The quality of an AI assistant's answers depends largely on what it knows at the moment you ask. An AI asked for advice on price without being given the company's real hourly cost can only answer with general principles — useful, but not specific to the situation in front of you.
Instructions and context are different things
Instructions tell the AI how to behave — tone, style, what to do and what to avoid. Context supplies the data it reasons over: the true hourly cost, the customer history, the margins applied in the past. Without context, even the best instructions produce answers that sound plausible but are disconnected from the reality of the business.
Context that is current, not frozen
Context written once and never refreshed ages quickly: an hourly cost that has changed, a new customer, a revised target margin. A system that automatically injects up-to-date data into every conversation keeps the answers aligned with reality, without anyone having to rewrite the instructions each time something moves.
The line to hold
More context does not mean handing the AI decisions that remain human: context informs, it does not replace the judgement of whoever decides. A well-contextualised AI advises with real knowledge of the situation; the final decision, particularly on price and negotiation, always stays with the person running the business.