What actually gets automated
Writing a proposal is mostly assembly: take the client data — the enquiry, industry, volume, previous correspondence, stage of the deal — work it into the structure of the proposal, choose wording that fits the situation and lay it out in the company style. Routine, but not quick: the manager spends time on it that could have gone into the next conversation or into closing the current deal. The model takes over that assembly, working from the real data of the specific client — but not the decision about what to offer and on what terms, which stays with the manager.
The key difference from typing “write me a proposal” into a chatbot is that the model works from your approved templates and the actual deal data rather than inventing a structure each time. That is what keeps quality stable and removes the risk of drifting away from the format the company has agreed on.
What the AI does and what stays with a person
| Step | Who does it |
|---|---|
| Pulling client data from the enquiry and correspondence | the model, from the CRM |
| Drafting the proposal text on the approved template | the model |
| Choosing wording for the specific situation | the model, as a first draft |
| Checking figures, terms and deadlines | the manager |
| Final wording of complex terms (discounts, special conditions) | the manager |
| Deciding what offer suits this client | the manager |
| Sending it | the manager |
How to connect it
- •Load the approved proposal templates and the standard terms,
- •Connect the CRM so the model sees the enquiry and history,
- •Run it on past deals first and compare with what managers actually sent,
- •Keep sending manual until the drafts consistently need only minor edits.
Wiring the model into the CRM is covered in integrating AI with your CRM, and keeping the writing in your own voice in AI content generation that actually works.
Frequently asked questions
Is an AI-written proposal genuinely personal or just a template?
It depends on the setup. If the model receives the real client data — the enquiry, the industry, the volume, the previous correspondence — and works it into the proposal structure, the text is personal in substance. Swapping a name inside one universal template is not personalisation, and clients can tell.
Does a manager have to check every proposal?
Yes, without exception. The model prepares a draft from your templates and data, but a person sends it — after checking the figures, terms, deadlines and whether the tone fits. That is what prevents an error in price or terms that would later have to be explained to the client.
Can the model see client history from the CRM?
Yes, and it is the standard working setup — the model takes the enquiry data and conversation history from the CRM and builds the draft on that basis rather than working detached from the deal context, which noticeably improves relevance.