Builds a defensible hundred-day agentic AI plan for one entity and one use case, by questioning you rather than accepting a brief.
Private skill · shared by invitation · How access works
Every failed hundred-day plan covers everything, so nothing in it can be proved wrong.
Once the pilot is live the comparison no longer exists, and no later analysis recovers it.
“The data office” owns it being wrong, which means nobody does.
A diagram showing a lineage service gets treated as lineage being in place.
Documents first. Every extracted answer keeps its source and stays marked Claimed until you confirm it.
Fifty questions in nine sections, a few at a time. The five gates that stop you are binary, and a no is not softened.
What “better” means in a number, its value today, and who captures it before Day 1.
Plan page, foundations, phased sequence, Jira backlog, executive ownership, risk register, and what could not be answered.
One entity and one workflow, plus whatever documents already exist: architecture diagrams, a previous plan, policies, an org chart, a DataPulse score.
Estimate by the same method; newer than the rest of the portfolio.
A name and a title against who owns the entity being wrong, and who owns the workflow.
Thirty minutes with Harveer, on your own data problem.