AI coaching for CEOs
one-on-one · one day, then two weeks of follow-up · day rate on consulting
Most of what gets sold to executives as AI training is a talk about where things are heading. This is a day working on your own material instead: the board pack, the investor update, whatever decision has been sitting on your desk. Far Horizons sets the tools up on your accounts and you use them on real work before the day is out. The aim is that you're still using them a month later.
01
Start with your desk
morning · ninety minutes
The morning starts with what actually comes across your desk in a normal week: the board pack, the investor update, the hiring call, the engineering estimate you're not sure you believe. The rest of the day is built around that list rather than a syllabus.
02
Set it up on real work
afternoon · on your own accounts
Claude and ChatGPT on your own accounts, pointed at your own documents. Voice typing, because you'll say far more than you'll ever type, and the models do much better with more context. Then a review step: one model drafts, and a second one, on a different model, pushes back from the point of view of your CFO, your head of engineering or an investor. When two different models agree on a problem, it's usually a real problem.
03
Two weeks after
async · email, then a call
Send through the first real things you run, the memo or the decision, along with the prompt and what came back. Far Horizons goes through them, points out where the model was confidently wrong, and adjusts the setup. There is a call at the end of the two weeks.
How Far Horizons works
Before Far Horizons builds anything, the plan goes past a set of reviews: a CEO review, an engineering review and a design review, each run on both Claude and Codex. On one engagement the plan was to move an API from GraphQL to REST, and both CEO reviews pushed back hard, because developers not liking GraphQL isn't a good enough reason to spend months on it. They were right. The same approach works on the decisions a CEO makes, and it takes an afternoon to set up.
Who leads it
Far Horizons is led by Luke Chadwick, who has been shipping software since 2001, has built production LLM systems since 2022 for Scout24, AutoScout24 and others, and still writes code with agents every day. There's more on the about page.
Before you write
I already have a CTO or a head of data on this. Why do I need it?
They're building AI into the product and the company. This is about your own work, the reading, writing and deciding that only the CEO does. CTOs are usually pretty happy when the CEO has a realistic sense of what the tools can and can't do, because it makes the conversations about estimates and vendors a lot easier.
Which model should I be using?
Usually both Claude and ChatGPT. Which one is ahead changes every few months, and the practice uses several every day. The habits last longer than the model choice: give it the real context, get a second model to check the first, and don't take a confident number at face value.
What about confidential board material?
The first hour covers what can go into which tool, on which plan, with which settings. Nothing goes into a model that you haven't decided should.
What's the first thing I should change?
Record your meetings. Most leadership teams spend three hours discussing something and the only record is in people's heads, and then somebody has to write it up. A transcript turns into the actions, the memo or a first draft of the plan in a few minutes.
We want this for the whole leadership team, or for the engineers.
That is a different engagement. For engineers it is team enablement, two to five days working on your repo. A leadership team usually starts with a conversation about what the company is actually trying to do with AI.
Write to hello@farhorizons.io with the two or three things on your desk you'd most like help with. You'll hear back within a business day. If it turns out your engineers need this more than you do, you'll be told, and point you at team enablement.