// in person training
Run leaner. Move faster. Because you actually know how.
You leave able to hand the slow parts of your work to an AI assistant, and to say what each one saved. Come as a team, as one engineer, or on your own. By Friday you will have done it to a real job you brought with you.
Who is in the room
Anyone who wants to get genuinely good at this. Engineers, people who have never written code, a whole team, or one person who booked a seat. What the week asks for is the willingness to work through the learning curve, and nothing else: no technical background, no team behind you, no permission from anyone.
Where a company sends several people, send them mixed: the owner, the managers who would run any of this, and the employees who would use it. Those three usually disagree about what the business is already doing with AI, and Day 1 starts by establishing which of them is right.
Each attendee brings one real job they do repeatedly, and a laptop. From Wednesday the work is on that job rather than a case study, whether it comes from a business, an engineering backlog or your own week. The laptop is not optional - an attendee without one spends the week watching somebody else learn.
The five days
Day 1
What this actually is
An honest inventory of what your people are already doing with AI, authorized or not. Then what it is doing when it answers you, and why it is confidently wrong. After that, the four things it is genuinely good at: reshaping something you already have, drafting from a pattern you supply, finding and classifying in text you provide, and explaining something unfamiliar at the level you ask for. Your team practises all four on their own material rather than watching a demonstration. You break it on purpose, then fix it.
By Monday evening they can explain to a colleague what the system is doing, predict the kind of mistake it will make before it makes it, and say what the business is already doing with AI whether anyone authorized it or not.
Day 2
Where it earns its place
Five questions that tell you whether a job is a candidate - what starts it, what the work is, how anyone would know it came out wrong, where it lands, and whose name is on it - applied to three of your own. The third stops most rooms: a job nobody checks today is not a job to hand over, and finding that out is the useful part. Then you find your own usage and read the rate card yourself, and set a limit on your own account before you receive a surprising bill.
By Tuesday evening they can take any job in the business and say, using the same five questions, whether it is a candidate and why -- and read a usage bill and set a limit before receiving one.
Day 3
Doing real work with it
Produce the actual thing from your own material and compare it line by line against the one a person made. Plus the hour that decides whether anyone uses this on Monday. We cover the tool itself and where a long-lived piece of work lives. Then the six failures that make people quietly stop using it: ignoring part of an instruction, inventing figures, drifting between runs, refusing something reasonable, wrong tone, and answering at length without saying anything. Each one comes with its cause and the single change that fixes it. Then build the check that catches it when it is wrong, and find out what your check misses.
By Wednesday evening they have produced a real artifact from their own business, written the instruction that produced it in their own words, and built a check that catches it when it is wrong.
Day 4
Staying in control
What you paste you have disclosed - read off your own vendor's terms, by you, live. Your own list of what must never go in, with somewhere else for that work to happen. The personal account, and why blame makes it worse. Then you make a document give the instructions and watch it work. Regulated data and who carries the obligation. Then the three incidents you will actually meet - something sensitive was pasted, something wrong went to a customer, somebody did something unauthorized and it worked - and the first ten minutes of each, written on a sheet you take with you.
By Thursday evening they can say specifically what must never be pasted in, recognize the three failure modes they are likely to meet, and know what to do in the first ten minutes of each.
Day 5
Putting it in the business
A one-page policy your own people write and the owner approves in the room. Who approves what. What must be true before anything runs unwatched. The five numbers worth measuring, found on your own account rather than described: how often it ran, how often it stopped itself, how often a person corrected it, what it cost, and the payback on the job they timed by hand on Wednesday. Plus the diagnosis for the sentence that stops most businesses - we tried it and it was not faster - which is almost never the tool, and is one of five things each fixed by an hour of this week.
By Friday the business has a written policy its own people wrote, an approval route, and one named job with one named owner and a review date.
What your team can do on Friday
Stated as capability, and the limits named too, because a course that oversells is one you cannot plan against.
Your people will use AI unsupervised, on their own work, and know when the output is not good enough to send. That is the bar that makes a team faster, and it is reached in five days.
On Friday they can do those four kinds of task on real material. They can write and reuse an instruction, so the job stops depending on who is asking. They can recognise and repair those six failures, say what must stay out of it, and tell when an answer is too weak to send.
Mastery compounds from there, and the week is what starts it: judging quickly which problems suit it, writing an instruction that works first time, and the instinct for when an answer is subtly wrong. Those come from use, and your team leaves equipped to keep building them without us.
What each attendee takes home
What they can do, and the thing they built that proves it. Every artifact is about your business, which is why nobody else can hand it to them.
Day 1
They can explain to a colleague what the system is actually doing and predict the kind of mistake it will make before it makes it. Proof: an honest inventory of what your business is already doing with AI, authorized or not -- which most owners find is the most useful hour of the week.
Day 2
They can look at any job they do and say whether it is worth handing over, and what it will cost. Proof: a ranked shortlist of their own jobs with the reason beside each, and a spending limit they set themselves.
Day 3
They can produce work to a standard you would send a customer, and catch it when it is wrong. Proof: a real finished artifact, the saved instruction that made it, the check that catches failures, and a written note of what that check still misses.
Day 4
They know what must never go in, and what to do in the first ten minutes when something has. Proof: the approved never-list with somewhere else named for each of those jobs, and three incident sheets.
Day 5
You can run this without us, and a team can run it without you. Proof: a 90-day plan naming one job, one check and a review date. Where a company is in the room, that plan comes with a one-page policy your own people wrote and approved there, and the route an approval takes.
How it is taught
The reason it is in person, and what it costs us to teach this way.
- The wrong way first, until it hurts. You never get the good answer first. You ask it badly and get a confident lie; you let a document give it instructions and watch that work; you compare your own output to a person's line by line and find the error. The habit then arrives as relief rather than as instruction.
- You write down what you think will happen, before it happens. A prediction on paper is remembered; a prediction in your head is not. Being wrong having committed is the thing that sticks, and we do not soften the result.
- The objection is answered before it is asked. Every unit has one question a sharp attendee raises, and every one of them is written into the material with its answer - including the ones where the honest answer is "you are right, and here is the trade".
- Said, drawn, then done. One idea reached three ways in the same hour beats three ideas reached once. When an hour is overfull we cut an idea rather than a channel.
- You leave with worksheets you filled in, not slides you were shown: the predictions you got wrong, the check you wrote, the plan you committed to.
- And it has a cost we will state up front. It is slower per topic, so the week cuts material rather than depth and we can tell you exactly what was cut and why. It needs a room under about a dozen people. And parts of it are deliberately uncomfortable, which we do not rescue early.
How it runs
- Scoped before it is quoted. A short conversation about who is in the room, what they already do, and what you want them able to do afterwards. The engineer who would teach it is the one who scopes it.
- On a real job. The week runs on a job your team actually does, which each attendee brings with them. Where something real cannot be shared we say so and use something that can.
- In person, five consecutive days. We come to you, or your team comes to us. One room, one week, nothing spread across afternoons.
- Small groups. Scoped per group rather than sold per seat, so the material can be aimed at the people actually in the room.
- You keep everything. The material, the exercises and the work your people produced stay with you, with no license to renew.
- It is not a funnel. Training is sold on its own, and nothing in it obliges you to buy anything else afterwards.