// step 0 · context
Who's across the table?
The role sets which metric leads the whole analysis — a CFO hears payback, a COO hears throughput, a CTO hears feasibility. Pick the lens.
// step 1 · discovery
Diagnose the pain, don't shop for features.
Answer in whatever order it comes. We're casting a wide net for where work piles up, what scales with volume, and what a delay costs.
// step 2 · score & rank · deterministic
Rank the candidates on four criteria.
Same gate for every idea — no favourites. Adjust any score; the weighting is seeded by role and fully editable. The math is visible, not improvised.
// step 3 · disqualifiers · hard gates
Kill the deal-breakers before the proposal.
These aren't scored — they're binary. Data-residency especially: surface it in the room, not after you've quoted.
// step 4 · cost & benefit · glass box
Their assumptions. Our math. Nothing hidden.
Every input below is editable and every figure is derived live. Two honesty knobs — coverage and the year-1 haircut — keep this credible.
// step 5 · scoped pilot
One use case. One metric. A time-box.
The smaller and more measurable the ask, the more likely it closes — and two-thirds of AI efforts die in pilot purgatory. Stay narrow.
Every figure is driven by the assumptions entered in step 4 and can be re-run by changing them.