My original note asked a blunt question: does AI reduce cognitive load, or does it create more of it? Intentional AI can compress a path to an output; messy adoption can multiply the decisions and coordination around that output. Those are not the same thing.
- Low / low A bounded assist may be easy to inspect.
- High decisions / low coordination Clarify the rule before adding more capability.
- Low decisions / high coordination Repair ownership and handoff before tooling.
- High / high Adoption is likely creating a new layer of work.
Count the decisions around the output
A team may be able to draft quickly and still be overloaded by choosing prompts, models, sources, reviewers and handoffs every time. Fewer tools and tighter rules are useful when they eliminate repeated choice; they are not a universal preference for a smaller stack.
Inspect the coordination path
Ask who owns the input, review and next action. If those answers change by person or by week, the workflow is asking the team to coordinate the process afresh. Start with one repeatable use case and observe the decisions it creates before expanding it.