/beta or already correctly rated. A follow-up deep-dive on the two manual-review flags produced one rating change: IBM T1 (on-prem) moves AMBER → GREEN after verifying — against IBM's own documentation — that watsonx BI is GA and installable on IBM Software Hub / Cloud Pak for Data (on-premises), correcting the prior (incorrect) view that it was IBM Cloud SaaS only. This makes IBM the sole on-premise leader at 89% of the theoretical maximum (Oracle 88%). The second flag (AWS A1 "Bedrock 2 Agent Builder GA") was confirmed to be an unofficial-blog myth — AWS A1 stays AMBER.

Cloud Assessment

On-Prem Assessment

No Change (verified accurate)

• Databricks (cloud) — Unity Catalog service policies still Beta (P2); Governance Hub still Private Preview (P3/P7); Databricks-managed MCP servers still Public Preview (A3); no native knowledge graph — Neo4j Virtual Graph is Public Preview only (K4). All stay AMBER.
• Microsoft (cloud) — Foundry Agent Service Memory still preview (A7). No change.
• Snowflake (cloud) — native agent memory still Threads-only (A7 AMBER); Semantic Studio still Preview (T1 GREEN holds on GA semantic views); no first-party knowledge graph or industry data models (K4/K5 stay RED).
• AWS (cloud) — AgentCore Harness GA is a BYO-framework runtime, not a first-party no-code builder (A1 stays AMBER); the third-party "Bedrock 2 Agent Builder GA" claim has no official AWS source, and Bedrock Agents Classic is closed to new customers — A1 holds AMBER; still no unified structured+unstructured storage layer (K1 stays AMBER); no new certified semantic-layer partnership (T1 stays AMBER); A2 knowledge-grounding gate holds.
• Teradata (cloud + on-prem) — on-premises "Teradata Factory" still not broadly GA (Q2 FY26 earnings: "still to go GA in the second half"); Tera Claw multi-agent orchestration remains research preview (YE2026 target). On-prem A1/A4/A7 and T1/T2/K4 stay AMBER.
• Oracle (on-prem) — Oracle AI Data Platform Agents reached GA (27 Jul 2026) but that is the OCI/cloud surface; Oracle's on-prem A1/A4/A2/P2 were already GREEN. No change.
• SAP (on-prem) — managed Joule Studio 2.0 still in Early Adopter Care with managed GA targeted Q3 2026 (not met as of Sept 2026), so A4/A7 stay AMBER; bidirectional A2A remains a Q4 2026 target.
• Dell (on-prem) — Data Analytics Engine AI Assistant still no GA (T1/T5 stay AMBER); ObjectScale S3 Tables / AI-optimized search remain Tech Preview (K2 stays AMBER); no new ISV blueprint GA.
• IBM (on-prem) — watsonx.data Context lakehouse-native semantic runtime is still Private Preview and SAL is enrichment metadata only, but watsonx BI (a separate installable IBM Software Hub service) already delivers a GA on-premises governed metrics semantic layer — so T1 moves to GREEN (see On-Prem table). T3/T6 stay AMBER pending agent-reach.
• SAS (on-prem) — Viya Copilot + Agentic AI Accelerator + Retrieval Agent Manager + MCP Server already reflected in the data and held AMBER per the first-party no-code/orchestration bar (A1/A4). No change.
• Cloudera (on-prem) — AI Studios / Agent Studio remain Technical Preview; agentic requirements stay AMBER/RED.

Narrative Updates

• editionDate (both) — set to 18 September 2026, 08:52 CEST.
• refreshScope (both) — set to RA (regular Red/Amber update; green-refresh did not run).
• methodologyPara2 (both) — "as of" date confirmed current (September 2026).
• Dell T1 detail (on-prem) — "not evidenced as shipped as of Aug 2026" freshness stamp advanced to September 2026 (Data Analytics Engine AI Assistant re-verified still not GA; rating unchanged).
• IBM T1 upgrade ripple (on-prem): RATIONALES.T1.ibm, REQUIREMENTS.T1.details.ibm, TECH.T1.ibm (→ "watsonx BI (Metrics catalog + semantic model)"); VLN.ibm.vision, VLN.ibm.translation.{why,capabilities,excels,gaps} and VLN.ibm.agentic.gaps rewritten (Translation no longer a hole; four of six T AMBER); vendorDiffs.ibm.gaps and vendorDiffs.teradata semantic-layer bullets updated.
• Cross-vendor T1 count (on-prem): every "two vendors GREEN T1" claim → three (IBM, Oracle, SAP) in execSummaryTakeaways, strategicFindings, VLN.teradata.translation, mustAccelerate.T1.
• Leadership reframing (on-prem): IBM now sole leader at 89% (Oracle 88%) — updated execSummaryTakeaways (1b, 3), strategicFindings (2, 3), competitiveLandscape.ibm/.oracle, and the "Against IBM" fieldGuidance bullet (semantic-layer wedge closed).

References Updated

On-Prem:

• IBM Product Documentation: watsonx.ai, watsonx Orchestrate, watsonx.data intelligence, IBM Software Hub — existing card extended with the watsonx BI evidence: GA installable on IBM Software Hub since 5.2 (30 Jun 2025), current v3.5.0 on Software Hub 5.4 (Jun 2026), air-gapped private-registry installs supported; governed Metrics catalog, semantic data model, business-term matching, row-level user filters and a natural-language interface — the source for the T1 on-premises AMBER → GREEN upgrade. T1 added to the card's informed-ratings list.

[2026-09-11] — Full Red/Amber/Green refresh (2-week cycle): no rating movement; two GREEN capabilities strengthened

🔴🟡🟢 Full Red/Amber/Green update

Files: Knowledge Platform Comparison - Cloud [2026-09-11].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-09-11].html

This was a full green-refresh pass covering the entire AMBER/RED gap matrix plus a deep re-review of GREEN-rated capabilities across both assessments, following a two-week gap. Every fast-moving item researched this cycle resolved to still preview/beta or already correctly rated — so no ratings changed. Two already-GREEN capabilities gained shipped developments and were strengthened (details only).

Cloud Assessment

On-Prem Assessment

GREEN Strengthened (details-only)

• K4 / Google (cloud) — BigQuery Graph reached GA (31 Aug 2026): ISO GQL / SQL-PGQ with native measures, Conversational Analytics chat, Looker integration and a visual modeler, joining first-party Spanner Graph (Enterprise Knowledge Graph remains Preview). Rating stays G.
• A3 / SAP (on-prem) — ABAP MCP Server reached GA (Aug 2026), connecting third-party AI tools to ABAP systems, and the open-source SAP Build MCP servers (CAP, Fiori, UI5) work with non-SAP IDEs. Rating stays G.

No Change (verified accurate)

• Databricks (cloud) — Unity AI Gateway is GA but its service policies guardrail surface is still Beta (P2); Governance Hub is Beta (P3/P7); core Databricks-managed MCP servers remain Public Preview, though Genie One SaaS MCP connectors reached GA 10 Sep 2026 (A3). All stay AMBER.
• Microsoft (cloud) — Foundry Agent Service Memory is still public preview (A7); Fabric IQ Ontology is still preview — K4 GREEN rests on Graph in Fabric GA, not the ontology. No change.
• Snowflake (cloud) — native Cortex Agents memory is Threads-only (persistent memory needs an external store), so A7 stays AMBER; no first-party knowledge graph or industry data models (K4/K5 stay RED); semantic-view variables are GA and Semantic Studio is Preview — T1 GREEN holds.
• AWS (cloud) — S3 Vectors GA + Bedrock Managed Knowledge Base GA are present but there is still no single unified storage layer (K1 stays AMBER); with T1/T6/K4 all AMBER the A2 knowledge-grounding gate holds (A2 stays AMBER); no first-party no-code agent builder (A1 stays AMBER).
• Google (cloud) — A2 stays AMBER (Looker data-agent publishing constraints unchanged).
• Teradata (cloud + on-prem) — the on-premises "Teradata Factory" is still not broadly GA (Q2 FY26 earnings: "still to go GA in the second half"); Tera Claw multi-agent orchestration remains research preview with a year-end 2026 target; no GA semantic layer or knowledge graph. On-prem A1/A4/A7 and T1/T2/K4 stay AMBER.
• SAP (on-prem) — managed Joule Studio 2.0 is still in Early Adopter Care with managed GA targeted Q3 2026 (not yet met as of Sept 2026), so A4/A7 stay AMBER; BDC Connect for Microsoft Fabric and bidirectional A2A remain Q3/Q4 2026 targets.
• Dell (on-prem) — the Dell-native Data Analytics Engine AI Assistant still has no GA (T1/T5 stay AMBER); ObjectScale AI-Optimized Search / S3 Tables remain Tech Preview (K2 stays AMBER); no native agent-memory, orchestration or governance control plane — agentic/policy stay AMBER via ISV blueprints.
• IBM (on-prem) — the watsonx.data Semantic Automation Layer is GA, but it is semantic-enrichment metadata rather than a metrics/business-terms semantic layer, and the newer watsonx.data Context federated runtime is still maturing — T1 stays AMBER (flagged for manual review).
• Cloudera (on-prem) — AI Studios remain Technical Preview; all agentic requirements stay AMBER/RED.

Narrative Updates

• editionDate (both) — set to 11 September 2026, 18:56 CEST.
• refreshScope (both) — set to RAG (full Red/Amber/Green update; green-refresh ran).
• methodologyPara2 (both) — "as of" date advanced to September 2026.
• VLN/RATIONALES/details K4 Google (cloud) and A3 SAP (on-prem) — strengthened as above (no rating field change).
• SAP A7 (on-prem) — "as of August 2026" date stamp on the Joule Studio 2.0 memory note advanced to September 2026 (still EAC).
• Partnership registry — SAP n8n and SAP–Microsoft (BDC Connect for Fabric) rows re-validated (Track B); both remain Q3 2026 targets, date_confirmed advanced to 2026-09-11; PARTNERSHIPS const rebuilt in both data files.

