For firm owners and fractional CFOs. Four live weeks, small group.
Learn to build your own AI, safely. On tools we have already built.
Four live weeks in a small group. You get all of The Sprint. Then we build your first workflow with you, live, and make the hard calls in the file. Your team builds the rest the same way. You keep the code and all you build on it.
45 minutes on your real close. No access. No install. Five a month.
Leave knowing how to build your own AI, safely.
What you leave with is a practical understanding of how to build your own AI systems safely — on tools we have already built. You build on your own firm, with us in the room, on code we have already built and tested. You learn what to build, what to buy, and what to leave alone. By the end, one client’s work runs end to end. And you have watched each control work before it went near a client.
Live use still needs checksNothing runs on real client work until it passes the checks. The link, the data, safety, launch and release. Both paths keep live use behind them.
Your first workflow, built with us.
We build it with you, live, on The Sprint’s code. We make the hard calls in the file. Your team builds the rest the same way.
Software you keep and can change.
In your own repo, on your own hosting, on a model key that is yours. One firm, all the seats, for good. Nothing stops if you stop paying us.
Rules for running it.
Which work to automate, and which to leave alone. What a control must do before a client sees its output. What it costs a month. Who opens the approval queue on a Tuesday.
What your best person knows, on paper.
A context file per client. Your firm’s written rules for AI use. Ninety minutes of your most senior person’s head, in a form the whole firm can use.
The rules have a name: AI-1. Six layers, built in order. May we do this at all. What the workspace may ask for, and what it may never do. What it reads, what it writes, and what it costs. How you prove it refused. Who uses it, and who says yes. Upkeep: keys that expire, models that change, drift. The Sprint is the three layers that are code. The Bootcamp is all six, against your own firm. The layers that let you defend it are not the ones that make it work. Most programmes teach the second set. You can change the code, because you have it. A new model vendor or a new ledger still needs an adapter, and tests to prove it works. That work stays under your control.
Read first. Then build, live, each Tuesday.
Six chapters run across four live weeks. Every week has the same shape. You read the chapter before Tuesday. Tuesday’s session runs on that week’s page. Wednesday’s office hour fixes what broke, on your own code. Below are the chapters’ real promises: not things you learned, but things that exist when you finish.
Six chapters, one systemRead as a list, it looks like six pieces of homework. It is not. Each one is a layer of the same thing. Chapter six is where they join up: one client’s work, running end to end.
-
Chapter 1
AI and the way the 99% use it
By the end: a written rule for your firm on which AI model does which job. It rests on your own measured numbers, from your own real work. Not on anyone’s benchmark.
-
Chapter 2
Make it stop starting from zero
By the end: one context file per client. Each is tested on a task you have already graded. Plus ninety minutes of your most senior person’s head, on paper. In a form the whole firm can use.
This chapter alone pays for the programme. You count the times your firm stops because an answer lives in one person’s head. You price them. Then you halve the number twice. It still pays back inside a month.
-
Chapter 3
Let it read your real systems
By the end: a live link to a system that matters, and it can only read. Plus one automation that runs off it. It tells you a thing you would have gone and looked up.
-
Chapter 4
Controls, in the order that makes the argument
By the end: one automation drafts work close to the books. Its controls are ones a reviewer would accept. A spend monitor shows the cost before a bill comes.
-
Chapter 5
Vibe coding, and the gap you can’t see from here
By the end: a thing you built yourself, and it runs. Plus a costed answer to the one question that matters. What would have to be true before you let it near client data?
-
Chapter 6
Bringing it together, and watching it refuse
By the end: the refusal tests from chapter four now drive your own code. They do not skip past it. One client’s work runs end to end. No one carries anything between the steps by hand. And you write down the one thing your system still cannot say.
All of The Sprint. Then four live weeks with us.
A seat includes The Sprint in full. That is the code, the plan, the tests and the checks. On top of it, you get four live weeks with us: the sessions, the office hour, a kickoff call, and a small room of firm owners doing the same work.
What we never getWe never get your ledger keys. Our access to your repo ends when the cohort ends.
- All of The Sprint
- The full source code. The plan, 1,965 tests and the checks. It all runs on your own servers. Your workspace builds from it. What is in The Sprint →
- Four live sessions
- One each Tuesday, for four weeks. You read before each one and come prepared. The session is the framework. Your week page is where it runs. Then we build the first workflow with you and make the hard calls in the file.
- One office hour a week
- Wednesdays at 1:00 pm Central. Sixty minutes. One standing Teams link, nothing to book. Bring what broke that week.
- A kickoff call
- Thirty minutes, before week one. We set up your repo access on it. A private talk is there when you ask for one.
- One pull request, yours to merge
- We work in a private pull request on your code. You read it. You merge it. Nothing lands that you did not read.
- A small room
- Firm owners doing the same work as you. Seats are capped. We will tell you if The Sprint alone would serve you better.
What you bring
- Your own workspace
- Your own ChatGPT Work local project, with Work locally on. Or Claude Cowork, with the permanent project folder set. What you can use depends on the app, your plan, its settings and your access. Neither service is bundled.
- Your AI key comes later
- The first two steps, Quick Start and First Proof, need no model key. When the build reaches the AI step, we show you how to set up an approved account, with keys your firm controls.
- One client workflow you know well
- The Outcome Lab uses made-up data. First Proof uses a test account, not a live one. Real client data comes in last. Two checks must pass first: who may allow it, and which path the data takes.
- Time to work each week
- Chapter one is the lightest. Chapter five is the heaviest, and you do not have to finish it.
Who this is for, and who it is not.
