Problem Framing Comes Before AI Agent Solutions

At Finite Computer’s agent camps, Skyler Westby teaches people how to work with AI agents. The first day is spent getting familiar with the technology, while the second focuses on something more fundamental. How do you figure out what the actual problem is before asking an agent to solve it?

The workshops introduce basic design thinking techniques, including Toyota’s five whys, which involves repeatedly asking why something is happening until you uncover the root cause. Skyler found that when he tried the same exercise with an AI agent, its answers were often poor. Without firsthand knowledge of the organisation or its circumstances, the agent struggled to understand what was really going wrong.

Pete Winn made a similar observation. An AI agent can start changing things, but it has no experience of its own to tell it whether something needs fixing in the first place. People encounter problems through their work and everyday lives, and that experience gives them an understanding of what needs to change. Once they’ve identified the problem, agents can be incredibly useful in exploring and building solutions.