[2026-08-28] — Weekly verification pass: no rating movement; on-prem Translation-Layer narrative corrected

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-08-28].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-08-28].html

Cloud Assessment

On-Prem Assessment

No Change (verified accurate)

• A3, P2, P3 / Databricks (cloud) — Unity AI Gateway reached GA 4 Aug 2026, but the Databricks release notes confirm service policies (the guardrail surface), agent services and the Databricks-managed MCP connectors all remain Beta as of 6 Aug 2026. All three stay AMBER; the newly announced Governance Hub (26 Aug 2026) is likewise Beta and does not move P7, which is already GREEN.
• A7 / Microsoft (cloud) — Foundry Agent Service Memory is still public preview. Stays AMBER.
• K4 / Microsoft (cloud) — GREEN rests on Graph in Fabric GA and the certified Neo4j Graph Intelligence workload, not on Fabric IQ Ontology, which remains Preview. GREEN holds.
• T1, T6, K4, A1, A2, A4 / Teradata (cloud + on-prem) — the Autonomous Knowledge Platform GA release (15 Jul 2026) names Teradata Cloud, Teradata Factory, AI Studio and AI Services only; no semantic-layer or knowledge-graph GA. TeraClaw multi-agent orchestration remains research preview with a year-end 2026 target. All stay AMBER.
• T1 / IBM (on-prem) — watsonx.data Context, IBM’s semantic-layer entry announced at Think 2026, is in private preview. Stays AMBER.
• A4, A7 / SAP (on-prem) — the managed Joule Studio 2.0 is still in Early Adopter Care with GA targeted Q3 2026; the target has not been met. Both stay AMBER.
• K2 / Dell (on-prem) — ObjectScale AI-Optimized Search (S3 Tables + S3 Vector) is still stated as "globally available beginning 2H26" with no GA announcement. Stays AMBER.
• T1, T5 / Dell (on-prem) — the Dell-native Data Analytics Engine AI Assistant remains unevidenced as shipped after its 1H CY26 target. Both stay AMBER.
• A1, A4 / Cloudera (on-prem) — AI Studios remain Technical Preview. Cloudera Anywhere Cloud, announced 19 Aug 2026 at EVOLVE Singapore, carries no GA date and no component-level detail, so it moves nothing.
• T1 / AWS (cloud) — no change to the SageMaker Catalog semantic-modeling position or the dbt Labs collaboration. Stays AMBER.

Narrative Updates

Eight corrections to the On-Prem assessment, all tracing to Oracle reaching T1 GREEN in an earlier cycle without the downstream prose being re-based. The GREEN truth table records on-prem T1: 2 GREEN → [oracle, sap], but four sections still described SAP as the only GREEN semantic layer — and one asserted the opposite rating outright.

• conclusionIntro[2] (on-prem) — "Only SAP now has a semantic layer rated GREEN (T1)" corrected to name both Oracle and SAP; Oracle removed from the list of vendors with no GA semantic layer, and the residual Oracle constraint restated as reach (the Oracle Analytics MCP Server is Preview and ships in no on-premises release) rather than absence.
• strategicFindings[1] (on-prem) — heading and body rewritten from "only SAP has a GREEN semantic layer" to the two-vendor position; Oracle removed from "lack an AI-ready semantic layer".
• strategicFindings[3] (on-prem) — removed the direct contradiction "the aging OAS/RPD semantic layer (T1 AMBER)", which conflicted with Oracle’s actual GREEN rating and with competitiveLandscape.oracle in the same file.
• strategicFindings[5] (on-prem) — re-based to the Aug 2026 agent-builder standard: A1 and A4 are AMBER on cloud and on-prem (framework-based, no documented first-party no-code builder or orchestration engine), with only A7 being a Teradata-Cloud-only GA awaiting Factory on-prem. The prior text credited "AgentStack AgentBuilder" and "AgentEngine" as GA on Teradata Cloud.
• competitiveLandscape.teradata.oneline (on-prem) — same A1/A4/A7 re-basing applied.
• fieldGuidance[0] (on-prem) — Oracle removed from "have none"; its GREEN-but-not-agent-reachable position stated instead.
• fieldGuidance[3] "Against Oracle" (on-prem) — rewritten. The old wedge ("Oracle Analytics Server is 20 years old… there is no clean answer") is no longer true given OAS 2026 Semantic Modeler GA; the wedge is now agent reach. Product name also corrected from Oracle 23ai to Oracle AI Database 26ai.
• VLN.teradata.agentic.gaps and VLN.teradata.translation.why (on-prem) — gaps still claimed AgentBuilder/AgentEngine were "GA only on Teradata Cloud", directly contradicting the excels field two lines above; why named only SAP as GREEN on T1. Both corrected.

Cloud narrative verified clean — all count and membership claims (T1 four GREEN, A1 four of six, A4 five of six, P2 three of six, P3 four of six, A6 five of six, A7 four, Snowflake two REDs) match the GREEN truth table exactly.

Manual Review — Resolved

The three items flagged for manual review in this run’s report were investigated and closed.

1. Oracle product naming (on-prem, 34 sites). Oracle AI Database 26ai replaced Oracle Database 23ai in October 2025 and reached on-premises GA in the January 2026 quarterly Release Update (23.26.1, 27 Jan 2026). The assessment still carried "23ai" throughout. All 34 sites were re-based to Oracle AI Database 26ai across REQUIREMENTS.details, RATIONALES, TECH, VLN, winsToday, mustAccelerate and vendorDiffs, plus one comparator in the Cloud assessment’s A7 rationale. A single deliberate "23ai" remains, in the references card title ("formerly Database 23ai"), because that card cites source documents spanning 2024–2026.

Two substantive corrections surfaced while doing the pass:

• Oracle A7 provenance was wrong. The file credited "Unified Memory Core (GA Jan 2026)". "Unified Memory Core" is Oracle’s architectural phrase, not the product name, and January 2026 is the on-premises GA date of 26ai itself, not of the agent-memory capability. The shipped product is Oracle AI Agent Memory, announced 24 Mar 2026, with versioned Oracle Help Center documentation and an oracleagentmemory package through release 26.6. A7 stays GREEN — the product is real and documented — but details, rationale, TECH, the IBM comparators and the references card were corrected.
• A3 TECH label read "DB 26ai native MCP server" while every peer names its MCP product; aligned to "AI Database 26ai MCP Server".

2. 4c-3 gate script bug (SKILL.md). The vendorDiffs block terminator was src.find('\n references:', vd_start). In the on-prem data file references: precedes vendorDiffs:, so the search returned −1, the block ran to EOF, and the trailing const PARTNERSHIPS rows were parsed as vendorDiffs bullets — producing two phantom Dell MISMATCH lines from partnership rating_impact notes. Replaced with a terminator that stops at whichever comes first after vendorDiffs: the next top-level key, a const declaration, or the object close. The corrected script was extracted from SKILL.md and run verbatim: 62 cloud and 94 on-prem rating-code references, all matching, no false positives.

3. Cloudera Anywhere Cloud. Launched 19 Aug 2026 at Cloudera’s EVOLVE event in Singapore: a single control plane across public-cloud, sovereign and on-premises tiers, with Apache Iceberg and a Polaris catalog, native zero-trust controls, lineage and access-boundary management, continuous compliance, self-service engine blueprints (Spark, Kafka, Trino and third-party engines), GPU-accelerated Spark 4.1 via NVIDIA cuDF, and an agentic copilot that turns natural-language instructions into executable operational tasks. Cloudera has published no GA designation and no component-level availability dates, and the launch materials address neither Agent Studio, MCP, a semantic layer, vector search nor guardrails — so it moves no rating. Recorded as Policy-Layer narrative context in VLN.cloudera.policy.capabilities and added to the Cloudera references card. The copilot is an operations copilot, not a data-agent builder, so it has no bearing on Cloudera A1/A4.

Partnerships

• google / Neo4j (K4) — resolved from needs-research to confirmed-certified. The prior cycle could not verify the relationship (Neo4j partner pages returned HTTP 403). Neo4j’s own blog for Google Cloud Next ’26, plus PR-wire and trade coverage, confirm an active co-marketed native partnership: the Neo4j AI Agent inside the Gemini Enterprise console, bidirectional BigQuery ↔ Neo4j Graph Data Science, Neo4j in Google Cloud Application Design Center with a Terraform provider for the Aura Admin API, and Enterprise + Community Edition Marketplace listings. No rating impact — Google K4 is already GREEN on first-party Spanner Graph.
• All other registry rows carry confirmed-* status dated 2026-08-14, inside the 12-month cache, so Track B re-validation was correctly skipped this run.

[2026-08-21] — Executive Summary deep-links: clickable RAG heatmap, score bars & layer summary

🛠 Platform / UX enhancement — no rating changes; applies to both the Cloud and On-Premise assessments.

The Executive Summary charts are now clickable and jump straight to the underlying evidence:

• RAG Distribution Heatmap — clicking a cell (a vendor × layer) switches to that vendor's tab and highlights the matching layer card.
• Vendor Score Comparison — clicking a colored bar segment (a vendor × layer) does the same.
• Layer-Level RAG Summary — clicking a row (a layer) switches to the Capability Matrix tab, expands that layer, and highlights its header.
• Return-focus — coming back to the Executive Summary restores the previous scroll position and briefly flashes the cell/segment/row you came from.

Keyboard-accessible (Enter/Space) with hover cursors and destination tooltips; reuses the existing tab-switching and return-focus machinery. No rating, score, or assessment narrative changed.

[2026-08-21] — Weekly Red/Amber sweep + manual-review resolution: even-handed agent-builder/orchestration re-basing (Teradata cloud A1/A4 + AWS A1 down; Teradata on-prem A1/A4/A7 down); IBM agent-identity (P3) refresh

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-08-21].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-08-21].html

Cloud Assessment

(Cloud A7 Teradata stays GREEN — AgentOps persistent memory is a genuine documented first-party GA capability. A3/A5 unaffected.)

Even-handed A1/A4 re-evaluation (all vendors, one bar)

After concluding Teradata's "AgentBuilder"/"AgentEngine" are marketing names over OSS frameworks + MCP with no product docs, the same "documented first-party product" bar was applied to every A1/A4-GREEN vendor:

• A1 (no-code builder): PASS — Snowflake (Snowsight Agents), Databricks (Agent Bricks), Microsoft (Copilot Studio), Google (Agent Studio), IBM (Orchestrate AI Agent Builder), Oracle (OCI GenAI Agents console), SAP (Joule Studio classic). FAIL → AMBER — AWS and Teradata (OSS SDK + runtime, no first-party no-code builder).
• A4 (orchestration engine): PASS — Databricks, Microsoft, IBM, Oracle; documented first-party piece (kept GREEN) — Snowflake, AWS, Google. FAIL → AMBER — Teradata only (no documented first-party orchestration product; SAP A4 already AMBER).
• Cloud GREEN sets: A1 6→4, A4 6→5. Scores: Teradata 80.5%→78.5%, AWS 80.5%→79.5% (both now below 80%) → "all six exceed 80%" and "all six GREEN on A1/A4" headlines flip to "four of six."

On-Prem Assessment

(A3 MCP Server and A5 Enterprise Vector Store stay GREEN — re-grounded on their independent GA on the existing VantageCore IntelliFlex/VMware estate (MCP Server Mar 2026, EVS Apr 2026), not Teradata Factory. On-prem A1/A4 are AMBER for a compounded reason: framework-based (no first-party no-code builder/orchestration engine, same as cloud) and Factory-gated; A7 is AMBER purely for Factory (the capability itself is genuine first-party GA).)

