One day. A real codebase.
What questions appear in your head?
One day, hands-on. You will spend most of it building agentic AI processes on a real, messy, inherited codebase - the kind you actually walk into on a job, not a clean tutorial repo. Then a two-week piece afterwards, on your own real work.
In a controlled study, doctors were given AI advice that was confident and wrong. Their accuracy got worse. The most experienced clinicians fell the hardest - because years had taught them that a well-articulated claim in their field is usually correct. They had everything they needed to catch the error. They just did not engage it.
That is the whole problem. Not that AI is too clever. That we stop asking.
So this is not a day about prompts or tools. It is about what questions appear in your head when a machine hands you something fluent - and whether the professional discipline you already have fires when the source is a chat window instead of a colleague.
By the end of the day and the two weeks after it:
The day is where you learn to think about it and build one. The two weeks is where you do the whole arc yourself, on real work.
Building an agent that actually works is a multi-disciplinary job, and I think that is the part most people underestimate.
You need design thinking, to work out what the process should even be before you automate it. You need psychology and UX, because you are modelling how someone really works, not how they describe their work when asked. You need software engineering, because the thing has to hold up when it is not being watched. And you need enough understanding of how the model actually works to know where it will drift.
Most people are strong in one or two of those. Being genuinely effective at this is not a small feat - and underestimating that is, I think, why so many agent projects quietly die.
This is intense. It is not an introduction to AI. You will get the most from it if you already work with AI or with software, and if you are the kind of person who likes taking a problem apart to see how it actually works - because that is most of what the day is.
Mixed on purpose: working engineers alongside people earlier in their career. Each half is the other half's exposure.
$999 · scholarships available
The day above is for people who build. There is a sister course for people who scope - independent consultants, in-house transformation leads, anyone who has to tell an organisation what to automate and what to leave alone.
Same shape, same price, same two-week tail. You spend the day on a real client with a real problem: interview them, work out what the job actually is, decide what an agent may own and what stays human, and design it. The failure it is built around is different - a bad build fails in ten seconds, and a bad recommendation gets acted on for a year before anyone notices.
Saturday 8 August 2026 - the week before the building one. Same day length, same place. Pick it below.
Places are limited and I read every one of these myself. If it is a fit I will send you the brief with the two setup steps.