This page is illustrative and may reference vendors for comparison. The baseline itself remains vendor‑neutral.

📊 Ratings — Current Platforms (UKM v2)

Platform Overall UKM (0–4) Gating Notes
Salesforce – Agentforce (OpsAI) 4 Typed corpora in Data Cloud; governed retrieval with citations; knowledge graph linking; profile‑aware surfacing.
ServiceNow – AI Agent Orchestrator 4 CSDM/Service Graph + KB; governed retrieval; citations; versioned ontology & query packs.
Databricks – Agent “Bricks” 4 Lakehouse tables + vector functions; lineage graph; governed RAG; profile/context aware.
Dataiku – AI Agents 4 Schema‑validated ingest; knowledge bank + vector; lineage/registry; governed retrieval with feedback.
IBM – AskIAM 3 Typed corpora (policies, roles, access); identity graph; governed retrieval; limited diagram/story composition.
Opsera – Hummingbird 3 Unified DevOps insights model; hybrid index + traceability graph; governed retrieval; citations on summaries.
AWS – Strands SDK 2 BYO storage/index; vector search optional; provenance via tool traces; no native graph/governance → cap 2.
Google – Conversational Agents 1 KB/FAQ index; minimal provenance; no vector/graph in base product → L1.
GitHub – Copilot 1 Ephemeral repo context; no persistent KB or graph; no citations → L1.
Termius – Gloria 1 Live host responses only; no ingest/index; no governance beyond auth → L1.

Caps applied per gates: lack of exportable citations, RDF/SHACL (or typed cross‑corpus retrieval), or enforceable visibility ⇒ max L2.

🧾 Platform Notes — Evidence‑Aligned (per UKM phase)

Salesforce – Agentforce (OpsAI) — UKM 4

  • Ingestion & Validation: Connectors into Data Cloud ingest KB, tickets, docs with schema validation; provenance captured (source, owner, revision).
  • Normalization & Enrichment: Topics ontology and entity tagging (service/team); PII redaction and version metadata; visibility by account/role.
  • Indexing & Linking: Hybrid keyword+vector index; links to CMDB/graph entities; shapes (schema) drive embeddings; cross‑record linking of cases/incidents/runbooks.
  • Retrieval & Composition: Typed retrieval (capability/knowledge/story); RAG with citations/snippets; composes “action plans” and Mermaid‑style diagrams for processes.
  • Governance & Evolution: Policy‑enforced visibility; per‑record provenance in answers; freshness monitoring; feedback loops refine results.

ServiceNow – AI Agent Orchestrator — UKM 4

  • Ingestion & Validation: Ingests KB articles, incidents, changes; validates against CSDM; dedupe via KB workflows; provenance/version retained.
  • Normalization & Enrichment: CSDM/Service Graph adds owner, service, env; PII masking; alias/version policy for KB.
  • Indexing & Linking: KB + semantic index; deep linking into CMDB/graph; query library mapped to service context.
  • Retrieval & Composition: Typed retrieval across KB/incidents/runbooks; citations on answers; diagram assembly for service runbooks.
  • Governance & Evolution: Strict RBAC by assignment groups/tenants; audit trails; feedback & recency windows; managed deprecation lifecycle.

Databricks – Agent “Bricks” — UKM 4

  • Ingestion & Validation: Structured ingest into Lakehouse tables; MLflow/Delta lineage; schema checks and dedupe.
  • Normalization & Enrichment: Unity Catalog entities, ownership, tags; PII policies; versioned docs and profiles.
  • Indexing & Linking: Vector functions + metadata in Delta; lineage graph links datasets↔models↔agents; shapes→vectors via table schemas.
  • Retrieval & Composition: Governed RAG on trusted data; profile‑aware answers; composes code/SQL/diagrams with citations.
  • Governance & Evolution: Central governance (ACLs, audits), quality dashboards; freshness SLOs; continuous evals & feedback.

Dataiku – AI Agents — UKM 4

  • Ingestion & Validation: Multi‑connector ingest with schema validation; DLQ and data quality checks; provenance fields on each item.
  • Normalization & Enrichment: Owner/visibility/entity tags; LLM Mesh enrichment; versions and aliasing.
  • Indexing & Linking: Knowledge bank with vector + metadata; lineage graph across datasets/flows/models; SHACL‑like shape checks where applicable.
  • Retrieval & Composition: Typed retrieval (capability/knowledge/story/profile); RAG with citations; agent can assemble Mermaid diagrams and narratives.
  • Governance & Evolution: Guardrails and permissions centrally enforced; feedback traces; freshness windows and dashboards.