Manual-review resolution (adversarial verification)

Two items flagged for manual review last cycle were re-researched with fair-weighed adversarial agents:

• Teradata Factory on-prem GA — CONFIRMED not broadly GA. Both the 15 Jul GA press release ("available now, on-premises," best read as controlled/early availability via Dell) and the 4 Aug Q2 earnings call (CEO: Factory "still to go GA in the second half"; CFO: "about to roll out"; uncorrected analyst: "not available yet") converge on H2 2026 broad GA. Under the strict scale (controlled/limited availability = A at most), Teradata's on-prem AI Studio / AgentStack capabilities (A1, A4, A7) are not GA on-premises → downgraded G→A. The Enterprise MCP Server (A3) and Enterprise Vector Store (A5) GA'd independently on IntelliFlex/VMware and survive as GREEN.
• Cloudera Agent Studio — CONFIRMED still Technical Preview. The "generally available" line in the Agent Studio 3.0.0 release note is overridden by (a) the parent "AI Studios Overview (Technical Preview)" page stating "not recommended for production deployments" and (b) the still-active administrator gate "Enable AI Studios (Technical Preview)" in Feature Flags (present in 1.5.5 SP3 and /latest/). No 1.5.5 SP4/1.5.6 exists. All Cloudera ratings held (A1/A4/A6/P2 AMBER, P3 RED, K2 AMBER via RAG Studio).

Ripple (inverted-trigger checks)

• On-prem GREEN sets shrank: A1 4→3 (IBM, Oracle, SAP), A4 3→2 (IBM, Oracle), A7 2→1 (Oracle is now the sole on-prem vendor GREEN on agent memory). Teradata on-prem weighted score 80.5% → 77.4%.
• IBM excels corrected ("Teradata now also GA on multi-agent" removed); all "only two vendors" / "one of two" A4/A6 claims re-verified against the truth table and hold (IBM + Oracle).

No Change (verified accurate — regular sweep)

• All near-misses from last cycle held pre-GA. Databricks managed MCP server still Public Preview and Unity AI Gateway guardrail service policies still Beta (only the AI Gateway core went GA 4 Aug); Google Looker→Gemini Enterprise publish path still Preview (A2); Microsoft Foundry agent memory still (paid) Preview (A7); Snowflake data-quality dashboard still Public Preview (K6). All stay AMBER.
• IBM A2 held AMBER despite a genuine GA uplift — watsonx Orchestrate now lets one agent bind multiple knowledge sources (GA 13 Aug 2026) — because the A2 GREEN discipline gate blocks it: IBM T1, T6 and K4 are all AMBER, so the grounding substrate is still document-RAG, not a governed-knowledge layer.
• SAP: every Q3/Q4 2026 target still pre-GA — Joule Studio 2.0 managed / HANA-backed agent memory in Early Adopter Care (A7), n8n embedded on roadmap (A1/A4), A2A outbound-only with bidirectional Agent Gateway targeted Q4 2026 (A4), BDC Connect for Microsoft Fabric not yet GA (K1). A3 (MCP) and T1 (Knowledge Graph semantic layer) GREEN re-verified.
• Oracle: OCI GenAI Guardrails GA re-confirmed (P2 GREEN); A2/A3/A7 GREEN and K3/K5/K7/T2 AMBER all unchanged — no new GA in the window.
• SAS (A1/A4/A2 AMBER, A7 RED) and Dell (ObjectScale S3 Vector still "planned H2 2026" native-not-GA; Data Analytics Engine AI Assistant still slipped 1H CY26; no new ISV-blueprint GA): no change.

Narrative Updates

• STATIC_NARRATIVES (both): editionDate → "21 August 2026, 07:57 CEST"; refreshScope → RA (Red/Amber).
• On-Prem: full Phase 4b re-evaluation for the Teradata downgrades — rewrote VLN.teradata vision + agentic (why/capabilities/excels/gaps), execSummaryTakeaways[6], strategicFindings[6], winsToday (removed A1/A4/A7 rows), mustAccelerate (added "On-premise agentic GA (A1, A4, A7)" row), competitiveLandscape.teradata, vendorDiffs.teradata (strengths + gaps), the Teradata reference card and the MCP standards card (corrected a stale "Teradata AMBER" A3 line), plus IBM excels cross-vendor fix. Re-grounded A3/A5 rationale + details on the IntelliFlex/VMware GA. Six additional spots that asserted AI Studio/AgentStack/AgentOps GA on-premises were corrected to Factory controlled-availability (A2/A6/P4 details & rationales, T1 details/rationale).
• On-Prem: refreshed IBM P3 (rationale, details, TECH label) to name the GA Agent access overview in the watsonx Orchestrate Security control center (17 Aug 2026) — a shipped agent-inventory / shadow-agent-detection capability that supersedes the prior "not yet production-hardened / features emerging" wording. Rating stays AMBER (visibility and governance only; no full agent-identity runtime — IBM Verify-to-Orchestrate credential wiring remains a developer tutorial).

References Added

On-Prem:

• IBM Product Documentation card updated: added the watsonx Orchestrate Security control center Agent access overview (GA Aug 2026) and extended the informed-ratings list to include P3.
• Teradata Autonomous Knowledge Platform GA card updated: clarified that AKP GA is on Teradata Cloud, with on-prem via Teradata Factory in controlled availability (broad GA H2 2026); MCP Server (A3) and Enterprise Vector Store (A5) GA on IntelliFlex/VMware independent of Factory.

[2026-08-14] — Weekly Red/Amber sweep: AWS Translation-Layer breakout, Oracle semantic layer GA — and six proposed upgrades killed by adversarial verification

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-08-14].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-08-14].html

Cloud Assessment

On-Prem Assessment

Adversarial Verification (6 proposed upgrades rejected)

All MEDIUM-confidence proposals were re-tested by independent agents briefed to refute them. Six of eleven did not survive:

• AWS T1 A→G — REFUTED. A Quick Sight Topic has no metric object; business rules are free-text natural-language instructions an LLM reinterprets per question, and AWS guidance tells users to verify the generated SQL. No versioning, no promotion path, no access control on definitions (they are visible to all dataset viewers), and no query surface outside Quick Sight. Held AMBER.
• Cloudera A1/A4/A6/P2 A→G and P3 R→A — ALL REFUTED. The Agent Studio 3.0.0 GA release note (24 Jun 2026) is contradicted by three independent Technical Preview signals, decisively the administrator enablement control labelled "Enable AI Studios (Technical Preview)" — Agent Studio is unreachable except through a gate marked Technical Preview, and Cloudera excludes Technical Preview from production support. Also found: zero results for A2A/agent-to-agent across Cloudera docs, stdio-only MCP transport, and guardrails scoped to Agent Studio workflows only (RAG Studio and AI Inference endpoints unguarded).
• IBM P3 A→G — REFUTED. No documented wiring between IBM Verify/Vault 2.0 and watsonx Orchestrate: the control-plane announcement omits both products, the Verify agentic-identity page never mentions Orchestrate outside site navigation, and the sole integration artifact is a tutorial where the developer hand-registers an OIDC client with a placeholder redirect URI "because the agent does not have an dedicated URL". The Microsoft/Entra precedent verified stronger than assumed, so the consistency argument runs the other way.
• IBM A2 A→G — BLOCKED by the A2 GREEN discipline gate: IBM T1, T6 and K4 are all AMBER, so A2 cannot be GREEN. Same rule that held Teradata A2 on 7 Aug.

Two upgrade rationales were also corrected rather than accepted as written: AWS T3 (Quick Autonomous Agents do not consume Topic semantics and have no verifiable GA) and AWS T6 (the claimed grounding mechanism was factually inverted).

No Change (verified accurate)

• Teradata: 0 of 16 AMBER moved. No governed semantic layer shipped since the AKP GA (15 Jul 2026) — the GA release names only Teradata Cloud, Teradata Factory, AI Studio and AI Services. The GA Data Analyst Agent (AWS Marketplace, 3 Aug 2026) grounds on physical data models, not governed semantics; Tera workspace modes remain "Coming soon".
• Snowflake, Databricks, Google, Microsoft, SAP, SAS, Dell: no rating changes. The near-misses are Databricks A3/P2 (managed MCP and guardrail service policies both still Preview/Beta despite Unity AI Gateway reaching GA 4 Aug 2026), Google A2 (Conversational Analytics API GA but the Looker→Gemini Enterprise publish path carries a Preview banner), Microsoft A7 (Foundry memory is a paid preview since 1 Jun 2026), and Snowflake K6 (data quality dashboard Public Preview 29 Jul 2026).
• Teradata Factory on-prem GA timing softened in the Q2 FY26 earnings call to "H2 2026 / later in 2026" versus the GA release's "available now" — monitor next cycle.

Factual corrections applied

• SAP A4 — our data had the A2A directions inverted. SAP's Architecture Center states the Agent Gateway is not GA, so A2A is unidirectional; the two research agents disagreed on which direction, so the text now asserts only what the official document proves.
• Databricks P2 cited Aporia, acquired by Coralogix (Dec 2024) and no longer an independent going concern. Removed and replaced with verified Unity AI Gateway ecosystem partners.
• Dell T1/T5 described the Data Analytics Engine AI Assistant as "recently launched" — it slipped its 1H CY26 target and does not appear in Dell Data Lakehouse 1.8.0.x documentation.
• Oracle T4 leaned on MDX/XMLA for OLAP, which is deprecated and unsupported beyond 26ai premier support. Oracle P4 claimed "no AI-specific token metering" — false since AIDP Units GA (Jul 2026). Oracle K3/K7 were written against the superseded OCI Data Catalog. Oracle K5 implied an industry-model breadth that no longer exists (Communications Data Model stalled at 12.1 on a 2015-era doc set).
• Product renames: Databricks Vector Search → AI Search; IBM Knowledge Catalog → watsonx.data intelligence (May 2025); SAS Information Catalog → SAS Data Governance and SAS Studio → SAS Data and AI Studio (Viya 2026.06); Snowflake Cortex Code → CoCo; Google Dataplex → Knowledge Catalog. Also corrected: QuickSight was not renamed to Amazon Quick — Quick is the umbrella and Quick Sight is the BI component within it.
• Removed a first-party product wrongly listed as a partnership (IBM's own Watson Orchestrate MCP import).

Narrative Updates

• STATIC_NARRATIVES (both): editionDate → "14 August 2026, 09:09 CEST"; refreshScope → RA (Red/Amber).
• Cloud: rewrote strategicFindings 1 and 2 (leader now 93%; all six vendors exceed 80% for the first time, AWS crossing this cycle), execSummaryTakeaways 2, competitiveLandscape.aws, fieldGuidance (the "AWS has no semantic layer at all" line is now false), vendorDiffs.aws, and VLN.aws across vision/knowledge/translation/agentic.
• On-Prem: rewrote strategicFindings 3 and 4 and execSummaryTakeaways 3 (Oracle drew level with IBM at 87.7%), added a new finding on the shared on-premise lead, updated competitiveLandscape.oracle, mustAccelerate T1 target (benchmarks now Oracle and SAP), vendorDiffs.oracle, and VLN.oracle. Corrected SAP's VLN claim that it was "the only vendor GREEN on a native on-premises semantic layer" — Oracle now is too.

