Role Copilot

The reference layer for twenty-four business roles: CEO, CFO, CTO, COO, CPO, CRO, CHRO, CMO, General Counsel, Chief of Staff, Executive Assistant, Accountant, Business Analyst, Financial Analyst, HR & People Ops, Legal Counsel, Management Consultant, Marketing Manager, Operations Manager, Procurement Manager, Product Manager, Recruiter, Sales Manager and Strategy Analyst. 989 entries: 264 priority signals, 281 KPIs, 142 decision frameworks and 302 deliverable outlines. Ask what a role watches and measures; or send the things actually on your desk and each is matched against that role's signals with the matching keywords shown, so you can see why something came back rather than trusting that it fits. IT RETURNS NO RECOMMENDATION, and that is deliberate. The agents this consolidates had an advise tool that produced "options, tradeoffs and a recommendation with a not-professional-advice disclaimer" — assembled from one sentence about a decision it knew nothing else about. A disclaimer does not make a recommendation better grounded. This returns the FRAMEWORK and the factors it says to weigh; deciding belongs to you, or to the model holding the context this agent never sees. IT CARRIES NO PRIORITY SCORE. Those agents ordered a person's day by multiplying an urgency by an impact and cutting the result by a cadence weight — and when the caller supplied neither number, filled both in from their own table, so someone who supplied nothing received an ordering that was entirely the tool's invention printed beside their own labels. A ROLE IT DOES NOT CARRY COMES BACK NAMED, with the full list of twenty-four. Handing you a CFO's KPIs under the heading you asked for is worse than answering nothing. Matching is by keyword and keyword matching is crude — a question about hiring a vendor matches a framework about hiring an executive on the word "hire" — so the keywords each match was made on are in the result, and a match one word deep is flagged. It replaces twenty-four separate agents that carried roughly 108,000 lines between them running twenty-four copies of one engine. Every entry is verified against those source repositories by a test that reads their source files directly, so "nothing was lost" is checkable rather than claimed.