One day AI course · 10:00 to 17:00

Everyone talks to AI. You’ll build one in a day.

You arrive with a job you do by hand. By 5pm you have a working agent on your machine, doing that job on your own work.

Next dateMon 12 October 2026
WhereCentral London
Total£540 inc VAT
Book this date →

We run this course regularly. New dates are added here as they are set. £450 + VAT. Questions first? Email us.

People from these companies have trained with us
BoltEYAngiTrip.comEgon ZehnderMollieCheckatradeBlue Light CardBanco de Crédito BCPDoorFeedMatrix OneSprintsBoltEYAngiTrip.comEgon ZehnderMollieCheckatradeBlue Light CardBanco de Crédito BCPDoorFeedMatrix OneSprints
What you’ll build

The job you want to do better, or faster.

You bring your own. These are the two kinds that work, with examples from different industries. There are as many more as there are jobs you repeat.

An agent that saves you time

A job you repeat, that follows the same shape every time. Low risk, easy to check, and the hours come straight back.

  • Document generatorProposals, quotes, engagement letters and reports, built from your own templates and the last one you sent.Law, trades, agencies, consultancies
  • Internal knowledge baseAnswers pulled from your own policies, procedures and past work, instead of hunting through folders.Any business with more than one person
  • The checkerInvoices against orders, bids against criteria, drafts against your own standards. Every difference flagged.Construction, retail, bookkeepers, charities
  • Team planMeeting notes to actionsWritten up, assigned, and chased when nobody replies.Any business with meetings
An agent that improves your outcome

A job where doing it better changes a number you care about, not just how long it takes. Harder to build. Worth more.

  • Team planContract and document reviewEvery clause checked against the list you already argue from, with the misses flagged.Law, insurance, surveying, procurement
  • Content and marketing engineWhat your competitors rank for, where you are silent, and the first draft of what to write.Agencies, clinics, shops, trades
  • Customer feedback analysisMonths of reviews and complaints, sorted into what to fix first and what to ignore.Hospitality, retail, services, clinics
  • Profitability analysisTime, materials and scope creep against what you charged, job by job.Law, studios, trades, consultancies
Or your own idea. Bring the problem you want solved. The morning sessions will help you decide if it is the right one to build first.
The day

Morning you understand it. Afternoon you build it.

You build in Claude or Codex, whichever you already have. The skills carry to either.

Morning · understand it, then decide
1Discussion

AI foundations

  • From ChatGPT to today, in four key dates
  • Chatbot vs agent, and why the difference matters
  • The five core parts: brain, tools, context and memory, loop, gate
  • Useful AI terms: LLMs, tokens, context windows
  • Why AI sounds completely certain even when it is wrong
Know: what an agent is, and is not
2Discussion

Pick the right job for your agent

  • What makes a job worth handing to an agent, and what to avoid
  • The four-question filter, separating a good agent job from a bad one
  • Choosing and scoping the job for your afternoon build
Choose: your afternoon build
3Hands on

Security and privacy

  • What commercial AI tools do with your data
  • Live privacy check, finding and switching off the training toggle
  • The three conditions that cause a data leak, and how to break one
Set: an account safe to use at work
4Hands on

Set up your workspace

  • The workspace folder an agent reads before it starts
  • What belongs inside it, and what must stay out
  • Connecting one of your own tools (optional)
  • A throwaway test agent, checking your context reads back correctly
Build: your workspace
Afternoon · build it
5Hands on · the long one

Build it

  • Your agent structure: brief, memory, out tray
  • Briefing it in plain English, the way you would brief a person
  • Hard limits: what it may do alone, what it must never touch
  • Run it, break it, sharpen the brief
  • Your coach beside you, the whole session
Build: a working agent on your machine
6Hands on

Check it works

  • A scorecard two people would mark identically: pass, fail, or cannot tell
  • When it fails, fix the criterion, not the work
  • Why an honest agent can still be wrong about the world
Write: a scorecard for your own agent
7Theory, no build

Agent teams

  • The ladder: agent, team, handoff, coordinator, trigger, guardrail
  • The two team shapes, and why you never mix them
  • When a longer brief beats adding a second agent
Know: when one agent is enough

One day, you leave with a working agent that you built.

Book this date →
Who teaches it

Operators, not lecturers.

All three have run the work they are teaching you to hand over. One of them runs your day, and is in the room for all of it, including the whole afternoon build.

Raj Ashan
Raj AshanAI for product

Ran product teams at Trainline and CarNext, and has been in product since 2013.

Seb de Bandt
Seb de BandtFounder

Ran growth at Checkatrade, Super Payments, CarNext and Trainline, over fifteen years.

Ali Goldsmith
Ali GoldsmithAI for marketing

Ran marketing and brand at Auto Trader and Checkatrade, with work across BT and EE.

Before the day

Two things to sort. Neither is a barrier.

Install the toolClaude or Codex. Link and a short guide when you book, a few minutes.
Ask about tool accessIf the answer is no, bring exports. Same agent either way.
FAQ

What you’d want to know.

Do I need to be technical?
No engineering team needed. If you can write a clear brief for a person, you can build these agents, and a coach builds alongside you for most of the afternoon. Briefing an agent well is the real skill, and it is the one you practise all day.
What do I bring?
A laptop you can install software on, and one piece of work you currently do by hand more than once a month. That second one matters more than the first. The whole day runs on your own work, so the agent you leave with is one you will actually use.
What if I cannot connect a work tool?
You build the same agent from exports instead. Some people arrive with tool access sorted and some do not, so the day is designed for both. Nobody spends the afternoon waiting for permission from an IT team.
Where does my data go?
One of the morning blocks is spent entirely on this, on your own account rather than in the abstract. You will set the training toggle, see what your plan does with your conversations, and decide what an agent may see and do where you work.
Do I need a Team plan?
For most of the jobs on this page, no, your existing account is fine. Two of them read other people's personal data, which needs a data processing agreement, and that only comes with a commercial plan. The security block is where you learn to tell which is which.
How is this different from the four week course?
This is one day and you leave with one agent working. The four-week course goes considerably further: a new agent every week, agent teams, and the work of keeping them reliable over time. Neither is a preview of the other.

More questions? Email Raj any time.

One day, you leave with a working agent that you built.