References Added

Cloud:

• Amazon Quick Sight documentation (multi-dataset Topics, Dataset Q&A, custom instructions) — informs AWS T1, T3, T5, T6.
• AWS HealthLake and Amazon Connect Decisions data-model documentation — informs AWS K5.

On-Prem:

• Oracle Analytics Server 2026 "What's New" and Semantic Modeler documentation (doc F91008-09, March 2026) — informs Oracle T1, replacing partner-blog sourcing.

Validation

• All gates pass: vendorDiffs rating-code integrity (cloud 63 / on-prem 92 refs), TECH coverage and rating-sync, winsToday co-GREEN membership, comparative-claim scan against the GREEN truth table, score consistency, roadmap-date scan (all past-due strings are completed GA events or explicitly-resolved slips), and template structural audit.
• Skill bug found: the 4c-3 vendorDiffs script assumes references follows vendorDiffs, but the on-prem file has the reverse order, so the scan ran into the PARTNERSHIPS rows and produced two false MISMATCH lines. Bounded the scan at the next top-level key instead.
• Partnership registry rebuilt comprehensively: 88 rows (41 cloud, 55 on-prem), up from 34, with Track B validity outcomes recorded for every incumbent.

[2026-08-07] — Full Red/Amber/Green refresh: agentic-layer GA wave (Teradata AKP MCP/RAG/memory, Snowflake & Databricks semantic-layer + guardrails, Oracle grounded agents, Dell/SAS catalog + orchestration)

🔴🟡🟢 Full Red/Amber/Green update

Files: Knowledge Platform Comparison - Cloud [2026-08-07].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-08-07].html

Cloud Assessment

On-Prem Assessment

GREEN-Refresh (Phase 2g) coverage

• Full deep GREEN sweep executed across all GREEN requirement+vendor pairs in both assessments (fan-out research agents per vendor). The material developments on already-GREEN capabilities this cycle manifested as the AMBER→GREEN upgrades above; existing GREEN narratives already carried the prior-cycle GA proof points, so no standalone details-only GREEN edits were required. All GREEN ratings re-verified and held (no downgrades).

No Change (verified accurate)

• A2 / Teradata (both assessments): held AMBER — grounding is Enterprise Vector Store RAG + Enterprise MCP access-control; a GA governed-knowledge substrate (semantic layer T1/T6 and knowledge graph K4 are all AMBER) that would ground agents beyond document retrieval has not shipped. The requirement excludes "just retrieve documents via RAG." An A2 GREEN discipline rule was codified in the skill so this is applied consistently going forward (Snowflake/Databricks earned A2 GREEN via a GA semantic layer; Oracle via a GA knowledge graph).
• IBM / A7 (on-prem): held AMBER — watsonx Orchestrate memory reachable only via an --experimental-package-root flag; fails GA bar.
• SAS / A2 (on-prem): held AMBER — citation-grounding is the standalone RAM (same standalone caveat as A4); capability fragmented, not an integrated GA runtime.
• AWS / P6 (cloud): held AMBER — European Sovereign Cloud strengthens sovereignty but AWS remains single-cloud/no hybrid (P6 GREEN bar is genuine hybrid; only Teradata qualifies).

Narrative Updates

• STATIC_NARRATIVES (both): editionDate → "07 August 2026, 09:29 CEST"; refreshScope → RAG (full Red/Amber/Green); methodologyPara2 → "as of August 2026".
• Cloud: rewrote execSummaryTakeaways (agentic + governance), strategicFindings (agentic, governance, Teradata, leader distribution now 5-of-6 >80%), winsToday (+A3/A5/A7 Teradata rows), mustAccelerate (RAG-lifecycle row → agent-evaluation A6), competitiveLandscape (Teradata/Snowflake/Databricks/AWS), fieldGuidance (vs Snowflake/Databricks/AWS), vendorDiffs (Teradata/AWS/Snowflake/Databricks) and VLN for AWS/Databricks/Snowflake/Teradata.
• On-Prem: rewrote execSummaryTakeaways (Oracle all-7 agentic, SAS, Teradata), strategicFindings (Oracle/SAS/Teradata), winsToday (+A3/A5/A7 Teradata rows), mustAccelerate (RAG-lifecycle row → A6), competitiveLandscape (Teradata/Oracle/SAS/Dell), fieldGuidance (vs SAS/Dell), vendorDiffs (Teradata/Oracle/SAS/Dell) and VLN for Oracle/SAS/Dell/Teradata.

References Added

Cloud:

• AWS Bedrock AgentCore Memory GA: managed persistent short/long-term cross-session agent memory (Oct 2025) — informs AWS A7.
• Snowflake release-notes card updated: Semantic Views standard-SQL GA (Mar 2026) + Cortex AI Guardrails GA — informs T3/T6/A2/P2.
• Databricks Business Semantics card updated: T6 + A2 grounding.

On-Prem:

• Oracle Private Agent Factory card updated: A2 (knowledge-graph + source-attributed RAG grounding) → GREEN.
• SAS Viya card updated: Viya Copilot for Information Catalog (K7 GREEN) + Retrieval Agent Manager orchestrator (A4 R→A).
• Dell card updated: Data Orchestration Engine native discovery (K7 GREEN) + Dell/Palantir reference architecture (K4/T2 R→A).

[2026-07-31] — Platform hardening: skill roadmap-date gate + template dead-snapshot cleanup

🛠 Platform / tooling enhancement — no rating changes.

Two maintenance fixes surfaced during a deep validation review:

• Skill 4c-5 roadmap-date gate now catches half-year and CY-prefixed target dates (1H CY26, 2H 2026, H2 2026, 1H-2H 2026, Q1 CY26), not just Q1 2026-style strings. The prior regex was blind to exactly the formats used for Dell's roadmap items, so stale targets could pass review unflagged.
• Both templates: removed the dead, pre-rendered static vendor-card snapshots baked into every #vendor-* div. These were fully overwritten by renderVendorTabs() on page load (so never visible), but had rotted — carrying stale dates and apostrophe-truncated sentences — and polluted validation greps. The vendor cards are, and always were, rendered from VLN data. Net effect: identical rendered output, "zero authored text in templates" now literally true, and each shipped HTML is ~27% smaller (Cloud 522K→379K, On-Prem 567K→391K).

[2026-07-31] — Weekly refresh: no rating changes (validation + edition refresh)

🔴🟡 Red/Amber update — full pipeline re-run; no rating changes this cycle.

Files: Knowledge Platform Comparison - Cloud [2026-07-31].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-07-31].html

Cloud Assessment

On-Prem Assessment

No Change (verified accurate)

• Deep re-validation of all 19 AMBER/RED ratings carrying a forward-looking qualifier ("preview / targeted / roadmap"). Every one confirmed still correct against primary sources: Microsoft Fabric IQ reached GA at Build 2026 but its Ontology component is still Preview (T2 AMBER holds); Databricks Managed MCP still Public Preview per docs (A3 AMBER); Databricks Genie Ontology still Preview/announcement (T2/T6 AMBER); SAP native agent memory GA planned H2 2026, not shipped (A7 AMBER); Cloudera RAG Studio still Technical Preview (K2/A2 AMBER); Snowflake CoWork memory still preview (A7 AMBER).

Narrative Updates

• Dell K2 / T1 / T5 (on-prem) — details-only refresh, rating stays AMBER. The Dell Data Search Engine (Elasticsearch-based, NVIDIA cuVS GPU acceleration) is now available within the Dell AI Data Platform, so the stale "targeted 1H CY26 / 1H-2H 2026" roadmap wording was replaced with present-tense availability across details, RATIONALES, vendorDiffs, and the TECH label. Ratings stay AMBER because the capability is Elastic-partner-delivered (one level of indirection) per the Dell ISV rule; Dell-native ObjectScale S3 Vector remains emerging.

References Added

On-Prem:

• Dell + Elastic — Data Search Engine (registry): Elasticsearch-based vector/hybrid retrieval with NVIDIA cuVS GPU acceleration, available now in the Dell AI Data Platform; documents the K2/T5 provenance and why both stay AMBER.

Validation

• All consistency gates passed: vendorDiffs rating-code integrity (cloud 67 / on-prem 97 refs), TECH coverage & rating-sync, winsToday co-GREEN membership, comparative-claim scan vs. GREEN truth table, and score consistency (cloud leader Google 92.8% ≈ narrative "92%").
• Edition date refreshed to 31 July 2026; refresh scope Red/Amber.

[2026-07-24] — Vendor Technology → vendor-tab deep-link (Architecture tab)

🛠 Platform / UX enhancement — no rating changes.

Vendor technology entries in the Architecture tab are now clickable and jump straight to the vendor's evidence:

• Clicking a vendor's technology card in the requirement drawer switches to that vendor's tab and highlights the matching requirement row.
• Clicking a vendor's technology pill in the diagram (when a vendor is overlaid) does the same.

Reuses the existing Capability-Matrix → vendor deep-link (focusVendorRequirement); keyboard-accessible (Enter/Space on the drawer cards).

[2026-07-24] — Editorial: Teradata product naming alignment

🛠 Platform / editorial update — no rating changes; naming consistency only.

Aligned all Teradata product references across both assessments to the current Teradata Product Architecture naming (effective May 2026):

• ClearScape removed as a product name — in-database capabilities are now described as standard in Teradata Database; ClearScape Feature Store → Teradata Feature Store.
• Vantage removed from our materials — Vantage Industry Data Models → Teradata Industry Data Models; on-premises Vantage 20.00 → Teradata Factory.
• Tera Analyze / Tera Code / Tera Claw → TeraAnalyzer / TeraCode / TeraClaw; "Tera workspace" → "Tera family".

Cloudera's own "AI Workbench" product was intentionally left unchanged (the naming guide governs Teradata products only). Ratings, scores, and assessment narratives are otherwise unchanged.

[2026-07-24] — Platform enhancement: Vendor Technology overlay (Architecture tab)

🛠 Platform / UX enhancement — no rating changes; applies to both the Cloud and On-Premise assessments.

The Architecture tab now surfaces the specific vendor technology behind each functional requirement, colored by that vendor's rating (GA / Preview / None):

• Requirement drawer always lists every vendor's product for the requirement as rating-accented cards with GA / Preview / None badges.
• "Overlay vendor tech in diagram" dropdown (default None) adds a rating-colored technology pill beneath each requirement in the diagram for the selected vendor, keeping the existing layout.
• With the dropdown at None, hovering a requirement shows a tooltip comparing all vendors' technology for that requirement at a glance.

Driven by a new per-requirement × vendor technology map; no rating or assessment narrative changed this entry.

