Many companies roll out AI by bolting it onto an existing process, with the same steps and the same roles as before. As I wrote in AI Belongs in the Workflow, tools that stay outside the system where the work happens often go unused. A better approach is to rebuild the process, and often the responsibilities and skills of the people in it.
The reason is that the original process was designed around what people could do before AI. Once AI can draft, sort and check, some steps are no longer needed, others change, and the people involved need different skills. For example, if AI drafts your customer quotes, the estimator’s job moves from typing to checking prices and assumptions. Keep the old process and the old job description, and you may only speed up the wrong steps.
Prosci’s 2026 study of 1,553 workers shows why this matters. How well AI fit a team’s actual work was the strongest predictor of that team’s adoption progress, accounting for nearly one quarter of the differences between teams.¹ The study cannot show that good fit causes adoption, but it makes a clear case that fit matters. Prosci also notes that the people closest to the work need a route to influence how AI gets applied to it.
That suggests a straightforward approach, and it begins with the people who run the process today:
Ask them how the process actually operates now, including the workarounds, and what is working and not working.
Ask them to help rebuild the process with AI in it, one step at a time.
Ask them to define the new skills they will need to perform their augmented roles, and plan the training together.
In Ask Your People Before You Automate, I suggested asking which tasks people want to hand over. Here, the people who do the work also help design what comes next, including the skills their roles will need. They know the exceptions and shortcuts no process map shows, and they have a reason to make the new approach work because they helped build it. You still make the final decisions, but you make them with better information.
Pick one process where AI was added on top of the existing steps. Bring together the people who do it and ask three questions: how does it run today, what works and what does not, and what would the job look like with AI built in? The answers are a strong start on a process that is built for AI from the beginning.
Sources
¹ Prosci, “Closing the Gap Between AI Use and Business Value,” 2026, drawing on Prosci’s 2026 State of AI Adoption study of 1,553 workers in the United States and the United Kingdom. Prosci states that the study does not show that better fit causes adoption to advance.