AI at work — building things that survive you

What it looks like

A narrow window you snap beside whatever AI you are using. The lesson and the tutor are one conversation, and the strip along the bottom always says what you were doing — because you will spend most of your time in the other window.

The course panel: module and progress at the top, the lesson text in the middle, and a task strip along the bottom reading 'Read this section — Done, next'. — A real section, photographed from the running course. Open it and have a read — no sign-up.

This has happened three times before

Spreadsheets. Departmental databases and macros. Low-code. Each one let people build without waiting for IT, and each one was genuinely good for the organisations that lived through it.

Each one also ended the same way: something the business depended on, that worked, that nobody could explain, built by someone who had since left. Every finance department has found one. Several public inquiries have turned on one.

AI is the fourth wave, and it is faster because it removed the last friction — you no longer have to learn the tool's language at all. What has not got faster is the time it takes to understand what you have built.

Which of these are you?

You run the business

Somewhere in your organisation, something important is running on a tool one person built in an afternoon. You do not know which tool, or which person. You will find out when they leave, or when it is wrong and nobody notices for a month.

This course does not stop your people building. It makes what they build something you still own in three years.

You manage a team

Your people are already doing this. Banning it does not work and would cost you the gains. The realistic option is that they do it with a method — one that produces a specification, a record of decisions, and a handover pack, without slowing them down.

Everything here is designed for someone with no engineering organisation behind them.

You are the one building

You can already produce more in an afternoon than you can properly understand in a week. That is not a criticism — it is the actual problem, and nobody has told you it is a known one with a known shape.

The skill that will still be worth something is not building faster. Everyone can build faster now.

What it costs to skip this

Stated plainly, because the alternative is fear and fear is not useful.

The key-person problem. One person understands the thing. When they go, the knowledge goes, and what remains is a system nobody dares change.
Silent wrongness. Most quickly-built systems have no way of telling you they have started producing wrong answers. You find out from a customer.
Data in tools nobody chose. Company material in personal accounts, under consumer terms, because it was what someone had open.
The dependency nobody registered. It is not in the disaster recovery plan, because nobody knew it existed.
The direction of regulation. The EU AI Act phases through 2026 and 2027, and sector regulators are issuing their own guidance. Being able to explain what your automated systems do is moving from good practice towards an expectation.

What you actually get

Nine modules, about seven and a half hours, in sittings of ten to fifteen minutes. It sits in a window beside your own AI window while you do real work.

Before you build anything — what you already do, what you are driving, and one conversation done properly. free
What this technology actually is — how it works, how we got here, what it can be trusted with, which system for which job, and what you must never feed it. free
Driving Claude — the workbench, why long conversations degrade, how much rope to give it, and connecting it to your systems.
Frame it — should this exist at all? Ten minutes that kill bad projects before they cost anything.
Discovery — being interrogated about your own work, and the list of things your system must never do.
Spec, tests and foundations — the load-bearing stage. Includes version control and house rules from scratch, for people who have never used either.
The slice loop — one thin slice at a time, verified, and reverted rather than patched.
The launch gate — going live changes the blast radius, so it gets its own gate.
Run and keep true — five minutes a week, and the handover pack that means it survives you.

You do the work in your own tools, on your own systems. The course reads what you produce and marks it against explicit criteria. There are five pieces of work, and they are about something you actually build.

What you get at the end

A record of what you actually did — the modules, the work you submitted and had marked, the dates — that anyone can verify.

A printed certificate on cream stock with a green and gold border, guilloche pattern, seal and signature. It reads: Record of completion. A. Learner completed The Platinum Build Method. 9 modules, 38 sections, 5 pieces of work submitted and marked, with dates, curriculum version and a verification reference. — See it full size.

It lists the work you did and the dates you did it, and every one carries a reference anyone can check. Facts, in other words, rather than an opinion about you.

The honest bits

What you need. Chrome or Edge on a computer, and an AI account you are permitted to use for work. The folder features do not work in Safari or Firefox, and a phone is read-and-review only. You cannot build a system on a phone and we are not going to pretend otherwise.

Who this is not for. Professional software engineers. You already have the disciplines this teaches; you would be bored. This is for people who build things at work without an engineering organisation around them.

Who wrote this

Paul Bappoo. Thirty years building systems inside other people's organisations — SunSystems and Maximo implementations, integration work, and a long run of being the person called in afterwards.

That second part is what this course is made of. A recurring job through the 2000s was arriving at an organisation running on something a department had built years earlier, which worked, which nobody could explain, and whose author had left. The technology changed three times over those decades. The ending did not.

This is not a course about AI. It is a course about the thing that goes wrong after the building is done, taught by someone who has been paid to clear it up.

Questions people actually ask

Do I need to be able to code?: No. It is written for people who build things at work without an engineering background. You will read code the AI writes and decide whether to accept it — that is a different skill from writing it, and the course teaches it.
My employer has locked down AI tools. Is this still worth it?: Most of it, yes. Module 1 covers what you may and may not put into these tools and how to find out your employer's actual position, which is often "nobody has decided". Two later sections need software you can install; they say so plainly rather than failing oddly.
Can I expense it?: Usually, yes. You get an itemised VAT receipt and a completion record with a verifiable reference. If your finance team needs something in a particular format, email and ask.
How long does it take?: About seven and a half hours of material, designed in sittings of ten to fifteen minutes because nobody does this in one go. Most people take two or three weeks.
How long do I have access?: Twelve months from purchase. That is a live service — a tutor that answers you and marks your work — and it costs money to run every day it is open, so it runs for a term rather than forever. Twelve months is far longer than the course takes.
What happens when the twelve months are up?: Your record and your certificate are permanent — the verification link keeps working, and you can still export everything you did as a single file, at any point, without asking anyone. It is the course content and the tutor that close.
Is the AI marking my work?: Yes, against criteria published to you before you submit — you can see exactly what it is looking for. The AI judges each criterion; the pass or fail is decided by ordinary code from those judgements, not by the AI. That distinction is one of the things the course teaches, and it would be odd to teach it and not do it.
Will this be out of date in six months?: The parts that describe software will be, so those are kept in one module, dated on the page, and replaced when they change. The method does not date — it has survived three previous waves of this.
Refunds?: Two modules are free and complete, which is about two hours. Do those first and you will know. If you pay and it is not what you expected, email within 30 days and it is refunded.

Start with the free part

The first two modules are free and complete — about two hours, nothing held back. Finish them and you will know whether the rest is worth it to you.

$499

One payment, twelve months of access, when you decide to carry on. For comparison, instructor-led AI training runs into the thousands per person, per course.