[2026-07-24] — Cloud agentic & governance update (P3/P4/A1/K4 upgrades; Databricks K5 off RED)

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-07-24].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-07-24].html

Cloud Assessment

On-Prem Assessment

No Change (verified accurate)

• A2/all cloud vendors: knowledge-grounding-beyond-RAG frontier held AMBER across all six despite AWS Bedrock Managed KB GA and Databricks Agent Bricks GA — production-scale governed-semantic grounding not yet proven.
• T3/T6 databricks (cloud): GA semantic layer is real, but external-BI consumption breadth (Tableau late 2026) and anti-hallucination proof kept both at AMBER this cycle (deferred by decision).
• A3/A5/A7 teradata (both): AKP-component GA proposed, but Teradata's own live pages still list AgentBuilder as "private preview" — held AMBER pending verification; A1/A4 GREEN retained but flagged for manual review.
• A3 databricks (cloud): Managed MCP servers still Public Preview per official docs — stays AMBER.
• P3 ibm (onprem): IBM Verify Agentic AI Identity Management shipped, but self-managed/on-prem GA unconfirmed — held AMBER (strong next-cycle candidate).
• A4 sas (onprem): Agentic AI Accelerator is a GitHub project; true multi-agent/supervisor coordination unconfirmed — held RED.
• SAP (onprem): Joule Studio 2.0 in Early Adopter Care (GA targeted Q3 2026); agent memory pre-GA; bidirectional A2A roadmap — no changes.

Narrative Updates

• Full VLN + vendorDiffs re-evaluation for the five changed cloud vendors (Snowflake, Microsoft, AWS, Google, Databricks).
• Cross-vendor ripple: A1 now GREEN across all six vendors (winsToday, Executive Summary / Strategic Findings item 3); P3 now GREEN for four of six — Snowflake, Microsoft, AWS, Google (item 4, competitiveLandscape, mustAccelerate, fieldGuidance); K4 now two GREEN (Google, Microsoft); P4 now two GREEN (Snowflake, Databricks).
• Currency fixes: AWS "Bedrock Agents" → AgentCore (now maintenance-mode) in A7; Google "Dataplex" → "Knowledge Catalog (formerly Dataplex)" in P1.
• editionDate updated to 24 July 2026 (both assessments); refreshScope RA.

References Added

Cloud:

• Microsoft Graph in Fabric & Neo4j Graph Intelligence GA: documents the GA first-party graph engine and certified Neo4j workload informing Microsoft K4 (AMBER→GREEN).
• AWS Bedrock AgentCore Identity GA: documents OBO token exchange + Cedar policy informing AWS P3 (AMBER→GREEN).
• Databricks Industry Data Models: documents the 40+ industry data-model accelerator repo informing Databricks K5 (RED→AMBER).
• Updated: Snowflake Summit 2026 card (A1→GREEN), Google Cloud Next 2026 card (Agent Identity GA, P3→GREEN), Databricks Unity AI Gateway card (Spend Controls GA, P4→GREEN).

[2026-07-17] — Teradata Autonomous Knowledge Platform GA: agent builder + multi-agent to GREEN

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-07-17].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-07-17].html

Cloud Assessment

On-Prem Assessment

No Change (verified accurate)

• A5, A6/teradata (both): AgentOps shipped with the AKP GA, but the full RAG-Ops evaluation lifecycle and dedicated agent evaluation remain roadmap — stays A
• T1, T3/teradata (both): AKP GA did not ship a governed metrics/business-terms semantic layer — stays A
• K6, P4/teradata (both): ModelOps monitoring / AI Studio Consumption & Showback FinOps are GA, but dedicated data observability and per-agent AI FinOps controls remain roadmap — stays A
• A3/databricks (cloud): Managed MCP servers remain Public Preview (official docs Jul 14 2026) — stays A

Narrative Updates

• Teradata VLN.vision + .agentic (both assessments) — rewritten from "AKP forthcoming / private preview" to "AKP GA Jul 15 2026; AgentBuilder + AgentEngine GA (A1, A4 GREEN)"; held-AMBER capabilities reframed as "not yet among shipped GA features" rather than "pending not-yet-GA AKP"
• Teradata VLN.translation + .policy (both) — refreshed T1/T3/P4 AKP framing to post-GA
• STATIC_NARRATIVES.winsToday (both) — added A1 and A4 GREEN rows
• STATIC_NARRATIVES.mustAccelerate — removed A4 row (cloud); removed A1 + A4 rows (on-prem); refreshed T1/T3/A5/P4 rows to post-GA framing
• STATIC_NARRATIVES.execSummaryTakeaways + strategicFindings (both) — Teradata now GREEN 8/27 (cloud) and agent-builder/multi-agent GA on-prem; cloud A4 "five vendors GREEN" → "all six"
• STATIC_NARRATIVES.competitiveLandscape.teradata (both), fieldGuidance (both), vendorDiffs.teradata (both) — updated for A1/A4 GREEN and AKP GA
• Cross-vendor fixes: Microsoft "both A1 and A4" membership (cloud) now names Teradata as joined; IBM & Oracle "one of two vendors running multi-agent" (on-prem) reframed to the still-exclusive multi-agent + agent-evaluation (A4+A6) pairing
• editionDate → 17 July 2026, 08:17 CEST (both)

References Added

Cloud & On-Prem:

• Teradata Autonomous Knowledge Platform GA — Teradata, 2026 — replaced the prior "Announcement" card; documents the July 15 2026 GA across cloud/on-prem/hybrid and the A1/A4 upgrade, with T1/T3/A5/A6/K6/P4 held AMBER as not-yet-shipped capabilities.

[2026-07-10] — Weekly Red/Amber sweep — no rating changes; T1 co-GREEN count correction

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-07-10].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-07-10].html

Weekly regular (Red/Amber) sweep one week after the 2026-07-03 full Red/Amber/Green edition. Researched the highest-velocity AMBER/RED candidates; all remain pre-GA and hold their current ratings. One stale cross-vendor count from last week's Teradata T1 downgrade was corrected.

Cloud Assessment

On-Prem Assessment

No Change (verified accurate)

• T1/T3/K6/A5/P4 teradata (both): Autonomous Knowledge Platform / Teradata Factory still "expected Q3 2026" (press release 19 May 2026) — pre-GA, AMBER holds.
• T2/K4 snowflake (cloud): Cortex Sense enters private preview mid-July 2026 — not GA; T2 stays R, K4 stays A.
• A3/P2/A7 databricks (cloud): Managed MCP servers and managed agent memory remain Beta/Preview — ratings unchanged.

Narrative Updates

• VLN.google.translation.excels (cloud) — corrected stale count "five GREEN T1 vendors" → "four GREEN T1 vendors" (Teradata T1 was downgraded G→A on 2026-07-03, leaving Databricks/Google/Microsoft/Snowflake as the four T1-GREEN vendors).
• STATIC_NARRATIVES.editionDate (both) — refreshed to 10 July 2026, 07:45 CEST.
• STATIC_NARRATIVES.refreshScope (both) — set to RA (regular Red/Amber run; prior edition was RAG).

[2026-07-03] — V3.1: uniform shipped-GA standard — Teradata GREEN correction + cross-vendor GA wave

🔴🟡🟢 Full Red/Amber/Green update

Files: Knowledge Platform Comparison - Cloud [2026-07-03].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-07-03].html

This edition re-applied one consistent standard to every vendor's GREEN: a rating stays GREEN only if backed by a shipped, generally-available product — not an announcement, private preview, or a general-purpose tool standing in for an AI-specific requirement. The GREEN-Refresh sweep found Teradata resting on the not-yet-GA Autonomous Knowledge Platform in several places, and the same lens tightened the AI-specific Cost/FinOps (P4) column across vendors.

Cloud Assessment

On-Prem Assessment

No Change (verified accurate, held to the standard)

• A2/all (cloud + on-prem): the knowledge-grounded-agents frontier stays AMBER — evidence is RAG/connector grounding, not governed-semantic reasoning (Google/Microsoft/IBM/Oracle candidates held).
• P6/Google (cloud): GDC air-gapped GA, but held AMBER for consistency with AWS Outposts / Azure Stack (also AMBER).
• P2/Microsoft (cloud): Prompt Shields/Content Safety GA, but groundedness/hallucination detection still preview — held AMBER.
• T6/IBM, T4/SAP (on-prem): held AMBER — grounding/openness advancing but the semantic layer proper is not GREEN/is proprietary.
• P4/SAP (on-prem): held GREEN — AI Core Consumption Monitor provides GA token metering (genuinely AI-specific).

Narrative Updates

• VLN.teradata (both assessments) — full rewrite: GREEN now only on K1, K2, K5, P1, P5, P6; Translation and Agentic layers entirely AMBER pending the AKP.
• STATIC_NARRATIVES.winsToday (both) — removed Teradata T1, T3, A5, P4 (cloud also K6) rows now that those are AMBER.
• STATIC_NARRATIVES.mustAccelerate (both) — added T1, T3, A5, K6, P4 as Teradata gaps.
• execSummaryTakeaways / strategicFindings (both) — semantic-layer count corrected (cloud: four GREEN vendors, Teradata not among them; on-prem: only SAP GREEN on T1); A4/A6 counts updated (cloud A4 now 5 GREEN, A6 now 5); leader recomputed (cloud Google 92%, on-prem IBM 88%).
• competitiveLandscape / fieldGuidance / vendorDiffs (both) — rating codes and comparative claims re-synced for every changed vendor; cross-vendor VLN contradictions (IBM K2/T5/P2/P7, Oracle T5/P2/P3, SAP K4, SAS A3, Cloudera K2/A2, Dell A7, Google K4/K6/T2/A2/A4/A6/P6/P7, Microsoft K6/P7, Snowflake P4) reconciled.

References Added

Cloud:

• Google Cloud Next 2026 — Spanner Graph, Knowledge Catalog & Gemini Enterprise Agent Platform: informs Google K4/K6/T2/A4/A6/P7.
• AWS Bedrock Multi-Agent Collaboration & AgentCore A2A GA: informs AWS A4.
• Microsoft Purview Unified Catalog DQ, DSPM-for-AI & Agent 365 GA: informs Microsoft K6/P7.
• Snowflake AI Credits & Cortex Agents Resource Budgets GA: informs Snowflake P4.
• Teradata AKP card updated to record the T1/T3/K6/A5/P4 downgrades under the shipped-GA standard.

On-Prem:

• IBM/Oracle/SAP/SAS/Cloudera/Dell vendor cards updated to record this cycle's upgrades (IBM K2/T5/P2/P7, Oracle T5/P2/P3, SAP K4, SAS A3, Cloudera K2/A2, Dell A7) and the P4 downgrades. Teradata AKP card updated with the on-prem downgrades.

