The AI Executive · a mini-MBA in AI, without the code
The work that eats your week is the same work every week.
You run a business. The repetitive part of it — the invoices, the inbox, the same answer typed for the ninth time — is the part a model is genuinely good at. This is the sequence that hands it over. You come to the classes, the recordings are there for the weeks the business wins, and you never open a terminal to do any of it.
Your week, as a ledger
Cost the repetition before you buy anything
Tick the chores you or your staff actually do, put real hours against them, and the sheet totals what they are worth a year. Every figure in it is yours to change, and the last column is the class that hands that chore over. Your inputs stay in this browser: the sheet adds up here, and the figure it reaches is saved in it.
Hours per week, your figures
| Operation | Hrs / wk | Addressed by |
|---|---|---|
| Automate Your Email operate a no code automation that calls a model | ||
| Automate Your Email produce a reusable system prompt | ||
| Custom GPTs design a packaged assistant over your documents | ||
| Context Engineering design a recursive summarizer for oversized documents |
Both are yours to set. 48 weeks assumes four off; say 44 or 52 and the sheet follows.
Arithmetic, not a forecast
That share is yours to set. Nobody here has seen your invoices. Set it low, then argue with yourself.
Which class is on the hook
- Automate Your Email: No Code AI Automation 7.0 h/wk
- Custom GPTs 3.0 h/wk
- Context Engineering 2.0 h/wk
There is a great deal here to work through — an estimated 58.6 hours of recorded sessions to work through, none of it on somebody else's timetable. At the recovery you set, the time it gives back covers the time it takes in 45 weeks, and every week after that is profit.
In money it is $250 a month — about 6.2 hours of your own time at the rate you just entered. That is the whole path recorded, with the standups every week.
Whatever that came out at, the six classes that hand it over are further down this page, and $250 a month opens all six of them: the whole path on your own clock, and the standups every week. For $500 you are in the room while they run, starting with Automate Your Email: No Code AI Automation on 4 September.
Join us $250/monthGovern AI safely for your team
$400 suggested / $300 floor
Own your team's infrastructure
$750 suggested / $500 floor
Add on anytime — keeps your current path. One-time purchase, classes yours to keep.
Put it in the diary
The next one is 4 September
Every class on this path runs live, and $500 a month is a seat in all of them — the same seat next time it runs, and the time after that. You bring your own invoices, your own supplier who does not pay, your own nine replies, and you get an answer about those rather than about the example.
Then the week goes wrong, because it does: someone calls, a delivery is late, the person covering the counter does not turn up. Every session is filmed, so the class you missed is a recording you play at the moment you are actually stuck — and 58.6 hours are already in there before you attend anything at all.
$250 a month is the filmed half — every session, every exercise, the written curriculum, and the standups every week to bring the stuck thing to. $500 is both halves: the same, plus your chair in the room while it runs.
See what is in it ↓Filmed for one tier, your chair for the other
- 4Sep Automate Your Email: No Code AI Automation Friday
- 12Sep Control AI Spending Saturday
- 16Sep Context Engineering Wednesday
Hours
- Control AI Spending 2 · 3.1 h
- Automate Your Email: No Code AI Automation 8 · 10.8 h
- Custom GPTs 7 · 9.0 h
- Context Engineering 27 · 35.7 h
Before you begin
Everything this asks you to have
You are the person who has to keep the thing running on a Tuesday when nobody is available to help. So everything you build here, you build in something you can open, read and fix yourself.
The whole list
- A computer and a browser Every tool on this path opens in a tab. The automation step is a builder you drag things around in; the assistant step is a form and a set of uploaded documents.
- An account with a model provider The models cost money to run, which is why the first class is about pricing a job before you point it at a thousand of them.
- Your own documents Your invoices, your policies, the nine replies you could write in your sleep. The exercises are worked on those rather than on a toy example.
- Room to install one thing Step one has you running an open model on your own machine — an installer and a download. It is how you handle the documents you would rather not upload anywhere.
- An hour, in the room or on the recording Come to the session and ask about your own invoices, or take it back in the gaps — twenty minutes before you open, an hour after the till is counted, the whole of a slow Sunday.
Not once, across the six
- A programming language Nothing on this path is written in one, and nothing you build here has to be handed to somebody who can read one.
- A repository No git, no branches, no pull requests. Your work lives in the tool that runs it, which is also where you edit it when it goes wrong at 6pm.
- A terminal The local model is an installer you double-click. Everything after it is a browser tab and your own files.
- Somebody technical on call What you build, you build in the tool that runs it, so the person who opens it up when it misbehaves is you.
- A clean run of free Tuesdays Every session is filmed as it runs, so a week the business wins costs you the room and not the class — you take that one back at 11pm, and you are in the next session with everybody else.
The sequence
Six classes — short enough that you will actually finish it
Costing before building, building before packaging, packaging before scale, and then the two that stop somebody taking it all off you. Each step stands on the one before it, and by the last one you are designing the system rather than following somebody else's.
-
01
Control AI Spending
- run an open model locally
- produce an inventory of what you pay for and what it holds
- produce an itemised bill for what your running system costs
- segment text into tokens and cost it
- characterise what a tool costs you beyond its price
- characterise what one user costs you to serve
- classify each recurring ai cost by keep downgrade or replace
- classify models by fit for a workload
- verify a spending limit declines the charge you did not intend
Why it is first: everything after it costs money per token, and a business owner should be able to price a job before automating it. You leave able to look at a task and say what it will cost to run a thousand times.
