The CambridgeAI Labs
The CambridgeAI Labs
A four-week live cohort for professionals who need to build workflows themselves, not delegate them. Applied hands-on in Perplexity, Claude, Copilot and Grok. Practitioner-led, vendor-free.
By the end of Week 4, every professional in the room has these in hand — for a real workflow of theirs.
One job of yours, drawn out end-to-end, with the bits AI touches marked.
In your language. You wrote it, you can defend it.
Evals, pass-rates and the failure patterns to watch for — so you can prove the agent works, not hope it does.
The specific war-story references you can quote in your next board meeting.
Saved as files, portable across Claude, ChatGPT and Copilot.
What you decided, why, and what you'll revisit. Keeps growing after the cohort ends.
Product, ops, transformation and functional leaders. Senior professionals with the authority to change how their team runs. No coding needed — you bring the workflow, we bring the method.
Every module has three parts: what you leave with, the applied hands-on work, and the AI tools you use for it.
A shortlist of jobs where AI pays off AND you can measure success — unmeasurable work doesn't make the cut.
Audit your team's recurring work live. Score every candidate against seven questions, including 'can we measure it when we ship it?'
Perplexity or Claude for the audit conversation.
One real job mapped end-to-end, with the bits AI touches marked AND the success criteria written down before you build anything.
Walk a real workflow through analyse → deconstruct → design → build → test → run → improve. First version working, first scorecard drafted.
Claude or Copilot to prototype. Perplexity for research.
A small library of your own prompts, checks and eval cases — saved as files you own, versioned alongside the agents they test.
Move a working prompt out of the chat and into a file. Add the test cases. Version both so a regression next week shows up on your dashboard, not in a customer complaint.
Claude, ChatGPT or Copilot as targets. A folder or GitHub for storage — no coding required.
Three or four skills, each with its own eval set: test cases, pass-rate, blind A/B result. Evidence that each piece earns its place.
Turn one thing you do repeatedly into a skill. Write the test cases BEFORE you tune. Compare two versions side by side and pick the winner on the scorecard, not the gut.
Claude and ChatGPT (both support skills). Perplexity for evaluation research.
Two production workflows plus the eval that fits each — deterministic test cases for the on-rails one, human-review rubrics for the collaborative one.
Ship both. Run them against a real week of work. Measure them against the scorecard you wrote in Module 02.
Whichever AI tool your org uses in production — Copilot for M365 shops, Claude or ChatGPT for others. Grok for the research-heavy variants.
A single-agent workflow doing your highest-value job automatically, with a scorecard the exec sponsor can defend. Optional multi-agent extension if the single one is stable.
Build the single-agent version live. Attach the scorecard. Discuss when multi-agent is worth the complexity (usually: not until the single one measures clean for two weeks).
Claude, ChatGPT, Copilot, Gemini — whichever your team uses. Grok for real-time-data agents.
Fourteen-day full refund from the day the cohort starts, no questions asked. If you complete the four weeks and haven't built and shipped at least one working workflow of your own, we refund the whole fee.
No. The programme is designed for professionals who direct work, not build software. Everything is done in AI chat tools you already use — Perplexity, Claude, Copilot, ChatGPT, Grok.
Fine. Every module runs on the tool of your choice. Copilot-only cohorts happen regularly. We show you how to get to the same outcomes inside the M365 stack.
You never share confidential data with the cohort. All applied work runs on your own account, in your own tool, with your own data. We share methodology, not data.
Every live session is recorded and shared within 24 hours. Homework catches you up. The private cohort chat keeps you in the loop.
Yes. We can supply a reimbursement letter on request that most L&D teams accept.
The next cohort opens in November 2026. Join the waiting list at the top of this page and we'll email you the moment the date locks.
Two or more people from the same organisation get 20% off. Everyone in the same cohort so the workflows compound.
Talk to us about team pricing →A closed cohort for one organisation, delivered around your team's real work and tools. Book a call to scope.
Book a scoping call →Join the waiting list and we'll email you the moment seats open. First cohort is limited to 12 places.
🔒 Secure checkout via Stripe when the cohort opens · 14-day refund policy