When AI arrived, ideas came from all over — product, editorial, sales, engineering — and most of them stalled. Submitters felt ignored, prototype choices looked arbitrary, and people were losing faith that anything would come of "we're doing AI." The fix was a funnel anyone in the org could watch.
How it works
A public intake board where anyone in the organization can submit an AI idea and watch its status; a triage process with a promise — every submitter gets a substantive feedback summary within two weeks; and a fast-track path that takes low-risk ideas from concept to validated prototype inside a single sprint.
What made it work
Make the funnel public. The funnel's numbers are public inside the org. When an idea gets killed, everyone can see why. That's a big part of why people trust the ones that advance.
A fast track with explicit entry criteria. Treating every idea as equally risky kills momentum. Scoping a low-risk lane — bounded data, no customer exposure, reversible — let us validate ideas in days instead of waiting a quarter for the default process.
Quarterly partner check-ins. Bottleneck reviews with product partners kept the pipeline pointed at strategy, and grew into a shared roadmap board aligning AI work with product priorities.
The numbers
In one fiscal year, 29 proposals came in, 16 became working proofs of concept, and 10 reached production or active development — we'd targeted 6–8 PoCs with 3–4 making it that far. The stat I care most about: all 29 submitters got substantive feedback within two weeks, including every "no." Experimentation got cheaper and more visible, and people started trusting the process again.
I designed the funnel, ran it, and did much of the prototype engineering myself. The fast track only works if the person running the funnel can also build what's in it. I wrote up what rapid prototyping taught me along the way.