Transform.

Everything you built in Stages 1 through 4 decays. Not because you did it wrong, but because the ground moves.

Three things break a working setup:

• Models update. An attorney's contract-summary prompt that produced tight two-page memos can start producing four-page ones after a model update. Nothing announces the change. The output drifts, and if she isn't checking, she ships worse work.
• Tools die. A teacher built her quiz-generation routine on a free tool. The company pivoted and shut the feature down with 30 days' notice. Her workflow went with it.
• Your job shifts. The tasks on your leverage map are not permanent. New role, new clients, new season: the map from six months ago describes a job you no longer have.

None of this is a reason to skip Stages 1 through 4. It is a reason to add Stage 5. A snapshot decays. A maintenance habit doesn't.

The fix is cheap: about an hour per quarter, on the calendar, with a worksheet. That hour is the difference between a system that compounds and a setup you quietly abandon by next year.

Your leverage map is a living document. Re-run it four times a year. Block 60 minutes on the first business day of each quarter. Set the recurring calendar event now, before you finish this page.

The re-run is faster than the original. You are not starting from zero; you are checking what changed. Two things to look for:

• Tasks that moved. An accountant's January map is full of tax-prep tasks. Her July map is client advisory and cleanup work. Same person, different leverage. A workflow built for the January map sits idle by summer.
• Scores that moved. Tools improve. A task you scored 2-out-of-5 for AI fit last year, say first-pass review of scanned receipts, might score 4 now. A military spouse running a remote admin business found that meeting-notes summarization went from "not worth it" to "core workflow" in two quarters as transcription got cheaper and better.

The Quarterly Re-Assessment Ritual below walks you through it: re-score the map, audit the library, pick one new workflow to build. One. A single finished workflow per quarter beats three half-built ones.

Every workflow you built in Stages 3 and 4 gets one of three verdicts each quarter: keep, fix, or retire.

• Keep: you used it in the last 30 days and the output is still good. Touch nothing.
• Fix: the output got worse or the prompt broke. A nonprofit grant writer noticed her letter-of-inquiry drafts had gotten padded and vague after a model update. Twenty minutes of prompt tightening fixed it. If a fix takes more than an hour, that's a retire.
• Retire: you haven't used it in 90 days. Delete it from the library and move on.

Retiring is not failure. A veteran in transition built a resume-tailoring workflow in Stage 3, used it eleven times, landed a role, and retired it. The workflow did its job. Keeping dead workflows around makes the library a junk drawer. You stop trusting it, then you stop using it.

Track one honest number each quarter: hours per week the library saves you. If that number drops two quarters in a row, your map and your job have diverged. Re-do Stage 1 in full, not the quick version.

Most AI news is noise. Vendors announce more than they ship. Benchmark scores don't predict performance on your work. A new tool that's impressive in a demo is often worse than your current setup on your actual tasks.

So don't try to keep up. Build a filter instead:

• Pick two sources. One newsletter or publication that covers AI tools for your field, one general. Read them monthly, not daily. Unsubscribe from the rest.
• Run every tool through the five-question filter (below) before giving it any time. Most tools fail at question one: they don't touch a task on your leverage map. No mapped task, no test. A teacher doesn't need an AI video generator, however good it looks, if video isn't on her map.
• Test at most one tool a month, 30 minutes, on your own real work, side by side with your current setup. Keep the winner.

This feels like falling behind. It isn't. The person who tests one relevant tool a quarter and adopts it beats the person who reads everything and changes nothing. You built judgment in Stage 2, what AI does well, where it fails. The filter is that judgment, applied on a schedule.

The quarterly rhythm covers most changes. Three events warrant breaking it:

1. A tool you depend on shuts down or changes its data policy. Move within the week. If you followed Stage 3, your workflow is documented as prompts and steps, not locked inside one product, so porting it to another tool is hours, not weeks. This is why the library matters.
2. The model behind your workflows gets a major update. Spot-check your three most-used workflows against past outputs. An attorney should re-verify citations and summaries before trusting drift-prone outputs with client work; an accountant should re-check any workflow that touches numbers. Updates usually help. Usually is not always.
3. Your job changes. New role, big move, new client base. A military spouse relocating mid-year shouldn't wait for October to re-map a job that changed in June. Run the full ritual within 30 days of the change.

Everything else waits for the quarter. A new model announcement is not an emergency. A competitor adopting some tool is not an emergency. The calendar, not the news cycle, decides when you spend attention.