You will receive 70 AI ideas this quarter. You will fund 6. The triage process between those numbers is what separates organizations that ship from organizations that talk.
The 8-week funnel
- Weeks 1–2: Intake. A standard one-page intake form. Problem, owner, expected impact, rough scope. No solutions yet.
- Weeks 3–4: Triage. Three-person panel scores each idea on impact, feasibility, and strategic fit. Cuts to ~20.
- Weeks 5–6: Discovery sprints. 2-week scoping sprint per surviving idea. Output: a shaped proposal with effort, ROI estimate, and risks. Cuts to ~10.
- Weeks 7–8: Investment committee. Funded vs. parked vs. killed. ~6 funded; the rest get clear feedback.
How to score
Three numbers, each 1–5: impact, feasibility, strategic fit. Multiply them. Sort. The top of the list is not always obvious; the bottom of the list almost always is.
How to say no without crushing people
- Say no quickly. The cost of a slow no is higher than the cost of a fast no.
- Say no with reasons. Generic rejections feel arbitrary; specific reasons feel fair.
- Say no with a path. "Not now, but if you can demonstrate X, we'll revisit in Q3." gives people something to do.
The portfolio shape
Of your 6 funded projects, target: 3 high-confidence operational improvements, 2 medium-confidence revenue plays, 1 high-uncertainty capability bet. This shape protects the program from a few inevitable failures while still creating optionality.
The one thing every AI PMO needs
A single dashboard that shows: funded projects, milestone status, in-production count, ROI realized to date. The dashboard is what the CEO will ask for. Build it before they ask.