.webp)
Kore.ai has launched Autoloop, an optimisation engine for enterprise AI agents. The product is available to customers using the Artemis edition of the Kore.ai Agent Platform.
The software is designed to improve AI agents in production by measuring them against business goals such as task completion, accuracy, safety, customer experience, and cost. It then identifies where performance falls short, makes changes automatically, and rechecks the outcome.
The launch comes as Kore.ai's own research highlights operational problems in enterprise AI deployments. In its 2026 Agent Productivity Index, 79% of enterprises said they had reversed an action taken by an AI agent, while 70% reported a failure their teams could not trace.
Many organisations still maintain agents manually, often reviewing interactions one transcript at a time. Kore.ai argues that this can create knock-on problems when one fix affects another part of the system.
How it works
Autoloop begins by asking teams to define the outcomes they want from an agent. These goals can include task completion, whether answers are grounded in company data, compliance with business rules, adherence to guardrails, consistency across channels and languages, the smoothness of the interaction, and overall cost.
It then runs what Kore.ai describes as a continuous loop of building, evaluating, diagnosing, optimising, and verifying against those measures. Before deployment, the system uses an enterprise's operating procedures to create the agent and its test coverage, then iterates until targets are met. Once the agent is in production, live interactions trigger further optimisation cycles.
Each change is scored across all selected goals at the same time. The aim is to reduce the risk that an improvement in one area, such as lower token use, harms another, such as safety or task completion.
Two elements sit behind the system. One is StateTrace, which tracks an agent's production execution context, including hand-offs, delegation, state, tool calls, and the surrounding context across an agent network. The other is Agent Blueprint Language, or ABL, which maps those actions back to an executable state machine so changes can be made at the specific point where a problem occurred.
Kore.ai says a five-layer validation architecture makes most checks deterministic, helping keep ongoing optimisation affordable at enterprise scale.
Executive view
Raj Koneru, Founder and Chief Executive Officer of Kore.ai, said the product is centred on business-defined targets for AI systems.
"Every enterprise knows what it wants from its agents: finish the job, follow the rules, stay safe, and do it at a sensible cost," Koneru said.
"With Autoloop, you set the goals and the agents keep getting better against them. The companies that scale AI will be the ones using AI to build, govern, and optimize AI," he said.
Prasanna Arikala, Chief Technology Officer and Chief Product Officer at Kore.ai, linked the product to the company's tracing and blueprint tools.
"You can't optimize what you can't see, or fix precisely what you can't express precisely," Arikala said.
"StateTrace lets Autoloop see exactly what agents did, and ABL lets it change exactly what needs changing. That's what lets optimization run automatically instead of by hand," he said.
Market backdrop
Kore.ai positioned the launch against a broader shift in enterprise AI management. It cited a Gartner forecast that autonomous learning techniques will be present in a majority of AI agents by 2030, up from less than 5% in 2026.
The company also pointed to its own internal engineering operation as an example of the governed AI development it supports. AI agents now produce about 6,500 commits a month on a production codebase of 2.6 million lines, overseen by 68 always-on guardrails, according to Kore.ai.
Kore.ai says it serves more than 500 Global 2000 organisations and has focused its platform strategy on combining agent building, deployment, management, and optimisation in a single layer. Autoloop extends that approach by adding continuous optimisation tied to the same governance structure used to create and run the agents.