An AI-SRE vendor bets the hard part of production agents is keeping them grounded

Komodor, the AI-SRE vendor backed by $90 million in venture funding and used inside Fortune 500 shops including Nebius, Dell, Cisco and Prudential, launched its Agentic Operations Platform on September 16. It combines more than 50 pre-built "specialist" agents spanning AI SRE, AI software operations and cost optimization with a shared backbone: persistent memory, context management, cost controls, security guardrails and audit trails. A platform team can run those agents as-is or build and govern custom ones on the same infrastructure. Co-founder and CTO Itiel Shwartz named the bottleneck: "The hardest part of running agentic operations in production is not building the agents. It is keeping every agent grounded in the right context."

Gartner says governance kills these projects

The launch leans on Gartner's projection that more than 40% of agentic AI initiatives will be decommissioned by 2027 over governance gaps rather than because the agents don't work. That claim is sharper than the usual AI-project failure rate because it places the failure after a successful pilot. An agent that works in a demo but has no persistent memory, no cost ceiling and no audit trail gets shut off once it faces production traffic and a CFO.

Why it matters

This is one vendor's press release, so read its adoption numbers accordingly. The product's shape still shows where enterprise AI spend is getting stuck this year. The question has moved from "can an agent do the task" to "what is it costing, is it still doing the right thing six months in, and who signed off on it." A cost-optimization agent that needs its own cost ceiling and audit trail is an admission that agentic tooling is now a line item to track in its own right.