Dataiku — Agent Management: standalone cross-platform agent inventory/certification product, GA moved to October 2026 (product page + Forkast analysis, September 2026)
Tag: S-2026-09-12-dataiku-agent-management-ga-slip Type: article (vendor product page, undated, state as at 15 Sep 2026; plus Forkast analysis, 12 Sep 2026 — both fetched in full) Author(s): Dataiku (product page); Dana Ellison, “Forkast mind” — an AI-authored correspondent per the site’s own labelling (analysis) Date of source: 2026-09-12 (Forkast publication; product page undated, observed 2026-09-15) Date ingested: 2026-09-15 Authority weight: low — the primary is an undated marketing page for a pre-GA product; the dated secondary is AI-authored commentary whose historical claims (9 Mar launch; September→October GA slip) were not independently verified against the March release or May blog Raw file: /_raw_sources/S-2026-09-12-dataiku-agent-management-ga-slip.md
What it claims
Dataiku’s product page presents Agent Management as “a standalone product that gives an organization one place to discover, monitor, and manage the risk of all its AI agents, wherever they were built”. Three functions are described: discover — scan agents across platforms into one inventory, “each agent with an owner and a purpose”; monitor — track usage, cost and quality against each agent’s purpose and “catch drift before users report it”, measuring “each agent as an agent” rather than the infrastructure it runs on; and control — “certify which agents are cleared to run, assess each against named risks and their mitigations, and re-run tests on a schedule”, so that “when an auditor asks, the record is already there to export”. Connected platforms are Microsoft Copilot Studio and Azure Foundry, Salesforce Agentforce, AWS Bedrock, Google Vertex, Databricks, Snowflake Cortex, n8n and Dataiku. The FAQ states it is not an AI gateway (“a gateway sits between agents and their tools and controls traffic. Agent Management sits above the agents and keeps track of them”), that it connects “through native APIs and log streams” with depth that “depends on what each platform exposes”, and that the product “will be available in October 2026”.
The Forkast analysis (12 Sep) says Agent Management was announced on 9 March 2026 as part of Dataiku’s “Platform for AI Success” launch and does not require the Dataiku platform — “a meaningful departure” for an integrated data-science vendor. It frames Dataiku’s bet as cross-platform oversight versus platform-native governance (Salesforce Einstein Trust Layer; Microsoft Agent 365 + Purview), which it calls “deep but proprietary” and “silent outside” their own walls. It flags that “Dataiku’s initial press release pointed to a September 2026 general availability date. The current product page lists October 2026”, and that a May 2026 Dataiku blog “references a September launch with early access available now” — advising buyers to “verify the GA timeline directly with Dataiku before mapping it into their own deployment plans”. It places the product in a cluster of four standalone agent-governance launches between 24 Aug and early Sep (Okta Agent SSO, IBM watsonx Orchestrate AgentOps Agent, Broadcom AgentMinder, Dataiku), and raises “governance tool sprawl — managing the tools that manage the agents” as an emerging problem.
Notable quotes
- “Certify which agents are cleared to run, assess each against named risks and their mitigations, and re-run tests on a schedule. When an auditor asks, the record is already there to export.” (product page, control section)
- “The depth of integration depends on what each platform exposes, with deeper integrations actively being developed.” (product page, FAQ)
- “Dataiku Agent Management will be available in October 2026.” (product page, FAQ)
- “For a product whose pitch is reliability and auditability, the moving date is a minor signal, but it is worth flagging.” (Forkast, para 8)
What’s speculative vs. asserted
Asserted (vendor statements about a future product): standalone; nine connected platforms; inventory with owner/purpose; usage/cost/quality/value tracking; certification, named-risk assessment, scheduled re-testing, exportable record; not a gateway; October 2026 availability.
Asserted by the secondary (unverified here): 9 Mar 2026 announcement; original September GA; May blog early-access reference.
Vendor marketing / unverified: “ready for auditors”; drift detection; any capability, since the product is pre-GA.
Not stated (gaps): any regulatory standard; any customer; what “certify” means procedurally (who signs, against what criteria); the schema of the exported “risk record”; pricing; the relationship to Dataiku Govern (the watchlist’s name for Dataiku’s existing governance module).
Vault inference (not a source claim): the inventory + certification + scheduled re-test + exportable record set maps onto EU AI Act inventory/record-keeping, ISO/IEC 42001 asset and risk-treatment records and SS1/23 / SR 11-7 model-inventory expectations — but as vendor design intent for an unreleased product.
Topics this feeds
- AI Governance Platforms — the fourth product in the Aug–Sep “standalone agent governance” cluster; the inventory/certification (rather than identity, orchestration or runtime) layer; first GA-date slip recorded for an agent-governance product.
- Dataiku — company page created this run (second source after Cobuild).
Open questions raised
- What did the March 2026 release actually state for GA, and why did the date move — capacity, integration depth, or demand? (Neither source says; the March release was not fetched.)
- If the record is exportable “when an auditor asks”, in what format, and does it carry the test evidence or only the certification decision?
- How does Agent Management relate to Dataiku Govern and to the LLM Mesh gateway for Dataiku-native agents — one governance record or two?
- With four vendors shipping standalone agent-governance layers in two weeks (Forkast’s cluster), which layer’s inventory is authoritative for a regulated firm’s AI-system register?