Can vs. Should
Twelve jobs product teams actually do, scored on two axes. One tells you where AI works. The other tells you what it costs you to hand it over. They are not the same question, and the gap between them is where teams lose their judgment without noticing.
Claim your jobs
4 minCheck the jobs you personally do, not your team, you. Then put rough hours per month next to each. Rough is fine. Wrong is fine. Directionally right is all this needs.
Score both axes
8 minGut-score each job you checked, 1–5 on each axis. Don't agonize. If you're stuck between two numbers, take the lower one.
CAN How well can AI do this today?
Input · Verifiability · Context · Volume
- 1Confident garbage. You'd redo it from scratch.
- 2Occasionally useful, mostly cleanup.
- 3Solid first draft. You do real work on top.
- 4Better than your first draft. You edit, not rebuild.
- 5At or above your level. You spot-check.
SHOULD What does handing it over cost you?
Judgment debt · Reversibility · Accountability · Presence
- 1Never delegate the decision. This is the job.
- 2Delegate research inputs only. Thinking stays yours.
- 3Delegate the draft, own the call.
- 4Delegate the work, review the output.
- 5Delegate freely. Nobody gains from you doing it by hand.
| The job | Step 01 | Step 02 | |||
|---|---|---|---|---|---|
| I do this | Hrs / mo | CAN 1–5 | SHOULD 1–5 | Pain 1–5 | |
If a job scores 3 and 3, it's scoped too broadly. Split it. "Prioritization" is fuzzy. "Assembling the inputs for prioritization" and "making the call" score completely differently.
Plot it, then pick one
5 min
Reveal the two-by-two
Closed on purpose. If you can see where
the answers land, you score toward them.
Start here: score a few jobs first
Your pick is the highest hours × pain in the Automate quadrant. Not the most interesting job. The biggest one you lose nothing by handing over.
The second blank is the one people skip. It's the one that turns a worksheet into a change.
Count what you're not taking home
90 secYou picked one job out of the Automate quadrant and you can go build that on your own. The instruction is build the workflow, and it's enough to act on. Now look at the other one.
AI is already good enough to do these, and the work is your judgment. The only instruction on this whole worksheet for all of them is one sentence: keep the draft, keep the decision.
That sentence is right. It's also useless. It doesn't tell you what the checkpoint looks like on prioritization, what AI-assisted-but-not-AI-decided strategy looks like on a Tuesday afternoon, or what to do when your PM scores a job a 4 and you score it a 1.
Signal to Strategy
From Customer Truth to Product Direction with AI
In the AI era, building is cheap. Deciding is the whole game.
- Cohort
- Sep 9 – Oct 3, 2026
- Format
- 4 live sessions90 minutes each, recorded
- You build
- 6 lessons, 3 projectsOn a real decision you bring
- Price
- $599 CADLifetime access to the templates
This audit sorts your jobs. It doesn't tell you what to do with the sorted ones. Sorting takes eight minutes. Building the decision system for the jobs that landed in your Trap quadrant takes four sessions.
- →The augmentation audit, run properly. Job by job, from discovery and prioritization through pricing, GTM and retention, with reference builds and the benchmarks underneath each score.
- →Signal into truth in minutes, not weeks. Your biggest Automate win, built as a reusable system rather than a one-off prompt.
- →A decision you can defend. Every product call tied to the ROI number that matters, with the evidence documented, so it holds in the room. This is what a checkpoint on a Trap job actually looks like.
Most AI-for-product training teaches you to build faster. Building faster isn't the problem. Most teams already ship more than they can justify. This is about deciding right, which is the part that's still yours in eighteen months when all the tools have changed.
20%+ off when you bring a teammate