Why AI needs your context more than a better model

By Devan Borsboom

Every new chat starts from zero

Every time you open a new chat, you are talking to a brilliant adult with short-term memory loss. Plenty of reasoning, and no memory of your practice, your clients or how you like things done. That is true no matter which AI you use.

AI without data is useless. AI with data is powerful.

The models have been smart enough for most office work for a while now. What they need is the right information. When I built a presentation for a meeting, I gave the AI every call, email and note I had with that person, then asked for something professional. The result was good because the context was good.

Set it up once and short requests start to work. When the AI already knows where your procedures, past examples and preferences live, a one-sentence request can produce work that would otherwise need a page of instructions.

More is not always better

AI can reach a huge amount of information, but it can only work with a limited amount at once, a bit like a computer's memory. The skill is giving it exactly what the task needs, and no more.

Connect it to your tools

An AI in a chat window can write you a file. It cannot go and do the thing. Its real value comes from its reasoning, plus the right information, plus what it can actually reach: your email, your calendar, your practice software.

The order I use to connect a tool:

1. The tool's official connector, if it has one.
2. The tool's own programming interface, if it does not.
3. Working through a web browser, as a last resort, because it is the slowest and least reliable way for AI to do work.

Before you ask which AI to buy, ask what it will be able to see and reach in your practice.