Kore.ai launches Autoloop to optimise enterprise agents

Kore.ai has launched Autoloop, an optimisation engine for its Agent Platform, now generally available to customers using the Artemis edition.

Autoloop is designed to build AI agents, measure them against business goals, and keep adjusting them from initial set-up through live deployment. It focuses on targets set by enterprise users rather than prompt-based tuning.

Kore.ai presents the release as a response to the challenges businesses face in maintaining AI agents manually. In the company's 2026 Agent Productivity Index, 79% of enterprises said they had reversed an action taken by an AI agent, while 70% had dealt with a failure their teams could not trace.

Autoloop measures agents across several criteria, including task completion, accuracy, adherence to business rules, safety, consistency, end-user experience, and cost efficiency. It runs a loop covering build, evaluation, diagnosis, optimisation, and re-verification.

Before deployment, the tool uses an enterprise's operating procedures to create an agent and its test coverage, then iterates until targets are met. In production, live interactions trigger new optimisation cycles, with each change assessed against all goals at the same time.

Kore.ai said this is intended to prevent gains in one area from undermining another, such as reducing token use at the expense of task completion or safety.

Tracing activity

The system depends on two internal technologies: StateTrace and Agent Blueprint Language, or ABL. StateTrace follows execution context across an agent network, including hand-offs, delegations, tool calls, and state changes, allowing Kore.ai to identify where a goal was missed.

The second component, ABL, compiles supervision rules, routing, hand-offs, delegation, tools, business rules, and guardrails into an executable state machine. Kore.ai said this allows Autoloop to change the specific part of an agent workflow that caused a problem without affecting unrelated steps.

Most checks are also made deterministic through a five-layer validation architecture, which Kore.ai argues keeps continuous optimisation affordable at large scale.

Raj Koneru, Founder and Chief Executive Officer of Kore.ai, set out the company's view of the market need for this kind of system. "Every enterprise knows what it wants from its agents: finish the job, follow the rules, stay safe, and do it at a sensible cost," he said.

"With Autoloop, your agents keep improving against the goals you set. The companies that scale AI will be the ones using AI to build, govern, and optimize AI," Koneru said.

Internal use

Kore.ai said it applies the same method in its own engineering organisation. According to the company, AI agents now generate about 6,500 commits a month across a production codebase of 2.6 million lines, under 68 always-on guardrails.

This internal use supports Kore.ai's argument that automatic optimisation requires both visibility into what agents are doing and a controlled way to change their behaviour. The release also extends the company's broader strategy of offering a single environment for building, deploying, managing, and optimising enterprise AI agents.

Prasanna Arikala, Chief Technology Officer and Chief Product Officer of Kore.ai, linked the product to those underlying systems. "You can't optimize what you can't see, or fix precisely what you can't express precisely," he said.

"StateTrace lets Autoloop see exactly what agents did, and ABL allows it to change anything that needs changing. These are the technologies that make automatic optimization a reality," Arikala said.