IBM – AskIAM — UKM 3

  • Ingestion & Validation: Structured ingest from IAM/HR systems (roles, entitlements, policies); validation on identity schemas; full provenance.
  • Normalization & Enrichment: Unifies identities, roles, apps; adds owner/visibility; PII redaction by policy.
  • Indexing & Linking: Identity/entitlement graph with linkable nodes; hybrid search (attributes + text); query packs for audits.
  • Retrieval & Composition: Typed retrieval (policy/knowledge/profile); cites back to access logs/records; limited diagram/story assembly.
  • Governance & Evolution: Strong RBAC/SoD enforcement; query auditability; feedback into certification cycles; high governance pass rate.

Opsera – Hummingbird — UKM 3

  • Ingestion & Validation: Pulls CI/CD, repos, tickets; validates against a DevOps insights schema; dedupe of runs and issues.
  • Normalization & Enrichment: Normalizes entities (service/pipeline/env); tags DORA metrics; versions by run/build.
  • Indexing & Linking: Hybrid index + traceability graph linking commit→build→deploy→incident; query library.
  • Retrieval & Composition: Typed retrieval across insights and KB; summaries with citations; diagrams for pipelines available.
  • Governance & Evolution: RBAC by team/project; audit logs; feedback loops for recommendation quality.

AWS – Strands SDK — UKM 2

  • Ingestion & Validation: No managed KB ingest; developers bring sources; agent tool traces provide minimal provenance.
  • Normalization & Enrichment: Light normalization via standard tool contracts; optional entity tags.
  • Indexing & Linking: BYO vector/metadata store; no native RDF/SHACL or tenancy graph → graph/governance gate caps at 2.
  • Retrieval & Composition: On‑demand tool retrieval; basic snippet assembly; citations optional if implemented by developer.
  • Governance & Evolution: IAM‑scoped access; developer‑managed visibility; limited platform‑level governance metrics.

Google – Conversational Agents — UKM 1

  • Ingestion & Validation: Manual KB/FAQ uploads; minimal schema validation; limited dedupe.
  • Normalization & Enrichment: Basic tags/intents; little entity mapping; optional PII redaction.
  • Indexing & Linking: Keyword/FAQ index; no vector by default; no graph.
  • Retrieval & Composition: Template/LLM responses; no guaranteed citations; single‑corpus retrieval.
  • Governance & Evolution: Basic role separation; conversation analytics; limited visibility enforcement → cap 1.

GitHub – Copilot — UKM 1

  • Ingestion & Validation: Ephemeral repo/PR context only; no KB ingest.
  • Normalization & Enrichment: None beyond repo metadata.
  • Indexing & Linking: No persistent index/graph.
  • Retrieval & Composition: Generates code/snippets; no citations/provenance.
  • Governance & Evolution: Org policy for usage; no KB governance → L1.

Termius – Gloria — UKM 1

  • Ingestion & Validation: No KB ingest; operates on live SSH outputs.
  • Normalization & Enrichment: Minimal tagging; no ontology.
  • Indexing & Linking: No persistent index/graph.
  • Retrieval & Composition: On‑the‑fly Q&A; no citations.
  • Governance & Evolution: Key‑based auth; minimal retrieval governance → L1.

📌 Summary & Guidance

  • Leaders (UKM 4): Salesforce Agentforce, ServiceNow AI Agent Orchestrator, Databricks Agent Bricks, Dataiku AI Agents — all demonstrate typed corpora, hybrid vector+keyword search, governed retrieval with citations, and graph/lineage linking, with feedback/freshness loops.
  • Strong (UKM 3): IBM AskIAM, Opsera Hummingbird — domain‑specific graphs and governed retrieval, but limited story/diagram composition or cross‑domain reasoning keep them at L3.
  • Builders (UKM 2): AWS Strands SDK — flexible but BYO for index/graph/governance, so gate‑capped at L2 absent formal exportable citations and tenancy controls.
  • Foundational (UKM 1): Google Conversational Agents, GitHub Copilot, Termius Gloria — keyword or ephemeral context only; no exportable citations or graphs.

Common gating failures that cap vendors at ≤ L2:
1) No exportable citations/provenance with results.
2) No RDF/SHACL graph or lack of typed cross‑corpus retrieval.
3) No demonstrable visibility/tenant enforcement under test.

Evidence to collect for bake‑offs: retrieval logs with citations, schema/shape files, visibility test scripts (pass/fail), freshness dashboards, and any ontology/graph query packs.