On 22 August 2026 I stood on the AI Connect stage and spoke about something I get asked in almost every conversation: how do you actually give a business an AI brain, rather than another subscription nobody opens after week two. The recording runs 13:29.
The recording is below. This article is the written version, because a talk disappears and a page does not. I am Egils Boitmanis, Chair of the Board of FullDigital, a Latvian nonprofit association. This article and FullDigital's public resources are free.
A chat window is not a business brain.
Most companies I meet have the same setup. Several people pay for a chat tool. Each of them has learned a few tricks. None of the tricks survive the person going on holiday, and none of them know anything about the company. The model is capable. The situation around it is empty.
A chat window starts every conversation from zero. It does not know your clients, your pricing logic, your last three decisions or the reason you rejected a supplier in March. So every useful answer has to be assembled by hand, by a human, from memory, every single time. That is not a brain. That is a very fast stranger.
The shift that makes the difference is boring to say out loud: write down the context, keep it in files you own, and let the AI read those files. That is it. The whole talk was thirteen minutes of unpacking that one sentence.
Three things have to be true before any of it works.
Context. The AI needs to know what the business is, who it serves, what it refuses to do, and what good work looks like here. Nobody has ever written that down in one place. Until somebody does, every prompt is a re-explanation.
Files. Plain text and Markdown, in a folder, in your own storage. Not inside one vendor's database. Files can be read by any model you choose today and by whatever replaces it next year. A knowledge base locked inside a product is a rental agreement with extra steps.
Control. A human decides what the system may read, what it may write, and what it may never do without approval. On stage I only had time to name this. It matters enough that it gets its own article later this month.
“The model is not the hard part. The context around it is.”
Five layers, built in order.
At FullDigital we teach this as five layers. It is an educational model, not a product. Each layer is useless without the one before it, which is exactly why so many AI projects stall: people jump to layer four on day one.
- Constitution. One file that says who you are, what you value and how the AI should behave. Read the layer.
- Memory. The second brain: notes, decisions, clients, projects, in files. Read the layer.
- Skills. Named jobs the AI knows how to run, written down once. Read the layer.
- Tools. Connections to the systems where the real work lives. Read the layer.
- Focus. The weekly loop that keeps all of it current. Read the layer.
In the room, the question that came back most often was some version of "where do I start". The answer is always layer one, and it takes an evening, not a quarter.
If you want the practical next step rather than the theory, the free starter kit walks through the first evening: the folder, the constitution, the first two skills. No signup, no payment.
Thanks to the AI Connect organisers for the stage. If your organisation needs a deeper paid company implementation than a nonprofit can responsibly provide, that work continues at AreteAIOS.com.
Buying tools before writing anything down
A new tool on top of undocumented context gives you faster confusion. Write the constitution first, then connect things to it.
Entrepreneur from Latvia. Building FullDigital so people stay free in the age of AI. I write about what actually works in my own week, not what sounds good on a conference stage.
Connect on LinkedIn