-
02
Automate Your Email: No Code AI Automation
- operate a hosted assistant over uploaded documents
- operate a no code automation that calls a model
- produce a reusable system prompt
- produce an itemised bill for what your running system costs
- segment text into tokens and cost it
- transform a transcript into a structured record
- transform an incoming message into ask deadline and claimed sender
- characterise what one user costs you to serve
- characterise where a teams work stalls
- constrain model output to a schema
- parameterise an agent persona
- measure a delegations outcome against what was asked
- measure whether a user pays more than they cost
- verify you would know your system broke before a user tells you
- design a system you can afford to keep running
- reconcile agents from different toolchains into one run
The workhorse. 10.8 recorded hours, and it is the class most of the ledger above points at — the invoicing, the triage, the scheduling, the data entry. You build the automation in a drag-and-drop builder and it runs whether or not you are at your desk.
-
03
Custom GPTs
- configure an assistants memory and outside connections
- operate a hosted assistant over uploaded documents
- produce a rubric a model can apply
- generalise a tool interface for model use
- parameterise a prompt template
- parameterise an agent persona
- verify a rubric against independent graders
- design a packaged assistant over your documents
- design an interaction protocol for responsible use
Where it becomes a thing you own: 9.0 recorded hours. This is where you stop prompting and start packaging — one assistant that knows your policies, your prices and your standard answers, that your staff and your customers can use without you in the room.
-
04
Context Engineering
- follow a model quickstart
- execute a tool call round trip by hand
- configure inference hyperparameters
- elicit output from a model
- express a data shape as a schema
- produce a refusal boundary that fires on cases you did not list
- produce a reusable system prompt
- produce a stopping condition an agent can check on itself
- segment text into tokens and cost it
- transform a corpus into an embedded index
- characterise what one user costs you to serve
- classify models by fit for a workload
- constrain model output to a schema
- generalise a reasoning prompt pattern
- generalise a tool interface for model use
- parameterise a prompt template
- falsify a models alignment guardrails
- falsify a prompt with a benchmark
- justify a vector index for a latency and corpus budget
- justify an embedding model and its dimensionality
- measure a models value alignment against a stated standard
- measure whether a fine tune changed behaviour
- measure whether a user pays more than they cost
- verify a model output you thought was random
- verify you would know your system broke before a user tells you
- construct a multi agent conversation with turn taking
- construct a retrieval system
- construct an agent that uses tools
- design a chunking strategy against measured retrieval
- design a packaged assistant over your documents
- design a recursive summariser for oversized documents
- design a system you can afford to keep running
- govern a swarm by information hierarchy
- re architect an agent as a stateless reducer
- reconcile a failing retrieval by reranking and rewriting
- reconcile a working set with the context window it must fit
- reconcile an agents run with a tool that failed
- select among agent architectures
- synthesise a memory hierarchy for an agent
- synthesise an agent that carries notes across its own runs
The heavy one, and the guided one. 35.7 of the path's 58.6 recorded hours sit in this one class, worked through the only fully guided workbench on the path — it walks you a step at a time and remembers where you stopped. Plan to live here for a while.
The two nobody expects
A pretext invoice will cost you more than the inbox ever did
Two of the six are security classes, and they may be the two that save you the most money. Business email compromise and account takeover are how a small business actually loses a large sum at once: someone emails an invoice that looks right, or takes over a login and waits. Those sums are bigger than anything you save by automating your calendar.
They also make the rest of the path safe to run fast. Once your business acts on messages without a person reading every one, the person who knows a forged instruction on sight is you — and by the end of these two classes you have practiced on your own accounts, with your own suppliers, until you can spot one cold.
Both are worked through Six Roses — a checklist you run against your own accounts, your own suppliers and your own recovery options, rather than a video you watch.
- transform an incoming message into ask deadline and claimed sender
What is actually in the box
Counted, not rounded up
A seat in all six, and 58.6 hours already filmed
58.6 hours of recorded sessions waiting the day you join, and the room itself whenever a class runs — starting with Automate Your Email: No Code AI Automation on 4 September.
Watch them in order or go straight to the chore that is costing you the most.
Hands-on exercises
Weighted where the work is: most of them in Context Engineering and the automation class, the rest across spending and assistants. You do them on your own business, not on a toy example.
The tools you work in
- The automation blueprint — the reference you keep open while you wire up the first automation that really sends mail
- The Context Workbench — the fully guided one — you work the exercises inside it rather than reading about them
Yours from the first day, for as long as you are with us.
And if you want the room as well
Everything above is yours at $250 a month. $500 a month adds the part that cannot be filmed: your own operation on the table, out loud, with the person teaching it looking at it — next on 4 September.
Two doors down the hall
Two people want the next ledger over
You want to build the systems yourself
If the appeal is the machinery — writing the agent, owning the code, running it on your own infrastructure — there is a path that starts at the terminal and goes all the way down. This one starts at your invoices.
Take The AI Builder →You are starting something, not running something
This sequence assumes there is already an operation with repetition in it. If nothing exists yet, you want the ordering that puts the offer before the automation — and it starts by naming what a month of your life costs.
Take The Founder →Same school either way, and the standups are the same room.
The close
Hand the repetition over
4 classes, 58.6 recorded hours, and the tools to do the work — plus the standups every week, so there is a room to bring it to even in the weeks the business wins. Want the chair as well? The next session is 4 September. No code anywhere in it, and two classes near the end that make sure nobody walks off with what you built.
$250 a month for the recordings, the exercises and the standups; $500 to be in the rooms as well. Cancel either yourself, any time. Or take a single class, live, at its own price if you would rather start with one — the automation class is where most of the ledger above points.