[2026-06-26] — Agentic-layer GA wave (5 upgrades) + Teradata A1 correction & product-naming refresh

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-06-26].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-06-26].html

Correction — On-Prem Teradata A1 (GREEN → AMBER)

Product-naming refresh (Teradata go-forward architecture): replaced legacy "ClearScape Analytics AgentBuilder" terminology throughout the on-prem assessment with the current naming — Teradata Autonomous Knowledge Platform, Teradata AI Studio, Teradata AgentStack (AgentBuilder/AgentEngine/AgentOps), Teradata Cloud, Teradata Factory, Teradata Fabric. (Other legacy names — Vantage*, QueryGrid, IntelliFlex — were already absent.)

Cloud Assessment

On-Prem Assessment

No Change (verified accurate)

• AWS A4 (cloud): AgentCore harness GA (Jun 18 2026) is single-agent; multi-agent still via Strands SDK composition — stays AMBER
• AWS A7 (cloud): managed memory already GA; no new in-window GA event — stays AMBER
• Snowflake/Databricks (Summit 2026): Cortex Sense, Advanced Semantics, Genie Ontology, Managed MCP, AI Spend Controls all confirmed Preview/Beta — no change
• Microsoft/Google: Fabric IQ Ontology, Foundry memory, Enterprise Knowledge Graph, BigQuery Graph, Vertex agent eval/identity all still Preview/Pre-GA — no change
• SAP/Cloudera/SAS/Dell/Teradata (on-prem): nothing crossed to GA in-window (SAP Joule Studio 2.0 / agent memory / bidirectional A2A H2-Q4 2026; Dell K2/T5 roadmap 1H-CY26-pending; Teradata AgentStack private preview)

Narrative Updates

• VLN.aws.agentic (cloud) — why/capabilities/excels/gaps: A6 GREEN (AgentCore Evaluations); clarified A4 harness single-agent, A7 memory GA-but-AMBER
• STATIC_NARRATIVES.execSummaryTakeaways + strategicFindings (cloud) — A6 now GREEN for four of six vendors (added AWS; corrected to also name Databricks)
• STATIC_NARRATIVES.competitiveLandscape.aws (cloud) — oneline notes A6 GREEN
• VLN.oracle.vision + VLN.oracle.agentic (on-prem) — agentic layer now GREEN overall (A4, A6 GREEN via Private Agent Factory 26.4); Translation now the lone gap
• VLN.ibm.vision + VLN.ibm.knowledge + VLN.ibm.agentic (on-prem) — watsonx.data intelligence GA self-managed (K7 GREEN); first-party MCP servers (A3 GREEN); removed stale "only vendor" multi-agent+eval claim (Oracle now matches)
• STATIC_NARRATIVES.execSummaryTakeaways + strategicFindings (on-prem) — Oracle agentic GREEN; IBM dual-claim de-duplicated; corrected stale "two vendors with agent builder" (A1 GREEN = four vendors)
• STATIC_NARRATIVES.competitiveLandscape.ibm + .oracle + mustAccelerate (on-prem) — refreshed
• STATIC_NARRATIVES.vendorDiffs.ibm + .oracle (on-prem) — catalog/MCP/multi-agent bullets updated to current ratings
• VLN.teradata.vision + VLN.teradata.agentic (on-prem) — A1 downgraded to AMBER (Karini.AI partnership path; AgentStack private preview); RAG Ops (A5) now the lone shipped agentic strength; legacy "ClearScape Analytics AgentBuilder" naming removed
• STATIC_NARRATIVES.execSummaryTakeaways[5] + strategicFindings[5] + competitiveLandscape.teradata + winsToday (removed A1 row) + mustAccelerate (added A1 row) + vendorDiffs.teradata (on-prem) — reframed Teradata agentic story around the corrected A1; IBM/Oracle/SAP named as the on-prem GA agent-builder vendors
• Partnerships registry: Teradata–Karini.AI row updated — A1/teradata AMBER on BOTH assessments (previously mislabeled on-prem GREEN); date_confirmed refreshed
• Conclusion-tab deep validation (both assessments) — corrected co-GREEN vendor counts against ground-truth ratings: cloud winsToday T3 (now Microsoft + Google) and competitiveLandscape.microsoft (Translation co-leader with Google, not "only"); on-prem winsToday K1 (+IBM), K5/A5/P4/T3 (+SAS), P6 count (all seven GREEN, not six), strategicFindings #6 (SAP ships GA agent builder but A4 multi-agent is AMBER, not GA), strategicFindings #3 (IBM = most complete shipped agent platform; catalog renamed to watsonx.data intelligence)
• Win/Loss Matrix (Conclusion tab, both templates) — added an adjustable tie band slider (default 1 weighted point): layer/overall margins within the band render as TIE instead of WIN/LOSS, so razor-thin leads (e.g. Teradata's +1-pt cloud Policy edge over AWS & Google, driven solely by its unique P4/P6 GREENs) no longer read as confident WINs. Matrix is fully JS-rendered with an explanatory caption; band 0 reproduces the prior strict behaviour. Skill 4c-7 audit extended to keep the control wired.
• editionDate (both) updated to 26 June 2026, 09:14 CEST

References Added

Cloud:

• AWS Bedrock AgentCore Evaluations GA — Amazon Web Services, 2026: documents the GA agent evaluation framework; informs AWS A6 (AMBER to GREEN).

On-Prem:

• IBM watsonx.data intelligence card updated: added self-managed Agentic Data Intelligence GA (Jun 17 2026) and first-party MCP servers; informs K7 and A3 (both AMBER to GREEN).
• Oracle Database 23ai / 26ai card updated to include Private Agent Factory 26.4 (GA Jun 17 2026); informs A4 and A6 (both AMBER to GREEN).
• MCP standards card updated: IBM A3 AMBER to GREEN; Cloudera corrected to AMBER.

[2026-06-19] — Conference-season GA wave: agent identity, observability, ontology, memory & open standards

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-06-19].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-06-19].html

Cloud Assessment

On-Prem Assessment

No Change (verified accurate)

• SAS A4 (multi-agent orchestration): stays RED. Verification refuted the proposed upgrade — the "handoff orchestration pattern" is Microsoft/Azure language, not SAS; the only SAS multi-agent supervisor (SAS 360 Agent) is announced/not-GA and confined to Customer Intelligence 360, not core Viya.
• IBM P2 (guardrails) & A2: stays AMBER — watsonx.governance real-time detectors appear to be shipping but no GA label could be pinned to a primary IBM source (kept AMBER per "don't inflate").
• Oracle P2 (guardrails): stays AMBER — OCI GenAI Guardrails GA (Feb–May 2026) are cloud-only; this is the on-prem assessment.
• Teradata (both): AgentStack / Autonomous Knowledge Platform still private preview / Q3 2026 future-dated.
• Cloudera (on-prem): all AI Studios remain Technical Preview — no GA.
• Dell (on-prem): ObjectScale S3 Vector, Data Search Engine, AI Assistant all still 1H–2H 2026 roadmap; Palantir on-prem AI OS announced May 2026 but undated/not-GA (registered as watch item — K4/T2 stay RED).

Details-only updates (no rating change)

• Snowflake P2: AI_REDACT PII detection/redaction now GA (Dec 2025) — note corrected ("lacks PII filtering" was stale).
• AWS A7: AgentCore UserPreferenceMemoryStrategy + episodic memory GA (May 2026) — dropped "lacks user-preference learning".
• Databricks P4: Unity AI Gateway hard spend caps added (Beta, Jun 2026); T6/K4: Genie Ontology grounding/context graph noted.
• SAS A1/K2: Viya Copilot now GA; RAM vector-DB maturity.

Narrative Updates

• Cloud strategicFindings[0] — leader Microsoft updated 85%→87%; "Five of six vendors exceed 80%" (was four).
• Cloud VLN/STATIC_NARRATIVES — Snowflake (2 RED ratings, K4 AMBER, P3 GREEN), Microsoft (A6/P3 GREEN), Databricks (T2 AMBER), AWS (T4 AMBER), Google (T3/A7 GREEN); competitiveLandscape, fieldGuidance, vendorDiffs, mustAccelerate P3 benchmark all refreshed.
• On-Prem VLN/STATIC_NARRATIVES — SAP (T4 AMBER), SAS (A5 GREEN); competitiveLandscape, takeaway/finding #5, fieldGuidance (Against SAS), vendorDiffs refreshed.
• Edition date set to 19 June 2026, 09:33 CEST in both assessments.

References Added

Cloud:

• RelationalAI Knowledge Graph Coprocessor (Snowflake Native App): documents the GA in-Snowflake knowledge graph — informs Snowflake K4.
• Microsoft Entra Agent ID & Microsoft Foundry Observability GA: agent identity + eval/observability GA — informs Microsoft P3, A6.
• Databricks Data + AI Summit 2026 — Genie Ontology & AI Gateway: native ontology Preview + hard spend caps — informs Databricks T2, T6, P4.
• Google Vertex AI Agent Engine Memory Bank & Dataplex Data Products GA: agent memory + data products GA — informs Google A7, T3.
• Open Semantic Interchange (OSI) card updated to record AWS working-group membership (informs AWS T4); existing Snowflake Summit card updated with AI Agent Identity + AI_REDACT.

On-Prem:

• SAP Datasphere/Business AI card updated: HANA Cloud Knowledge Graph Engine W3C standards (informs SAP T4 R→A).
• SAS Viya / Model Manager card updated: SAS Retrieval Agent Manager GA (informs SAS A5 A→G).

[2026-06-12] — June 2026 cycle: Snowflake CoWork multi-agent GA; SAP & Cloudera agentic updates; vendorDiffs integrity fixes

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-06-12].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-06-12].html

Cloud Assessment

On-Prem Assessment

No Change (verified accurate)

• Microsoft, AWS, Google (cloud): no tracked AMBER item moved to GA — Fabric IQ ontology, AWS native semantic layer, Google Enterprise Knowledge Graph & BigQuery Graph all remain Preview/absent.
• Databricks (cloud): MCP still Public Preview (A3 AMBER), Spend Controls still Beta (P4 AMBER) — Databricks Summit (Jun 15-18) falls after this window; recheck after.
• Teradata, IBM, Oracle: no GA in window. Teradata AgentStack midyear GA target due but unconfirmed — held at AMBER.
• SAS A4 (Agentic Accelerator is a GitHub starter kit, not GA) — stays RED. Dell K2/T1/T5 (ObjectScale S3 Vector, AI Assistant, NL interface still "targeted Q2/1H CY26") — stay AMBER.
• SAP A3 Commerce Cloud MCP — still Q2 2026 target, not GA.

