Can the AI do that? I had no idea.
AI the workflow doesn't use
The team has AI tools that could take on a step, but nobody has systematically compared the tool’s capabilities with what each step actually requires. People tend to use the small set of features they already know.
- People use a small fraction of what the tool can do.
- Workarounds exist for problems the tool already solves.
- Whoever set up the tool knows it can do more. The team doesn't.
Who: two people who do the work, and whoever looks after the tool. Look at: each step of the workflow. Ask of each: "Has anyone tried the tool on this step, side by side with the person who does it today?" Count: steps never tried. You're probably looking at this pattern if steps the tool might handle were never tested. Probably not if they were tested and the tool fell short. That's a sound decision, worth retesting as the tools improve.
The Change Barriers entry Tools, no redesign. In that pattern, the tools arrived and nothing about the work changed. In this pattern, the work may have changed, but not in the places where the tools could help most.
Pull 30 recent AI outputs. Have two people read them and write one line about each one that has a problem. Then count which problems repeat.
Write one page for your AI tool that lists what a new colleague would need to know in their first week. Include three things it should never do.
Observation. A caution belongs with this one: across 106 experiments, people working with AI did worse on average than the better of the two working alone, though combining helped when people outperformed the AI alone (Vaccaro, Almaatouq, and Malone, 2024, Established). So the point is to test a step side by side before handing it over, not to hand over everything the tool can do.