Score every intake request the same way, rank it, and staff it against who actually has capacity
PMNA Tutorial | Govincorp Innovation Center
Two issues ago we handed you a prompt to find every AI opportunity in your business. It’s very good at one thing: generating a big list.
What it doesn’t do is tell you which of those ideas are worth building, or who should build them.
If you’ve ever sat near an automation team, you know the shape of this. Internal customers submit ideas — “we do this every single day, can you automate it?” Some of those are genuinely great candidates. Some of them are a pile of manual judgment calls wearing an automation costume. Every intake form says high priority. Nothing gets scored the same way twice. The backlog just grows, and the work gets assigned by whoever argued loudest in the prioritization meeting.
Then there’s the second half of the problem, which is worse: even when you pick the right project, you hand it to the wrong person. Everything gets loaded on the team at once with a you guys figure it out, and three weeks later you find the blocker that made the whole thing infeasible.
This tutorial closes both gaps. You end up with a ranked, staffed backlog — and a draft reply already written for the people whose ideas didn’t make the cut.
A recommendation engine that reads two boards and gives you three outputs:
| Output | What it does |
|---|---|
| A scored, ranked backlog | Every intake submission rated against the same five criteria, ranked best-to-worst, with bad candidates flagged and explained |
| A staffing recommendation | Top opportunities matched to specific team members based on skills and current load — including “this one needs two people” and “your best match is underwater” |
| Drafted SME replies | For the submissions you’re declining, a ready-to-send note explaining why |
One important framing before you build it: this recommends, it does not assign. Claude doesn’t reach into your board and start moving work onto people. It hands you a proposal with the reasoning attached, and you make the call. That distinction is what makes this safe to run on a real team.
If you don’t have these boards yet, build them. The quality of everything downstream depends entirely on what’s in these columns.