Design Principles
Four principles that translate the North Star vision into decisions. Each takes a clear stance, settles trade-offs, and tells us what to build — and what to refuse.
Outcomes, not conversations
Default to presenting finished work for judgment — never a blank prompt waiting for instruction.
Enterprise users never wanted the workflow; a copilot just relocates the labor. The service model — go away, come back with finished work — is what the buyer actually wants.Grounded in their own data
Every generated artifact must visibly trace back to the customer's own locations, activities, and risks.
Trust follows recognition. “AI generated this” carries no authority; “built from your Frankfurt site and your three critical activities” does.Design for judgment, capture the signal
Where possible, present multiple options to choose between rather than one output to approve or reject.
A choice is faster and more confident for the expert than a binary verdict — and the unchosen options are preference data (B>A, B>C) that improve the next generation. The review UX is the data engine.Make trust visible
Actively de-risk presenting AI output — show confidence, reasoning, and sources, and let users own and showcase the result.
A BCM manager presenting AI plans to the board takes personal risk; consultants used to absorb the blame. We compress the trust curve with evaluation and traceability UX — confidence scores, judge panels, source trails.Using the principles
Principles are only worth writing down if they get used. Reach for them in these moments.