Having spent many years leading business growth in manufacturing, and transforming IT operations, I see changes organizations are beginning to experience as enterprises introduce AI agents into their day-to-day processes. Rather than viewing AI as a replacement for human effort, we should see it as an opportunity to rethink how we structure work — a perspective gained first-hand while helping enterprises implement agentic AI. It is also reflective of what we see through our own research at Kyndryl: 61% of global organizations are redesigning roles to work better with AI.
Today, enterprise IT is facing growing workforce and technology challenges. As experienced engineers retire or move on, organizations face the challenge of preserving the knowledge and experience that keep critical systems running. At the same time, cloud adoption and increasingly complex technology environments are making those systems harder to understand, manage and evolve. The result is a growing dependence on a small number of experts who understand how critical systems operate, making it more difficult to maintain speed and reliability while scaling operations.
So, what is the key to introducing AI effectively?
Kyndryl’s experience implementing agentic AI for a customer in the manufacturing industry highlighted several important lessons about what it takes to create lasting value.
As cloud-native system development and system modernization accelerate, configuration design teams are facing a growing burden: They must repeatedly evaluate assumptions and competing priorities, understand their potential impact and determine which approach is best for the business. To help address this challenge, we introduced a design-support AI agent that organizes the information teams need to make these decisions effectively. The objective was to reduce the time employees were spending on gathering and organizing information, and enable them to focus on the complex decisions and work that drive business value.
As the architect of the agent explained, “By presenting all of the available options alongside the reasoning behind each recommendation, even less-experienced team members can better understand how decisions are made and develop their expertise more quickly."
This highlights another important benefit: AI agents can serve as powerful tools for accelerating learning while also strengthening decision making.
The success of these AI agents depended on more than technology alone. It required operational knowledge, business context and practical experience developed throughout years of managing complex IT environments — and then structuring it all in a way AI could understand and apply.
As one member of Kyndryl’s implementation team observed, “Documented knowledge alone is not enough. Only when we combined the business context with operational knowledge did we create a foundation that the AI could successfully operate in.”
By developing a deep understanding of the customer’s business, their IT environment and operational requirements, our team was able to move beyond experimentation to real-world application.
Above all, the project demonstrated that successful AI implementation depends on more than the technology itself. The true promise of AI lies not in automating work, but in redesigning it — bringing together people, technology and expertise to deliver better business outcomes.
