Cohort 1 · Applications open
Eight weeks. Eight real projects. Proof you can do the work.
Most AI training leaves you with notes about someone else’s demo. This one leaves you with a published portfolio of work you did on your own job, your own data and your own problems — the kind of evidence that answers “what have you actually built?” The core eight weeks are free.
Information session
Thursday, September 17, 2026
7:00–8:30pm ET, online. Come with questions.
Applications close
Thursday, September 24, 2026
Everyone hears back within a week.
Week 1
Thursday, October 1, 2026
The cohort starts. Bring your task.
Graduation
Thursday, November 19, 2026
You present what you built.
Not five topics you sat through. Each one is attached to something you shipped and can show to someone who is deciding whether to hire you, fund you, or let you lead the work.
Define what is actually broken, and what better would measurably look like, before reaching for a tool.
Prompting as an engineering discipline: structured output, tested against a rubric, versioned like code.
Check the model's work, find where it fails, and state the limits out loud. The habit most training skips.
Ship a working automation and a tool-using assistant against a real system, with a human in the loop.
Demo in two minutes, write the one-page memo, and publish the evidence where someone can check it.
Awarded on completion, shareable, and checkable by anyone who asks.
In week one you name a real task you already do every week — chasing invoices, writing the same report, sorting a rota. Everything you build for the next eight weeks attaches to that task. That is what turns eight exercises into one portfolio.
Every session is live on a Thursday at 7:00–8:30pm ET and recorded. Projects are due before the next session and every one gets written feedback.
You ship: The Problem Brief
Pick one recurring task from your job, volunteer role or studies — something you personally do at least weekly. Time it honestly for a week. Map every step. Write a one-page brief stating the task, the current time and cost, who it affects, and what a measurably better version looks like. This task is your through-line for all eight weeks.
You ship: The Reusable Prompt
Turn week 1's task into a prompt someone else could run and get the same quality. Define the output structure. Write a five-point rubric for a good answer. Run it five times on five real inputs, score each against your rubric, and document every failure mode you hit.
You ship: The Red-Team Report
Attack your own week 2 output. Where does it fabricate? What does it get confidently wrong on edge cases? Whose data went into the prompt, and should it have? Produce a one-page limits-and-disclosure note: what this prompt must not be used for, what a human must check every time, and the disclosure line that goes to anyone receiving its output.
You ship: The One-Chart Answer
Take a real, messy public dataset connected to your problem or your community — a local budget, 311 calls, inspection records, public health statistics. Clean it with AI assistance, then answer exactly one question with exactly one chart. Write the three caveats a sceptical reader would raise, and answer them.
You ship: The Automation
Build a working automation for the week 1 task, end to end, in whatever your workplace actually allows — Power Automate, Apps Script, Zapier, Make, a script, a Claude Project. It must run on real input and produce real output without you babysitting it. Measure the new time-per-run against your week 1 baseline.
You ship: The Assistant
Build an assistant that uses at least two tools against a real system — reading a document store, querying a sheet, calling an API, sending a draft for approval. Put a human approval step in front of anything it changes. Write ten test cases, run them, and report your pass rate honestly.
You ship: The Two-Minute Demo
Record a two-minute demo aimed at a decision maker who is not technical, and write the one-page memo that goes with it: the problem, what you built, what it measurably changed, what it costs, what you recommend. No jargon. Two minutes means two minutes.
You ship: The Portfolio Page
Publish one public page covering the whole eight weeks: the problem, the baseline, what you built, the measured result, and the limits you found. Link the demo. This is the artefact you send to an employer instead of describing yourself as 'AI-familiar'. Present it at graduation.
There is no upsell inside the programme and no card required to finish it. Afterwards, graduates who want their portfolio taken further can apply for a paid capstone. Most will not need to, and that is fine.
8 weeks
The full eight weeks, the eight projects, every live session and the digital badge. No cost, no catch, no card.
5 further sessions
For graduates who want the portfolio taken further and put in front of people who hire. Capped at twelve, applied for after week 6.
No prior AI experience is expected. If you have never written a prompt beyond asking a chatbot for a recipe, you are exactly who this was built for.
You earn the badge by:
A digital badge issued by JMCB Technology Group. It is a record of work you completed and can be checked — not an accredited academic qualification.
Not sure this is the right thing for you? Take the free six-minute Career Compass first. It scores your job search across seven dimensions and tells you honestly whether the role you want is realistic today.
Take the Career CompassApplications close Thursday, September 24, 2026. It takes about five minutes, and there is one question that actually decides it.
Proof of Work runs as a private cohort for workforce programmes, associations, ministries and internal academies — same eight projects, your people, your systems, and a documented record of what the cohort shipped.