Enterprise

Dataiku's new Agent Management scans every platform for AI agents — and asks which ones actually earn their cost

A cross-platform inventory over Salesforce Agentforce, AWS Bedrock, Google Vertex, Microsoft Copilot Studio, Databricks and Snowflake, with business-KPI scoring on each agent. GA is set for October.

By Asher Glick· Enterprise software desk·2026-09-26·3 min read

Photo: Unsplash / Carlos Muza — Laptop screen showing analytics dashboard with charts and performance metrics

Dataiku unveiled Agent Management at its Succeed conference in New York on September 24, pitching a cross-platform inventory that scans for AI agents running on Salesforce Agentforce, AWS Bedrock, Databricks Agents, Google Vertex, Microsoft Copilot Studio, Azure Foundry, and Snowflake Cortex, plus anything custom pipelined through OpenTelemetry. General availability is set for October. The product scores each agent against business KPIs, not just latency and token cost, and flags agents by risk and cost-effectiveness. Access is priced per instance annually; monitoring is metered per agent.

The framing quote from co-founder and CEO Florian Douetteau is the whole pitch: “Ask a bank how many servers it runs, and you get an answer to the decimal. Ask how many AI agents it’s running, and you get a shrug or a guess.”

The numbers around the launch tell an interesting story about elite self-perception. IBM’s “AI in Motion” report says fewer than one in five organizations maintain a complete, current inventory of their AI systems. In a CIO cohort Dataiku surveyed alongside the launch, nine in 10 respondents said they were confident they had complete tracking of their agents. Then 81% admitted they lacked oversight of agents built outside approved systems or formal channels. The confidence and the reality are in different rooms.

That gap is landing at a bad moment for the buyers. SiliconANGLE reports 72% of respondents expect cuts or a freeze if this year’s performance targets are missed. Agent budgets are among the newest and softest line items on the sheet.

The category Dataiku is trying to define, per-agent business-KPI scoring across vendors, is what the current agent boom has quietly needed. Small teams running a Salesforce Agentforce outbound sales agent alongside a HubSpot Breeze prospecting campaign don’t get an honest apples-to-apples read from either vendor’s own dashboard. Every telemetry surface today is single-vendor by design, and every vendor grades its own homework.

Whether Dataiku becomes the neutral scorekeeper or whether the hyperscalers absorb the function themselves is the structural question. The IBM figure suggests the incumbents haven’t earned the benefit of the doubt yet.

Sources