Quotes and contracts with AI: how I do it without getting burnt

Writing a detailed quote used to take me a good hour. Today it takes fifteen minutes, and the result is better. But the saving doesn't come from asking AI to "write a quote": it comes from giving it the right material and knowing what never to delegate.

How I ask for a quote

What I always provide:

• One of my previous quotes as a model. It's the step that changes the result most: without an example it sounds like a brochure, with one it sounds like me.
• The real scope, with what's in and what isn't.
• The hours and the total. AI writes the breakdown; I set the numbers.
• Context on the client: sector, size and what they're worried about.

Then I review the whole thing. Always. The most common corrections: sums that don't add up, optimistic timescales and surplus adjectives.

With contracts, more care

Here AI is useful for redrafting clauses you already have, organising an agreement and spotting what's missing (intellectual property, payment terms, scope, what happens if the client vanishes). It isn't useful for inventing a contract from nothing.

A legal text generated without a base can sound impeccable and not hold up. If the contract matters, start from a lawyer's template and use AI only to adapt it and to ask what you're not seeing.

What I never give it

• Clients' personal data or confidential documentation.
• Third-party pricing or terms from agreements that aren't mine.
• The decision. What to charge is business judgement, and AI will hand you a reasonable average, which is almost never your price.

The workflow that works for me

1. Four notes from the project, exactly as they came out of the meeting.
2. A previous quote as an example + scope + amount.
3. Generated draft.
4. My review: numbers, timescales, tone and what's surplus.
5. Send, and save the final version as the example for next time.

That last step is what makes it cheaper every time: your folder of examples is the advantage, not the model you use.