Kyndryl launches service to oversee AI agents and improve return on investment

Kyndryl is launching a new service for companies deploying AI agents into their workflows. According to CEO Martin Schroeter, businesses need not just access to AI, but a clear control mechanism that ensures such systems deliver measurable returns.

Kyndryl's New Service

The company is bringing to market a solution designed to help customers manage AI agents and better justify investments in this technology. The emphasis itself is telling: the corporate market is gradually shifting from experiments with individual models to infrastructure around them. The question is no longer just whether an agent can complete a task, but how to integrate it into existing processes, who is responsible for results, and how to track work quality at scale.

For Kyndryl, this is a logical continuation of its course toward corporate services in complex IT environments. In large companies, AI agents can help with user support, processing internal requests, document handling, and automating routine operations. But once such tools move beyond pilot stage, the need emerges for a separate management layer: without it, even a useful scenario quickly runs into risks, costs, and chaos between teams.

This is when IT directors typically start asking not for another pilot, but for an industrial operational model.

Why Control Matters

An AI agent differs from an ordinary chatbot in that it not only answers questions but also acts: it reads data, launches steps in applications, passes results down the chain. Because of this, the cost of error is higher. If a company doesn't understand what exactly the agent does, what data it has access to, and when it should hand a task to a human, the promised automation easily turns into a new source of problems. This is why Schroeter's thesis about proper control sounds like a management requirement, not merely a technical one. In practice, business usually has to address several issues at once:

• agent access rights and action boundaries
• step logging and clear audit trails
• cost of each scenario and spending on models
• error frequency and share of tasks requiring human intervention
• unified rules for multiple teams and vendors

A strong model rarely becomes a corporate product on its own. It needs an observability loop, accountability, and metrics—otherwise managers see a beautiful demo scene, but not a real production tool. If Kyndryl's new service closes exactly this layer, the company hits one of the most sensitive demands of the corporate market.

Where to Count Returns

The second key emphasis is return on investment. For companies already spending money on licenses, integrations, and model setup, the question of payback becomes paramount. Executives want to understand not abstract transformation, but concrete metrics: how much time the agent saves, how many tasks it closes without escalation, how much employee workload it reduces, and what one successfully completed process costs.

Without this, an AI program quickly becomes a collection of disconnected initiatives with unclear impact. This is why agent management services are becoming a separate market category. They are needed not by those just trying AI as an experiment, but by those who have reached the scaling stage and now seek discipline in their operational model.

For such customers, what matters is not the most impressive model presentation, but the ability to gather in one place rules, monitoring, benefit assessment, and clear accountability for results. This is especially critical where the agent affects customer service, finances, or internal approvals.

What This Means

Kyndryl's launch shows that the AI market is gradually maturing: attention is shifting from demonstrations of capabilities to control, security, and the economics of implementation. The next stage of competition in the corporate segment will unfold not only around the models themselves, but also around services that help manage agents like a regular corporate system—with access rights, metrics, and clear business value.