Kyndryl looks at why AI struggles to move beyond pilots in enterprise IT

Kyndryl is introducing a structured approach to help companies shift from manual IT operations to AI-driven workflows.

A gap is forming between what AI systems can accomplish and what many IT environments are built to handle. While companies are investing more in AI, many are still struggling to see clear results. In many cases, older systems and processes were designed for human-led work rather than autonomous software agents.

Kyndryl has introduced a new approach called Agentic Service Management to address this issue. The model is designed to help companies move from manual service operations to setups where AI agents can take on tasks with some independence. It combines assessments, structured plans, and governance checks to guide that shift.

Data from the company's Readiness Report shows that more than two-thirds of organizations are putting money into AI. Still, close to half say they are not seeing strong returns. Many are working with workflows and controls that were built before AI became common. This can leave advanced tools stuck in testing instead of being used more widely.

Kris Lovejoy, Global Head of Strategy at Kyndryl, said many enterprise environments were built around people managing tickets and tools, not systems where autonomous agents carry out tasks across hybrid and multi-cloud setups. She pointed to this mismatch as a key reason AI often fails to move beyond pilot stages. Lovejoy further stated that companies cannot scale these workflows on operating models designed for manual work, and that clear controls, repeatable practices, and defined stages of adoption are needed.

At the same time, she noted that while AI agents can act on their own in some cases, people still need to remain accountable for governance, risk, and service outcomes.

Building a path to agentic service management

The company's framework is built around a maturity model. It looks at how ready an organization is to support AI-driven service operations and where gaps exist. These gaps can include weak governance rules, unclear workflows, or limited security controls. The aim is to give companies a clearer path to move from early AI use to more stable, scaled deployments.

This work is delivered through Kyndryl Consult, which runs assessments based on existing policies and systems. The process compares current practices with emerging standards, including ISO 42001, which focuses on AI management. Companies then receive a gap analysis and a phased plan. The plan outlines how to strengthen controls while allowing AI agents to take on more work over time, with human oversight still in place.

Alongside this, Kyndryl offers a separate service called Agentic AI Digital Trust. This service focuses on governance and risk. It is meant to help organizations manage how AI agents operate, especially in sectors where data protection and compliance rules are strict. The framework places security at the center, with the aim of reducing risks as AI systems scale across different environments.

Applying agentic AI in practice

Kyndryl is also applying this approach within its own operations. By using agentic methods in its service delivery, the company is changing how it manages IT systems for clients. Some of these capabilities are available through its platform, Kyndryl Bridge. The platform supports teams in monitoring systems, making decisions, and managing complex environments with input from AI-driven insights.

The company's current automation setup already handles a large number of tasks. It runs close to 200 million automated actions each month, supported by thousands of predefined workflows. Agentic Service Management builds on this base, with more focus on how AI agents and human teams work together.

The broader aim is to close the gap between what AI can do and what enterprise systems can support today. Without updates to governance, workflows, and controls, many organizations may continue to struggle to move beyond early-stage AI use.