AI Training
You can be brand new to all this. The smart money in 2026 is in learning to direct AI on your actual work, not someone else's example data.
AI Training
AI Training takes a person, or a team, from briefly used ChatGPT to confidently directing agents through real work. We don't teach the history of AI or the maths under the hood. We teach you (and your team) how to break down the work you actually do, codify it into prompt templates that are reusable, and use the kit (Claude, ChatGPT, Copilot, Gemini, the MCP servers we ship in The Build) so it pays back the same week. The smart money is on learning this stuff now. The ones who do will be the ones running it for the next decade. The career upside is real.
What you'll learn, how it runs, and what you walk away with.
What you'll actually learn
Mental models over maths. The stuff a senior practitioner uses every day.
- The mental models
- How an LLM actually works, in plain English. Why context matters more than the prompt. What an agent is, and what it isn't. We skip the history of AI and the transformer-architecture diagrams on purpose.
- Prompt template authoring
- Moving from ad-hoc chatting to codified prompts your team can reuse. System prompts, few-shot examples, structured outputs. The shift from "I'll just type something" to "the prompt template handles this case, every time".
- Agentic workflow design
- Taking a messy human task and breaking it into discrete, verifiable steps an agent can run. The skill that turns a one-person business into one that looks like ten people, and a marketing team into one that finally gets to do strategy.
- Tool selection
- When to use the foundation model directly, when to call a custom MCP server, when an Excel formula is still the right answer. Senior practitioners pick the right tool for the job.
- Evaluation and red-teaming
- How to know if an output is actually good, or just confidently written. Reading AI work for hallucinations, brand-tone drift, and the kind of subtle wrongness that's expensive in production.
- UK GDPR, privacy and security basics
- What goes into a prompt, what stays out. How to handle customer data, proprietary financials, anything regulated. The lines that matter for an SME.
How it runs
Adult-learning works when the data is real and the spacing is right.
- Small cohorts, six to ten people max
- Big enough for momentum, small enough that everyone gets airtime. Solo learners join via 1:1 programmes instead.
- Spaced sessions, not a single workshop day
- Ninety to one-hundred-and-twenty-minute sessions over three to four weeks. Adults retain almost nothing from an eight-hour cram session. They retain plenty from spaced practice on their own data.
- Bring your own backlog
- Exercises use your actual work. If you're an e-commerce team, we use your abandoned-cart data, not a fictional bakery. (This is where the AI Audit pays back: we already know what's eating your hours.)
- Async homework that does real work
- Between sessions, you apply what you learned to a task you were going to do anyway. The homework isn't a test. It's productive output.
- Remote-first, hybrid is better
- Remote works perfectly for screen-sharing prompt iterations. A single on-site kickoff helps the team admit "I don't quite get this yet" without it being awkward.
What you walk away with
Working artefacts on Monday morning, not a certificate.
- A shared vocabulary
- Your team stops saying "the AI is acting weird" and starts saying "the agent's hallucinating because the context window is cluttered, let's tighten the prompt template". This is what makes the next conversation productive.
- A versioned prompt library
- Five to ten battle-tested, codified prompt templates the whole team can use immediately. Not a generic "100 prompts for marketers" PDF. Templates built on your work, by your team, that you'll keep using.
- An internal AI owner emerges
- By the end of the cohort, one or two people will have stepped up as the in-house go-to. This is the person we hand the kit over to in The Build, so the training and the engagement that follows are properly joined up.
- Working artefacts on Monday
- Every person leaves with at least one automated workflow or codified prompt that saves them a couple of hours a week. Multiplied across the team and the year, that's a meaningful number of hours back, on the right side of the value chain.
The teams that get the most out of it.
Marketing, e-commerce, data and ops teams in SMEs who need to come up the curve together. One-person businesses who want to operate like a ten-person team. Senior leaders who've been experimenting alone and want to bring the rest of the team along. Total beginners are welcome. The curriculum starts with the mental models you actually need, not the maths under the hood.
The shape of the engagement.
We start by talking. What's the team like? What did the AI Audit surface as the highest-value work? What's the team's appetite for change? From there we shape the cohort. How many sessions, how spaced, what artefacts to ship. Sessions are remote by default with optional on-site for the kickoff. Between sessions, you work on your real backlog. By the end you have a shared vocabulary, a versioned prompt library, and an internal AI owner.
What the next conversation looks like.
A scoping call. Tell us about the team, where they are now, and what good would look like. We'll suggest the right shape (a half-day proof of concept, a three-week cohort, or a 1:1 programme for an internal AI owner). Most cohorts run alongside or just after an AI Audit, so the curriculum is grounded in the work that matters.