The workflow decides the technology. The engagement starts with how work actually moves through the company, then builds only what the evidence supports.
Watch the work as it happens now. Establish the baseline before changing it.
Decide what the system may use, where approved information lives, and which work deserves a repeatable path.
Use real work already due. The people who will own the workflow use it, break it, and help refine it.
Compare selected work before and after. If the result does not justify expansion, the recommendation should say so.
Platform decisions come after workflow and information requirements.
If a workflow produces bad output by hand, automating it produces bad output faster.
Every workflow has someone accountable for keeping it current.
Client-facing and consequential output has accountable review.
More AI use is not the objective. Useful, repeatable work is.