Sam Davis · Sydney · October 2026Bespoke proposals often take far longer than they should, because most of what goes into them gets rebuilt by hand, even though you’ve written it before. Give AI your best past proposals, your pricing rules and the notes from the client call, and it can assemble a first draft that you then make bespoke.
The slow part is putting old pieces back together
The call went well on Tuesday. You know what the client needs. It’s Thursday night and you’re still hunting for the last proposal that was close, copying its method section, rewriting the outcomes, swapping the case study for one that fits better, then working out the quote from rates you keep in your head.
Most of it is the method you used last month, reworded for someone new.
People tend to avoid templates because clients can tell when they’ve been sent one, and the work they sell depends on it fitting. So the evenings go on making each one feel bespoke. Quotes go the same way: if the formula lives in your head, each quote tends to get rebuilt from scratch.
Plain Claude can do most of this today
You don’t need any special product to start. A Claude Project (ChatGPT has a similar feature, also called Projects) holds documents and instructions that every chat inside it can see.
- Feed it your winners. Pick the three proposals you’re proudest of, ideally ones that won. Ask Claude to split them into reusable sections: who you are, your method, outcomes, case studies, pricing, terms. Save those sections in the Project.
- Write your pricing down in sentences. “Half-day workshop: this rate. Each extra session: this. Travel outside the city: this.” Add three past quotes as examples.
- Bring the call in. Record the discovery call with the client’s OK, or write your notes straight after. Paste the transcript and ask: “Draft a proposal for this client from my sections. Use their words for their problem. Only use case studies and claims that are in my sections, and mark anything missing as [NEEDS DETAIL].”
- Ask for the quote with its working shown. Line items, the rate each one uses, the total. Then check the total yourself. Language models can slip on arithmetic, and a wrong number in a quote costs more than the time it saved.
- Edit, then tell it what you changed. Paste the version you sent and ask it to list what you changed. Add that list to the Project’s instructions, because a new chat won’t remember the lesson otherwise. The next draft should land closer.
Whatever you paste goes to Anthropic (or OpenAI) under your own account, so check your privacy settings before you paste client material in.
The bespoke part comes from their words
Templates feel generic when they ignore what the client said. A draft built from the call transcript starts with their problem, described the way they described it, which is the part you’d struggle to fake. Your job becomes checking the scope and the price, and making sure the opening shows you listened.
AI is weakest at the judging. It doesn’t know which past proposal won or why, unless you tell it. It can invent a case study if the instructions are loose, which is why step 3 says “only use what’s in my sections”.
A Project waits to be asked, so the follow-ups stay with you
The Project route has limits you’ll probably hit. You’re the one carrying the transcript in. It only works while you have a chat open, so nobody notices that a proposal went out a fortnight ago and nobody replied. And the quote still gets typed into your accounting software by hand.
An AI assistant that lives in your working life closes some of that gap. Crads-AI is one. It’s open source and built on Claude, and it runs on a server in your own hosting account or on your own computer. It keeps your past proposals, meeting notes and pricing as pages in its own memory, so a proposal can become a skill you run by name. Run on a server and connected to your email and calendar, it can flag a proposal that’s gone quiet and draft the follow-up. It drafts; sending is always yours. Xero only connects read-only through your Claude account, so it can draft the quote lines and you enter them.
One call can show where your proposals stall
Some people take the steps above and never need more. Others want help working out where the time goes between a good call and a signed proposal.
If that’s you, book an AI problem-solving roadmap: a 30-minute call about how proposals and quotes work in your business, a written report with ways to solve it (the do-it-yourself route included), and a second 30-minute call to go through it together.