Kore.ai has announced the general availability of Autoloop, an optimisation engine for its Agent Platform, Artemis edition, designed to help enterprises continuously improve the performance of AI agents.
The platform automates the process of building, evaluating, diagnosing, and refining AI agents against business-defined objectives. It aims to reduce reliance on manual fixes by enabling agents to improve continuously while balancing task completion, accuracy, compliance, safety, user experience, and operating costs.
The launch comes as enterprises face challenges in managing AI agents at scale. According to the 2026 Kore.ai Agent Productivity Index, 79% of enterprises have reversed an action taken by an AI agent, while 70% have encountered failures that their teams could not trace. Gartner predicts that autonomous learning techniques will feature in a majority of AI agents by 2030, compared with less than 5% in 2026.
Raj Koneru, Founder and CEO of Kore.ai, said enterprises need AI agents that can complete tasks, follow business rules, operate safely and manage costs. He added that Autoloop is designed to help agents improve against the objectives defined by their organisations.
Continuous optimisation across the AI lifecycle
Autoloop evaluates agents across seven dimensions: task completion, accuracy and grounding, business-rule adherence, safety guardrails, behavioural consistency, end-user experience and cost efficiency.
Before deployment, the engine builds agents and generates test coverage based on an organisation’s operating procedures, refining performance until the defined goals are met. Once deployed, real-world interactions trigger further optimisation cycles.
Each proposed change is evaluated against all defined objectives to ensure that improvements in one area, such as reducing token consumption, do not compromise other requirements, including task completion or safety.
StateTrace and ABL underpin optimisation
Kore.ai has incorporated two technologies to support the optimisation process: StateTrace and Agent Blueprint Language (ABL).
StateTrace tracks agent execution, including handoffs, delegation, state changes, tool calls and contextual information. This visibility helps identify where an agent has failed to meet its objectives and why. The company said its patent-pending, five-layer validation architecture is designed to make continuous evaluation more cost-efficient at enterprise scale.
ABL converts elements such as routing, supervision, business rules, tool usage, delegation, and safety guardrails into an executable state machine. This allows Autoloop to map failures to specific components and modify the relevant sections without unnecessarily changing other parts of an agent’s workflow.
Prasanna Arikala, Chief Technology Officer and Chief Product Officer at Kore.ai, said StateTrace provides visibility into agent behaviour, while ABL enables precise changes to agent workflows, supporting automated optimisation.
Kore.ai also cited its internal engineering operations as an example of AI governance at scale. According to the company, AI agents generate approximately 6,500 code commits each month across a production codebase of 2.6 million lines, with 68 continuously enforced guardrails.
With Autoloop, Kore.ai aims to extend its enterprise AI platform beyond agent creation and deployment to continuous performance optimisation. The company positions the engine as part of an integrated approach in which agents are built, governed, managed, and improved within a common platform.