A way of working with AI can stay useful for a long time. It can also miss relevant possibilities because the team never went back to look at it. Updating your AI practice requires a moment to review needs, results and difficulties — not keeping up with every new release.
Start with an important task that already uses AI. Instead of looking for news in general, notice what still makes the work slow, confusing or hard to check.
Use friction as your research question
Do you have to repeat context? Does the deliverable require copying information between apps? Are the corrections always the same? Each difficulty points to a specific investigation.
A team that generates drafts and then redoes the formatting by hand should look for ways to produce the deliverable in its final format. That is a different question from looking for the newest model to write with. In another case, the problem may be outdated context, and the improvement may lie in how the documents are organized, with no new tool at all.
Compare the change against a known reference
Pick material the team has already worked on and test an alternative. Record what got better, what got worse and what would need to change to adopt the new path.
Consider the cost of switching: learning the interface, reorganizing instructions, reviewing access and explaining the practice to other people. A small improvement can be useful, or not worth the switch at that moment.
Set the next review date, or the event that would justify one. That keeps you open to new things without turning updates into an endless task.
Learning as part of operations
For leadership, this stance treats learning as part of the work. The team's knowledge of its own work guides what deserves investigation. In AI Thinking by Dani Hayoshi, this is the Question move: why do we do it this way, and what has changed since we last looked? See also the AI tool map.
Dani Hayoshi is the author of AI Thinking. Her work connects artificial intelligence, education and entrepreneurship. About Dani.