Kore.ai, a global provider of enterprise AI platforms and agentic applications, today announced the general availability of Autoloop, the optimisation engine of the Kore.ai Agent Platform, Artemis edition. Autoloop enables enterprises to build, evaluate and continuously optimise AI agents against business-defined goals, from initial development through production.
As enterprises scale AI adoption, maintaining agent performance through manual fixes is becoming increasingly challenging. The 2026 Kore.ai Agent Productivity Index found that 79% of enterprises have reversed an action taken by an AI agent, while 70% have experienced a failure their teams could not trace. Gartner predicts that autonomous learning techniques will be present in the majority of AI agents by 2030, up from less than 5% in 2026.
“Every enterprise knows what it wants from its agents: finish the job, follow the rules, stay safe, and do it at a sensible cost,” said Raj Koneru, Founder and Chief Executive Officer of Kore.ai. “With Autoloop, agents keep improving against the goals businesses set. The companies that scale AI will be those using AI to build, govern and optimise AI.”
Continuous optimisation, from development to production
Autoloop starts with business goals rather than prompts. It optimises agents across seven dimensions: task completion, accuracy and grounding, business-rule adherence, safety and guardrails, behavioural consistency, end-user experience and cost efficiency.
Its continuous optimisation cycle covers five stages: build, evaluate, diagnose, optimise and re-verify. Before deployment, Autoloop builds agents and test coverage from enterprise operating procedures, iterating until defined goals are met. In production, real interactions trigger further optimisation cycles.
Every proposed change is evaluated against all defined goals, helping prevent improvements in one area, such as cost, from compromising another, such as task completion or safety.
Two Kore.ai technologies underpin this approach:
- StateTrace provides visibility into agent execution, tracing hand-offs, delegation, state, tool calls and context across agent networks to identify where and why goals are missed. Its patent-pending, five-layer validation architecture is designed to make continuous optimisation more cost-effective at enterprise scale.
- Agent Blueprint Language (ABL) translates supervision, routing, hand-offs, delegation, tools, business rules and guardrails into an executable state machine. By mapping execution traces to the blueprint, Autoloop can target the specific components responsible for failures.
Kore.ai also applies this approach internally. Its AI agents produce approximately 6,500 commits per month across a production codebase of 2.6 million lines, governed by 68 always-on guardrails.
“You can’t optimise what you can’t see, or fix precisely what you can’t express precisely,” said Prasanna Arikala, Chief Technology Officer and Chief Product Officer at Kore.ai. “StateTrace lets Autoloop see exactly what agents did, and ABL allows it to change what needs changing. These technologies make automatic optimisation a reality.”
With Autoloop, Kore.ai extends its integrated approach to enterprise AI, bringing agent development, deployment, management, governance and optimisation together within a single platform. The result is a continuous, evidence-based approach to improving agent performance before and after deployment.