When is a process ready for an AI agent?
A process is ready for an AI agent when the work is repeatable, the systems are reachable and the result is testable. Only processes that meet all three conditions get a place in the build plan.
Explanation
Repeatable work: it happens often and follows a recognisable pattern, with exceptions you can name.
Reachable systems: the data and applications can be reached through an API, through MCP (the open protocol agents use to call tools) or through computer use, where the agent operates a screen the way a person does.
Testable result: you can establish whether the agent did it right. You need that to evaluate the agent before it goes live and to keep measuring afterwards.
Two out of three is not enough. Without a testable result you cannot test an agent and you never know whether it improves. Without reachable systems it stays a demo.
What it looks like in practice
At an insurer, assessing damage claims scores high: it happens daily, the systems are reachable through an API and past assessments form a test set. Strategic procurement negotiations score low: every conversation is different and the result is hard to test.