Narrative Updates

• VLN.snowflake (cloud): vision + agentic why/capabilities/excels/gaps rewritten for CoWork multi-agent GA (A4 GREEN) and User Memory preview (A7 AMBER).
• VLN.sap (onprem): vision + agentic why/capabilities/gaps updated for native agent memory (A7 AMBER) and Joule Studio 2.0 / A2A.
• VLN.cloudera (onprem): agentic why/capabilities/gaps updated for Cloudera Agent Studio + AI Workbench MCP Server.
• STATIC_NARRATIVES (cloud): execSummaryTakeaways[3], strategicFindings[3], competitiveLandscape.snowflake, fieldGuidance (vs Snowflake), vendorDiffs.snowflake — "three vendors now GREEN on A4". editionDate + methodologyPara2 → June 2026. Databricks P4 label "Public Preview" → "Beta" (3 spots).
• STATIC_NARRATIVES (onprem): execSummaryTakeaways[5], strategicFindings[5], competitiveLandscape.sap/cloudera, fieldGuidance (vs SAP, vs Cloudera). editionDate + methodologyPara2 → June 2026. A3 IBM details softened ("nascent" → publishes GA watsonx.data MCP servers). A4 SAP/Dell details refreshed.
• vendorDiffs integrity fixes (pre-existing rating-code errors corrected, 4c-3/4c-4 gates): cloud databricks & google K4 RED→AMBER (Neo4j); onprem teradata A1 AMBER→A6 AMBER (Teradata A1 is GREEN on-prem); onprem sap gaps A3 RED/A4 RED/A7 RED corrected to GREEN/AMBER/AMBER; onprem sas A3 RED→AMBER; onprem dell A3 GREEN→AMBER; onprem cloudera "all six T RED"→"four of six", RED count 11→10; onprem sap T5 GREEN→AMBER; SAS Policy "GREEN across all 7 P" → "5 of 7 GREEN" (3 spots, P2/P3 are AMBER).

References Added

Cloud:

• Snowflake Product Release Notes & Summit 2026 updated: added CoWork GA (Jun 2 2026) with multi-agent orchestration + GA MCP Connectors + User Memory preview; A4 noted as upgraded to GREEN.

On-Prem:

• SAP Datasphere and SAP Business AI Documentation updated: added Sapphire May 2026 — Joule Studio 2.0 + native agent memory (HANA Cloud runtime + Company Memory on Signavio); informed A7 upgrade.
• Cloudera CDP Documentation updated: added Cloudera Agent Studio + AI Workbench MCP Server; informed A4 upgrade.

[2026-05-20] — Snowflake agent memory R→A; Databricks MCP + FinOps details updated; IBM MCP GA confirmed

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-05-20].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-05-20].html

Cloud Assessment

On-Prem Assessment

No Change (verified accurate)

• A3/Databricks (cloud): Managed MCP servers still in Public Preview (May 7, 2026) — stays AMBER; details updated
• P4/Databricks (cloud): Unity AI Gateway AI Spend Controls in Public Preview (May 19, 2026) — stays AMBER; details updated
• A7/AWS (cloud): Bedrock Agents memory confirmed GA with 365-day retention + AgentCore memory — stays AMBER; details updated
• K4/Google (cloud): Enterprise Knowledge Graph still Preview as of May 19, 2026 — stays AMBER
• A3/IBM (onprem): Watson Orchestrate MCP import GA (May 5, 2026) — stays AMBER (client-side only); details updated
• A1/SAS (onprem): Agentic AI Accelerator on GitHub, no formal releases, early-stage — stays AMBER

Narrative Updates

• VLN.snowflake.vision — "4 RED ratings" → "3 RED ratings"; removed "no agent memory"; added "Cortex Agents Threads now provide basic cross-session agent memory (A7 AMBER)"
• VLN.snowflake.agentic.why — removed A7 from structural gaps; added Threads note (A7 AMBER)
• VLN.snowflake.agentic.gaps — updated: "(A7 RED)" → "(A7 AMBER)"; described Threads capability and remaining limits
• STATIC_NARRATIVES.competitiveLandscape.snowflake — oneline updated: "4 RED ratings (K4, K5, T2, A7)" → "3 RED ratings (K4, K5, T2)"; added Threads/A7 AMBER note
• STATIC_NARRATIVES.fieldGuidance vs.Snowflake (cloud) — updated: "4 RED ratings remain: … no agent memory" → "3 RED ratings remain: …"; added A7 AMBER context

References Added

(No new reference cards added this run — all sources were existing docs/release notes.)

[2026-05-08] — Teradata Autonomous Knowledge Platform announced; cloud A4/Microsoft + A6/Snowflake + on-prem A6/IBM upgraded to GREEN

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-05-08].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-05-08].html

Cloud Assessment

On-Prem Assessment

No Change (verified accurate)

• A3/Databricks (cloud): MCP still Public Preview — stays AMBER
• A4/Teradata (cloud): AgentStack still private preview, cloud GA targeted Q3 2026 per Autonomous Knowledge Platform announcement — stays AMBER
• T2/Microsoft (cloud): Fabric IQ ontology still Preview, billing meters not yet published — stays AMBER
• K4/Google (cloud): Enterprise Knowledge Graph + BigQuery Graph still Pre-GA — stays AMBER
• A1/SAS (onprem): Viya Copilot + Agentic AI Accelerator on GitHub, early-stage — stays AMBER (details updated)
• A4/SAP (onprem): Joule Studio multi-agent confined to SAP domain — stays AMBER
• K4/SAP (onprem): HANA Cloud Knowledge Graph Engine GA Q1 2025 confirmed — stays AMBER per SAP-domain confinement (details updated)
• T1/Oracle (onprem): Oracle Analytics AI Agents added but RPD remains aging — stays AMBER

Narrative Updates

• VLN.teradata.vision (cloud + onprem) — rewritten to frame around Autonomous Knowledge Platform announcement May 7, 2026 (Q3 cloud GA, Tera Factory on-prem later 2026)
• VLN.teradata.knowledge.capabilities (cloud + onprem) — added Pinecone partnership and Connected Data Foundation (Iceberg/Delta)
• VLN.teradata.translation.gaps (cloud + onprem) — added WisdomAI documented partnership context
• VLN.teradata.agentic.why/capabilities/gaps (cloud + onprem) — added Tera workspace (Tera Analyze, Tera Code, Tera Claw research preview), Tera Agents (platform infrastructure agents available now), Q3 2026 cloud GA timing, and revised competitor context (Microsoft now also GREEN on A4 cloud; IBM A6 GREEN on-prem)
• VLN.snowflake.vision + VLN.snowflake.agentic.why/capabilities/excels/gaps (cloud) — A6 GREEN with Cortex Agent Evaluations GA; Snowflake Intelligence positioned as agentic enterprise control plane (Apr 21 2026); Cortex Code expansion to AWS Glue, Databricks, Postgres
• VLN.microsoft.vision + VLN.microsoft.agentic.why/capabilities/excels/gaps (cloud) — A4 GREEN; Microsoft now second analytics platform GREEN on both A1 and A4
• VLN.databricks.agentic.gaps (cloud) — Lakebase GA Azure (Mar 3, 2026, 14 regions); cloud-portability gap on agent memory closed
• VLN.ibm.agentic.why/capabilities/excels/gaps (onprem) — A6 GREEN with runtime agent monitoring + Agentic Control Plane GA
• VLN.sap.knowledge.capabilities (onprem) — added HANA Cloud Knowledge Graph Engine GA Q1 2025 (RDF triple store, ontology rules)
• VLN.sas.agentic.capabilities (onprem) — added SAS Viya Copilot + Agentic AI Accelerator (Apr 28, 2026)
• STATIC_NARRATIVES.execSummaryTakeaways[2] (cloud) — Microsoft A4 GA + Snowflake A6 GA reflected
• STATIC_NARRATIVES.strategicFindings[2] (cloud) — Agentic stratification updated for Microsoft A4 GA
• STATIC_NARRATIVES.mustAccelerate A4 row (cloud) — benchmark updated to Databricks Supervisor Agent + Microsoft Copilot Studio multi-agent (both GA)
• STATIC_NARRATIVES.competitiveLandscape (cloud) — Snowflake (A6 GREEN), Databricks (Lakebase Azure GA), Microsoft (A4 GREEN) onelines updated
• STATIC_NARRATIVES.fieldGuidance vs.Snowflake/vs.Databricks/vs.Microsoft (cloud) — updated for new GA milestones
• STATIC_NARRATIVES.execSummaryTakeaways[2] + strategicFindings[2] (onprem) — IBM A6 GREEN reflected
• STATIC_NARRATIVES.execSummaryTakeaways[5] + strategicFindings[5] (onprem) — Teradata Autonomous Knowledge Platform announcement reflected
• STATIC_NARRATIVES.competitiveLandscape.ibm (onprem) — A6 GREEN added to oneline

References Added

Cloud:

• Snowflake Product Release Notes — Snowflake, 2025-2026: existing card updated to add Cortex Agent Evaluations GA (Mar 13, 2026) and multi-tenancy GA; informed A6 upgrade.
• Microsoft Copilot Studio Multi-Agent Orchestration GA — Microsoft, 2026: NEW card documenting April 2026 GA of A2A protocol, Fabric integration, and M365 Agents SDK orchestration; informed A4 upgrade.
• Databricks Lakebase GA — Databricks, 2026: existing card updated to reflect Azure GA Mar 3 2026 (14 regions).
• Teradata Autonomous Knowledge Platform Announcement — Teradata, 2026: NEW card documenting May 7, 2026 announcement (AI Studio, Tera workspace, Tera Agents, Pinecone + WisdomAI partnerships, Q3 2026 cloud GA).

On-Prem:

• IBM Product Documentation card updated: added Watson Orchestrate runtime agent monitoring GA, Agentic Control Plane GA (multi-framework: LangFlow, LangGraph, A2A), and watsonx.ai Graph RAG support; informed A6 upgrade.
• SAP Datasphere and SAP Business AI Documentation card updated: added HANA Cloud Knowledge Graph Engine GA Q1 2025, Joule Studio Agent Builder GA Dec 2025, HANA Cloud MCP Server GA Q1 2026.
• SAS Viya Documentation card updated: added Viya MCP Server GA 2025 and SAS Innovate Apr 28, 2026 announcements (Viya Copilot, Agentic AI Accelerator).
• Teradata Autonomous Knowledge Platform Announcement — Teradata, 2026: NEW card (same as cloud).

Partnership Registry additions (Data/partnerships.md)

2 new rows added:

• teradata | Pinecone (K2) — confirmed-documented; production vector retrieval inside AI Studio; details-only update (K2 already GREEN); Pinecone Nexus partner backing alongside Box, Unstructured, LlamaIndex, ThoughtFocus.
• teradata | WisdomAI (T5, A1) — confirmed-documented; agentic BI in AI Studio: NL queries, AI-powered dashboards from a single prompt, automated analytical workflows; documented partnership rule keeps T5 AMBER but updates rationale and details.

