Why does AI need your business context?

A short story

A family-owned machine builder gives its sales team an AI assistant to draft quotes faster. The first drafts look impressive. Then a quote goes out with the standard delivery time for a machine that, as the engineering lead knows, always needs a custom frame. Nobody had told the assistant, because the rule was never written down anywhere. The assistant was not wrong about machines. It was wrong about this company. (An illustrative example, not a client case.)

What goes wrong without the context?

• The first weeks: impressive drafts and fast answers on the standard cases.
• The first month: the first errors, always on the exceptions only the experts know.
• The second month: people double-check everything, and the time savings disappear.
• The third month: “AI doesn't work for us.” The pilot stalls — while the real gap was missing knowledge.

What does “business context” mean in practice?

• your core processes, and who owns each one;
• the decision rules and exceptions people apply;
• the history of customers, products and installations;
• the reasons behind past decisions;
• where each piece of knowledge comes from, and when it was confirmed.

Most of it is tacit knowledge: known, used every day, and written down nowhere — which is also why documentation alone rarely captures it.

How MentX captures that context

MentX captures how your company really works: the knowledge of your key people and the processes they run. We offer both a hands-on service approach — mapping your processes and capturing the knowledge behind them, from meetings, workshops and a guided intake — and our own personality-aware knowledge base, with full source attribution, in your own Knowledge Space. We build on the tools you already use and add best-practice solutions where they fit.

The four steps build the context AI needs: capture the knowledge of your key people; map your processes; capture your operations — owners, decision rules, exceptions and data, with their sources; and run on it with operational excellence, where AI assists with sources and people decide.

What if you already use AI tools?

Keep them. The problem is rarely the tool; it is what the tool does not know. An assistant in your office software, a chatbot on your documents or an AI feature in your ERP all get better when the knowledge they work from is complete, current and sourced. Capturing your business context is not a choice against the tools you have — it is what makes them useful for your company.

A good test: ask your AI tool a question only your most experienced colleague can answer, then ask that colleague. The difference between the two answers is the context you still need to capture.

Why can't the AI just learn it from our files?

Files help, and they are part of the capture. But files record decisions, not the reasoning behind them; they rarely say which exception applies; and they contradict each other over the years. An AI reading everything at once cannot tell which version is current or which rule an expert would apply today. That judgement has to come from the people who hold it, with a date and a source attached.

What to do this quarter

1. Pick one process where AI would help most.
2. List the decisions in it that only one or two people can take.
3. Write down the rules and exceptions of one of those decisions, together with the people who apply them.
4. Check whether the data that decision needs is available, and who may see it.
5. Only then test an AI tool on that process, and check its answers against the rules you captured.

Want a structured way to check a process first? Ask us for our AI-readiness checklist through the button below.