From laid-off to lead teacher: a playbook for turning years of craft into AI-era teaching

1. Turn your craft into a 90-minute lesson

A good lesson starts from one outcome a learner can demonstrate, not from a list of tools.

Professionals who try to teach often start with the tool. Start with the work instead. Write one sentence that begins "By the end, the learner can..." and make it observable: a checked claim file, a reconciled ledger, a tested prompt set, a reviewed drawing.

Use the template below for every lesson. The time boxes add up to 90 minutes.

Three rules keep the demos useful:

• Use practice material only. Never demonstrate on live client, patient or employer data.
• Narrate judgement, not clicks. Say why you chose the option you chose.
• One idea per demo. If a demo needs a second idea, it is two demos.

The failure case is the most valuable block. It shows learners that AI output can be plausible and wrong at the same time, which is the skill a seasoned practitioner already has.

2. Role translation: from classic roles to AI-era work

Much AI-era work is a new name for a judgement you already use: deciding whether an output is right, and what to do when it is not.

Job titles differ between employers and change quickly. The table shows the kind of work each classic role maps to and the judgement that carries over. Treat the middle column as examples of work, not as a guaranteed opening.

To use the table, find your row and write your own version of the middle column in one line. Then name the one judgement you would bring into your first week.

3. Clean room: bring your craft, not your former employer's secrets

You can teach everything you learned through practice without carrying a single confidential document.

A clean room is a set of rules for building teaching material from what you know, not from what you kept. Your method, your general judgement and your checklists, written fresh, are yours. Internal files, customer records, pricing sheets, source code, unreleased plans and screenshots of internal systems are not.

• Do not copy, photograph or forward internal documents, customer data, pricing sheets, source code, dashboards or internal chat logs.
• Rewrite every example from memory as a generic case. Remove names, numbers, identifiers and anything that points to one employer.
• Use public or synthetic data for every demo and exercise, and label synthetic data as synthetic.
• Read your former employer's agreements, including confidentiality, non-solicitation and return-of-property terms. If a term is unclear, ask a qualified lawyer before you publish anything.
• Describe methods in your own words, as how you work, not as a copy of an internal procedure.
• Keep a separate folder for teaching material, with no file that came from a former job.
• Ask a peer to read your lesson for anything that looks internal before it is shared.
• Your employment situation is not a required part of any lesson or profile. Share it only if you choose to.

This checklist is general practice, not legal advice. Pawanax faculty agree to clean-room terms in writing before any work starts.

4. Prove AI literacy in a weekend

AI literacy is shown by artefacts you can point to: a verification log, an evaluation set, a documented workflow, a failure analysis and a taught walkthrough.

Five tasks, each scoped to a few hours, make a folder of dated artefacts that anyone can review. No coding is required. Use generic or synthetic material throughout, so the folder stays clean-room compliant.

1. Verification log. Take one real task from your past craft, rewritten generically. Ask an AI tool for a draft. Find five errors or unsupported claims, and for each one record how you found it and which check would have caught it earlier.
2. Small evaluation set. Write 20 cases from your domain, each with the answer you would accept and one reason it is right. Run the same cases through two prompts or two tools, score both, count the disagreements and explain three of them.
3. One documented workflow. Automate one repetitive step with a spreadsheet formula, a no-code tool or a short script. Document the inputs, the outputs, the human check, and the two most likely failure modes.
4. Failure analysis. Find one confident wrong answer and classify its cause: missing context, ambiguous instruction, a bad source, or a case the tool was never built for. Write the fix you would try first.
5. Ten-minute teach-back. Record a five-minute walkthrough of tasks 1 to 4 for a peer. Write down the one question you could not answer and what you would check next.

Done means a folder with five dated artefacts. Share it with a peer or a lead teacher, and ask them to point at the weakest one.

5. Laid off? A six-week path to a visible AI portfolio

Three dated artefacts in six weeks, built around other commitments, with no former-employer material.

If you were laid off, this path gives you something a reviewer can read within six weeks. It does not promise a job, and it works the same whether or not you share your employment situation. Every artefact follows the clean-room checklist in section 3.

1. Weeks 1 and 2: write one evaluation rubric for a task from your craft, rewritten generically. Score five AI outputs against it, and note where the rubric was unclear.
2. Weeks 3 and 4: write one red-team note. Test only a tool or a practice system you are allowed to test. Try three ways to make it fail, and record the severity and the steps to reproduce each failure.
3. Weeks 5 and 6: record one teaching sample, a five-minute walkthrough of a single task that includes one deliberate mistake and the check that catches it.

Date each artefact and ask a lead teacher to review all three. If a step is not safe or not possible for you, skip it and say so in the notes. The outcomes this path produces are non-monetary: a portfolio, public credit, teaching evidence and references on request.

6. How to read AI labour data without the hype

Start with official statistics, prefer trackers that state their method, and treat prediction-market prices as moving opinions.

Headlines about AI and jobs often quote one figure with no source or method. A slower reading is more useful. Start with the official source for the question you are asking. In the United States that is the Bureau of Labor Statistics, which publishes monthly employment, unemployment and sector data.

For any AI-and-jobs claim, ask three questions: what is measured, over what period, and whether the source publishes how it counts a job loss as AI-related. A claim with no published method is a rumour until it has one.

Prediction markets such as Polymarket and Kalshi also host questions on tech layoffs and unemployment. We mention them as a topic only. Their prices move daily and reflect what traders are willing to stake, not verified outcomes. Pawanax does not publish those prices and is not affiliated with either platform.

Pawanax's own dated forecasts sit in the Signal Ledger, with each probability and the scoring rule. None of them is an AI-jobs forecast yet, and each probability is subjective and uncalibrated.

7. Evidence over credentials for AI roles

A reviewer can check dated work samples, rubric scores and a red-team report. A credential names where you studied; these show what you can do.

For evaluation, red-teaming and domain-expert AI work, reviewers ask what you can demonstrate. A credential helps, but it does not show that you can check an AI output and explain the result.

• Dated work samples, each with the task, the date and the constraints you worked under.
• Rubric scores with the reasoning behind each one, including the cases you scored low.
• A red-team report with the scope, the failures found, their severity and how to reproduce each one.
• A teaching sample or walkthrough that a lead teacher can review.

Build this evidence with the steps in sections 4 and 5, using generic or synthetic material. Label every estimate as an estimate, because a reviewer cannot check a claim that is not labelled. Evidence makes your work easier to read. It does not guarantee a role.

8. How Pawanax faculty works

Faculty teach real practice in a voluntary launch-phase role, under written terms that arrive before any work starts.

Pawanax Academy recruits experienced professionals as human faculty to teach practical AI. Learners are guided by AI coaches, and the work is signed off by real people.

Voluntary launch-phase role. Pawanax is not paying faculty or interns at this time, and we say so up front.

What the role offers, in writing:

• Public credit and a named profile.
• A portfolio and teaching evidence you can show later.
• Structured upskilling into AI roles, with each step written down.
• References on request.
• Flexible hours, planned around your existing work.
• Clean-room protection: you never have to bring a former employer's confidential material.

Written terms arrive before you commit. They cover scope, start date and cancellation. No work starts before those terms are agreed in writing, and Pawanax does not promise jobs, renewals or hours.