[2026-04-27] — Databricks agentic layer reaches GREEN: Agent Bricks GA (A1 + A4)

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-04-27].html

Cloud Assessment

On-Prem Assessment

Narrative Updates

Cloud VLN:

• VLN.databricks.vision — updated: Agent Bricks GA Apr 2026, Supervisor Agent GA Feb 2026; "first analytics platform to reach GREEN on both A1 and A4"
• VLN.databricks.agentic.why — AMBER → GREEN; Databricks completed agentic buildout in early 2026
• VLN.databricks.agentic.capabilities — fully rewritten to reflect GA Agent Bricks + Supervisor Agent
• VLN.databricks.agentic.excels — rewritten to highlight first GREEN on A1+A4 with enterprise proof
• VLN.databricks.agentic.gaps — removed Agent Bricks and Supervisor Agent (now GA); MCP Preview remains
• VLN.teradata.agentic.gaps — urgency note updated: "Databricks Supervisor Agent is now GA (Feb 2026)"
• RATIONALES.A1.databricks — updated to GREEN rationale citing Agent Bricks GA Apr 14, 2026
• RATIONALES.A4.databricks — updated to GREEN rationale citing Supervisor Agent GA Feb 10, 2026
• RATIONALES.A2.databricks — removed stale "A1 rated AMBER" reference
• RATIONALES.A1.teradata — updated to mention Karini AI documented partner + AgentStack cloud Q2 2026 target
• details.A1.teradata — added Karini AI as AgentStack integration partner
• details.A4.teradata — added Karini AI to supported frameworks list

Cloud STATIC_NARRATIVES:

• execSummaryTakeaways[2] — "universally immature, every vendor AMBER" → "maturing fast — Databricks leads with GREEN on A1+A4"
• strategicFindings[2] — "universally immature" → "stratifying — Databricks first GREEN on A1+A4"
• mustAccelerate A4 row — benchmark updated from "Multi-Agent Supervisor (Beta)" to "Databricks Supervisor Agent (GA Feb 2026)"
• competitiveLandscape.databricks — strongest reordered to Agentic / Knowledge; oneline updated to mention first GREEN on A1+A4

Partnership Registry additions (Data/partnerships.md)

3 new rows added:

• teradata | Karini AI — confirmed-documented; no-code agent builder + Teradata MCP + knowledge-grounded workflows; named in AgentStack press release Jan 2026 (applies to both cloud and on-prem)
• databricks | OpenAI — confirmed-documented; $100M multi-year partnership; GPT-5/5.5 natively in Agent Bricks; details-only (A1 already GREEN natively)
• snowflake | Anthropic — confirmed-documented; Cortex Agents optimized for Claude; VS Code + Claude Code plugin Apr 2026; no rating impact

[2026-04-27] — Partnership ecosystem scan: 8 rating changes across cloud + on-prem

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [2026-04-27].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-04-27].html

Skill Enhancement

Added mandatory Partnership & Ecosystem Research to SKILL.md Phase 2: a 4-step process (read registry → build gap matrix → search every row → completeness gate), vendor-specific partnership targets for all 11 non-Dell vendors, universal Partnership Rating Rule in Constraints, Partnership: field in Phase 3 recommendation blocks, and Phase 4d to write Data/partnerships.md after each run. Created Data/partnerships.md as a self-maintaining partnership registry (21 rows written this run).

Cloud Assessment

On-Prem Assessment

Details-only updates (no rating change)

• P2/Databricks: added Arize AI (agent eval + observability, documented) and Aporia (ML observability, documented) to details
• P2/Microsoft: added Arize Azure Native Integration (documented) for tracing and eval to details

No Change (confirmed not-found)

• K4/Snowflake — no Neo4j partnership found; remains RED
• A1/SAS — no LangChain/LlamaIndex partnership; SAS Viya MCP Server is native; remains AMBER
• K1/Oracle — no Unstructured.io partnership; Oracle Object Storage + AI Vision path; remains AMBER

Narrative Updates

Cloud VLN:

• VLN.databricks.knowledge.gaps — removed K4 RED language; added Neo4j certified partnership AMBER framing
• VLN.google.knowledge.gaps — removed K4 RED language; added Neo4j documented partnership AMBER framing
• VLN.aws.vision — updated "entire Translation Layer is RED" to reflect T1 now AMBER via dbt SCA
• VLN.aws.translation.why/capabilities/excels/gaps — full rewrite to reflect T1 AMBER via dbt SCA; T2/T5/T6 remain RED

Cloud STATIC_NARRATIVES:

• competitiveLandscape.aws.oneline — updated; weakest now "Translation" (not "Translation (RED)")
• strategicFindings[1] — "AWS's entire Translation Layer remains RED" → "T1 AMBER via dbt SCA; T2/T5/T6 remain RED"
• execSummaryTakeaways[1] — "AWS has no semantic layer at all" → "AWS improved T1 to AMBER via dbt SCA"

On-Prem VLN:

• VLN.ibm.knowledge.why/excels/gaps — updated K1 GREEN via Unstructured OEM; removed storage-assembly gap language
• VLN.oracle.agentic.why/excels/gaps — updated A1 GREEN via OCI EA Agents GA; removed "OCI GenAI Agents is less mature" language
• VLN.cloudera.agentic.why/capabilities/gaps — updated A3 AMBER via iceberg-mcp-server; added single-component MCP framing

On-Prem STATIC_NARRATIVES:

• execSummaryTakeaways[3] — Oracle finding updated to include OCI EA Agents GA (A1 GREEN)
• strategicFindings[3] — Oracle finding updated to include OCI EA Agents GA

Partnership Registry (Data/partnerships.md)

21 rows written covering: 4 Dell (pre-existing), Teradata/Unstructured.io, IBM/Unstructured.io, Databricks/Neo4j, Google/Neo4j, AWS/dbt Labs, Oracle/LangChain, Cloudera/iceberg-mcp-server, SAP/LangChain, SAP/A2A, Databricks/Arize, Databricks/Aporia, Microsoft/Arize, Snowflake/Neo4j (not-found), SAS/LangChain (not-found), SAS/LlamaIndex (not-found), IBM/LangChain (incidental), Oracle/Unstructured.io (not-found).

[2026-04-27] — On-prem references restructured to three-tier layout

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - On-Premise & Hybrid Platforms [2026-04-27].html

Structural Change — On-Prem References Tab

The on-premise References tab now uses the same three-tier layout as the cloud version:

No rating changes. No content added or removed — existing 15 reference cards redistributed into tiers. Template updated with renderRefCards() helper function matching the cloud template. SKILL.md schema comment updated to reflect unified structure.

[2026-04-27] — April 2026 cloud assessment update (P1 + P2 guardrails corrections)

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [Initial].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [Initial].html

Cloud Assessment

On-Prem Assessment

No Change (verified accurate)

• A3/databricks (cloud): MCP still Public Preview → AMBER confirmed
• A7/snowflake (cloud): No agent memory product → RED confirmed
• K4/google (cloud): Enterprise Knowledge Graph + BigQuery Graph still Preview → RED confirmed
• A4/teradata (cloud): AgentStack private preview, cloud GA targeting Q2 2026 → AMBER confirmed
• IBM A4/on-prem: Watson Orchestrate multi-agent GA → GREEN confirmed
• Oracle A3/on-prem: Oracle DB MCP Server GA → GREEN confirmed
• Oracle A7/on-prem: Oracle 23ai Unified Memory Core GA Jan 2026 → GREEN confirmed
• SAP A1/on-prem: Joule Studio Agent Builder GA Q1 2026 → GREEN confirmed

Narrative Updates

• VLN.snowflake.vision (cloud) — updated RED count from 5 to 4; noted AI guardrails upgraded to AMBER (Cortex Guard GA)
• VLN.snowflake.policy.why/capabilities/excels/gaps (cloud) — replaced "weakest of any major vendor (P2 rated RED)" with AMBER framing; added Cortex Guard GA Jul 2024 and Cortex AI Guardrails to capabilities
• VLN.microsoft.policy.why/capabilities/excels/gaps (cloud) — updated to reflect P1 GREEN: "OneLake security GA Mar 2026 closes the last Preview-stage gap"; removed "OneLake security still in Preview" from gaps
• STATIC_NARRATIVES.competitiveLandscape.snowflake (cloud) — updated oneline from "5 RED ratings (K4, K5, T2, A7, P2)" to "4 RED ratings (K4, K5, T2, A7); AI guardrails upgraded to AMBER"
• STATIC_NARRATIVES.fieldGuidance (cloud) — updated Snowflake bullet from "5 RED ratings: no AI guardrails" to "4 RED ratings; AI guardrails now AMBER"
• VLN.google.knowledge.why/capabilities/excels/gaps (cloud) — updated "Dataplex Universal Catalog" to "Knowledge Catalog (GA Apr 2026, formerly Dataplex)"; added GA data products + high-precision semantic search; noted BigQuery Graph still Preview
• REQUIREMENTS.K7.details.google (cloud) — same Knowledge Catalog rename and GA status

References Added

(No new reference cards added this run — evidence sourced from official Snowflake release notes (Jul 2024), Microsoft Fabric blog (Mar 2026), and Google Cloud Next 2026 blog (Apr 2026), all within existing source categories.)

[2026-04-27] — Q2 2026 agentic layer and semantic layer update

🔴🟡 Red/Amber update

Files: Knowledge Platform Comparison - Cloud [Initial].html · Knowledge Platform Comparison - On-Premise & Hybrid Platforms [Initial].html

Cloud Assessment

On-Prem Assessment

No Change (verified accurate)

• A3/teradata (cloud): Community Edition MCP only; no enterprise GA — stays A
• A3/databricks (cloud): Public Preview Jan 2026 — stays A
• A7/snowflake (cloud): No native agent memory product — stays R
• K4/google (cloud): Enterprise Knowledge Graph still Pre-GA April 2026 — stays R
• K7/dell (onprem): DOE adds discovery but unified catalog gap persists — stays A
• T1/dell, T5/dell (onprem): AI Assistant targeted 1H CY26, not yet GA — stays A
• A4/sas, A4/cloudera (onprem): No multi-agent products in portfolio — stays R

References Added

Cloud:

• Snowflake Product Release Notes — Snowflake, 2025: GA dates for Semantic Views (Oct 2025), Cortex Agents (Nov 2025), MCP Server (Nov 2025). Informed T1, A1, A3 Snowflake ratings.
• Databricks Unity Catalog Business Semantics GA — Databricks, 2026: GA announcement blog for Metric Views / UC Business Semantics. Informed T1 Databricks upgrade (A→G).
• Databricks Lakebase GA — Databricks, 2026: GA announcement blog for managed Postgres agent memory. Informed A7 Databricks upgrade (R→G).

On-Prem:

• Dell AI Factory documentation card updated: added Data Orchestration Engine GA Q1 2026 (GTC Mar 16 2026). Informed K6 Dell upgrade (A→G).
• MCP card updated: SAP entry revised to reflect HANA Cloud MCP Server GA Q1 2026 + Joule Studio MCP GA Dec 2025. Informed A3 SAP upgrade (A→G).