Before You Add AI, Follow the Work

The useful question is not simply whether a task can be automated. It is whether automating it would make the work better. An approval that serves no purpose is still an unnecessary approval, even when it arrives remarkably quickly.

Start with one workflow, agree on what a good result looks like, and keep someone responsible for reviewing the output. Then test a small change against the current way of working. The point is not to fill the business with impressive tools. It is to give people back some time and make the work easier to trust.

One useful question:

Where does the work wait? Pick one recent project. Look at the time it spent waiting for a decision, information, or approval. Ask the team which delay is worth fixing first.