The worked examples use a general ledger. That is the system we build against. The method does not care. A context file holds your firm’s views, its limits, and the work that needs a second set of eyes. It works the same for tax returns, workpapers and journal entries. The same control makes the workspace stop. You are buying the discipline, not the dataset.
It often tells you not to spendRunning models on your own hardware will not save you money. We do the sums to show it. A platform spend limit is sometimes only an alert. Even a hard one lags. The stop lives in your own code. QuickBooks Online has no read-only permission. The read-only you think you have, you have to build. Most of what your firm should automate needs no approval machinery at all.
It is for you if
- You run a firm. Or you lead a tax, audit, advisory or client accounting practice in one. Or you carry a book as a fractional CFO.
- You are liable when you are wrong. That shapes what you will accept.
- You are good with numbers and hard to sell to. An AI vendor has sold to you badly already.
- You want to know what to build, and what not to.
- You can make time each week for the work.
It is not for you if
- You want someone to do it for you. That is a build, and we will say so.
- You have never used ChatGPT or Claude. Book the free close teardown first. It takes 45 minutes.
- You want a certificate. We do not give one.
- You want to hear that AI will replace your staff. We argue at length that it will not.
Build it yourself. Or build it with us in the room.
Both start from the same tested seed. With The Sprint, your workspace builds the platform you chose. The Bootcamp adds four live weeks with us. Buy The Sprint on its own first if you like. Its full price comes off a seat later. So there is no added cost if you want the room after all.
Next cohortThe next cohort starts 13 October. Seats are capped. We will tell you if The Sprint alone would serve you better.
Build the first workflow with us in the room. Then build the rest yourself.
The Bootcamp
Full refund for two weeks from the first session.
- You get all of The Sprint.
- We build the first workflow with you.
- We make the hard calls in the file.
- Your workspace builds the rest the same way.
- Four live sessions, one each Tuesday.
- One office hour a week, on your own code. A private talk when you ask for one.
- A private pull request. You read it and merge it.
- We never get your ledger keys. The cohort ends. So does our access to your repo.
- Your firm keeps the code and all work built on it.
- Your firm’s know-how on paper. A context file per client. Your written rules for AI use.
- A small room of firm owners doing the same work.
The refund. Work through the first two weeks. If nothing is running that was not running before, we refund it all. Ask within two weeks of the first session.
Not sure? Book a close teardown → It is free, takes 45 minutes, and asks nothing of you.
Build the platform yourself. Use your own workspace.
The Sprint
Its full price comes off a seat later.
- You get The Sprint’s full source code.
- Your workspace builds the suite, add-ons, screens and AI features you chose.
- It tests and commits its own work.
- You get the plan, 1,965 tests, and the checks.
- It all runs on your own servers.
- Live use still needs checks. The link, the data, safety and release.
- It does not touch your books or data without your say. Nor your paid AI account.
- You keep the code for good. If we publish a fix, each buyer gets it.
Stripe takes you to the zip, the start guide and the build order. There is no email wait. See what is in the box →
The ones worth asking.
More answersFirms ask us the same things. We keep a page of them, on The Sprint and The Bootcamp. Read them →
Do I need to be technical?
No. The material assumes you are not. It does assume you are good with numbers. And that you know controls: what counts as material, and why the one who prepares is never the one who approves. Those are the ideas the whole thing is built on. Engineers learn this year what your field has known for a hundred years.
Will this be out of date in six months?
Parts of it. We say so on the page. Every price and product claim carries the date we checked it. That is the main reason it is web pages and not a PDF. Updates are easy and you keep access. The method will not go stale: measure on your own work, write decisions down, read-only first, and the control order in chapter four.
What does the Bootcamp add to The Sprint, really?
Not files. A seat includes all of The Sprint. With The Sprint, your workspace builds the platform you chose. With the Bootcamp, we build the first workflow with you. We make the hard calls in the file. Your workspace builds the rest the same way. You get the group calls, a weekly office hour on your own repo, and a room of firm owners. Both paths keep live use behind the same checks. If you want the room, take a seat. If you would rather build on your own, The Sprint is the honest choice. Its full price comes off a seat later.
Do I have to link a client’s live systems?
Not until chapter three. And not at all, if you would rather not. Chapter three covers what your engagement letter says about sending client data to an outside vendor. It covers what your standards say too. That includes the parts most firms have never asked about. You can do that chapter on your own practice data instead.
What if I finish and want to buy instead?
That is a fair answer, reached the right way. Chapter five has a section on that exact choice. If you finish and decide not to build, you did not fail. You did the work. And you can now question whatever you are shown far better than you could at the start.
And if you end up wanting it built, rather than building it yourself, that is the build. One month-end for the whole firm. Four weeks. Your firm keeps the code. It is scoped and quoted for your firm. A seat is the honest way to find out whether you want it.
Is there a refund?
Yes. Work through the first two weeks. If nothing is running that was not running before, we refund it all. The window is two weeks from the start of the first session. So you decide after you have seen the first two live weeks, not on faith. Email info@byram-advisory.com and it is done.
The only thing we ask in return is a phone call. The refund does not depend on it. No one will try to talk you out of it. We want to know what you expected and did not get. Right now that is worth more to us than the price. Every one of those calls makes the next version better.
If you would rather settle the fit question first, that is what the close teardown is for. 45 minutes, free, and you leave with one written page either way.
Take a seat. Or bring your close first.
Seats are capped. If you are not sure it fits, start where it costs you nothing. The close teardown is free. It is 45 minutes on your own close. You leave with one written page. If that is not useful, none of the rest will be either.
Not sure which?We will tell you if The Sprint alone would serve you better. What is in The Sprint →