How we do it

Most guides on this topic are written from the outside, by people describing a thing they do not run. This one is not. We run a persistent context system of our own, a second brain for the business, and it sits behind the tools we build. So rather than tell you what a brain is in theory, here is what it is actually like to run one, warts and all.

Two honest notes first. This is our own setup and how we use it, not a product we sell you. And nothing here needs a big budget or a technical team. The principles are identical whether your brain is a polished system or a well-kept folder of documents. The value is in the discipline, not the software.

What it actually is

Strip away the name and ours is a single, organised, always-available store of everything the AI needs to answer and act as our business would. The same four kinds of fact you have been writing down: our prices and how we scope work, our policies and how we run a job from enquiry to sign-off, the way we write to people, and the questions that come up again and again with the answers we actually give.

When we ask the AI to draft something or help with a task, it works from that, not from a blank slate. It is the difference between briefing a colleague who knows the business and briefing a stranger every morning. The “persistent” part is the point: it does not forget between conversations. What we taught it last month is still there this month.

What we learned putting things in

The first lesson was about what belongs and what does not. Good context is specific and stable: prices, policies, product details, tone, standard answers. Things true across many jobs that do not change by the hour. The temptation early on is to pour everything in, on the theory that more must be better. It is not. A brain stuffed with one-off details and half-finished thoughts that contradict each other makes answers worse, because the AI cannot tell what still matters. We got more value from a smaller, cleaner brain than a bigger, messier one.

The second lesson: write it the way you would explain it to a sharp new hire. The clearer we wrote the source, the better the AI performed, every single time.

The part nobody warns you about

If there is one thing to take from how we run ours, it is this. Setting a brain up is the easy, satisfying part. Keeping it true is the real job, and it never fully ends.

A business changes. Prices move, policies get revised, you drop one service and add another. Every one of those has to make it back into the brain, or the AI keeps confidently answering with yesterday’s facts. And a wrong answer delivered with total confidence is worse than no answer, because you might not catch it before a customer does. A stale brain is a genuine liability, not a neutral one. We treat maintenance as a standing habit, not a project with an end date. We go deeper on exactly how in the next lesson.

Why we bother

Here is the part that surprised us most. Once the context was solid and trusted, building things on top of it got dramatically faster. Every tool we make, from quoting to job scheduling to smaller helpers, would otherwise have to be told about the business from scratch. With the brain in place, that knowledge is already there. We stopped re-explaining who we are at the start of every build and started from a running position instead.

That compounds. The first tool is the hard one. The tenth is far easier, because it inherits the same shared understanding of prices, policies and voice that everything else uses. A good brain is not just a better chat assistant, it becomes the foundation the rest of your AI work stands on. That is the honest reason we invested in it, and the reason we now build the way we do. If you want this told as a standalone piece, it is written up in How we run our own AI brain.