Assess.

Most people start backwards. They pick a tool, watch a demo, then hunt for something to use it on. Two weeks later the subscription sits idle.

The right first question is not "what can AI do?" It is "what is my week made of?"

Your week is the raw material. An estate attorney who spends nine hours a week drafting client letters has a different opportunity than a litigator who spends those hours in depositions. Same profession, different map.

That is why ADAPT starts here. Stage 2 (Demystify) teaches you how the tools work, and Stage 3 (Apply) builds your first workflows, but both run on the task list you build now. Skip this stage and you will practice AI on toy problems instead of your real work. Spend the week. The map pays for itself.

Track one normal work week. Not your busiest, not a holiday week. A typical one.

Use the Task Inventory Worksheet. At the end of each day, write down what you did in blocks of 30 minutes or more. On day five, consolidate: group the entries into named recurring tasks and total the hours.

Aim for 10 to 25 tasks. "Email" is too broad. Split it into "answering parent questions" and "scheduling IEP meetings" if you teach, or "client status updates" and "vendor invoices" if you run a bookkeeping practice. Specific tasks score cleanly. Vague ones don't.

Include the unpaid work that matters to you. A military spouse rebuilding a career after a PCS move should log job-search tasks: tailoring cover letters, researching employers, prepping interview answers. The audit covers work you do, not only work you bill.

Common result: three to five tasks consume half the week. Those deserve the hardest look.

Run every task through the AI-Suitability Scorecard. Five questions, two points each, for a score out of 10:

1. Language-heavy. Is the work mostly reading, writing, or summarizing?
2. Pattern-based. Does the output follow the same structure each time?
3. Repeatable. Does it recur weekly or more often?
4. Checkable. Can you judge the output as good or bad within five minutes?
5. Safe inputs. Can you hand over the inputs without exposing client, student, or donor data, or redact them fast?

Examples. Drafting donor acknowledgment letters: language-heavy, templated, weekly, easy to check, scores 9. Reconciling bank feeds: pattern-based but numeric, and current AI models make arithmetic and lookup errors, scores 4. Representing a veteran at a disability hearing: scores 1. No tool replaces being in the room.

Be strict on question 5. Attorneys and accountants handle privileged data. A high score never means "paste confidential files into a chatbot." It means the task shape fits. Safe data handling comes in Stage 2.

Now plot. The Leverage Map is a 2x2 grid. Horizontal axis: hours per week. Vertical axis: suitability score. Draw the dividing lines at your median task hours and at a score of 7.

Four quadrants:

• Top right (high time, high suitability). Start here. This is where AI help compounds.
• Top left (low time, high suitability). Quick wins. Batch them later.
• Bottom right (high time, low suitability). Your human edge. Classroom time, hearings, client meetings. Protect these hours. The point of the rest of the framework is to give you more of them.
• Bottom left (low time, low suitability). Leave it alone.

A teacher changing careers might find lesson-plan adaptation and recommendation letters in the top right, rubric drafting in the top left, and parent conferences in the bottom right. A transitioning veteran might find resume tailoring and LinkedIn outreach top right, and the salary negotiation call bottom right, though prep for that call can score high on its own.

From the top-right quadrant, pick five tasks. Write them as a ranked list with hours per week and suitability score attached. That list is the only deliverable of this stage.

Why five? Stage 3 (Apply) builds one workflow per task, and five is what most working professionals can build and test in a month without dropping their day job. More than five spreads you thin. Fewer than three means the audit was too coarse. Go back and split your big tasks.

Resist the urge to start prompting tonight. You will get better results after Stage 2, where you learn what the models do well, where they fail, and how to feed them context. The map keeps. Your week, mostly, repeats.