Sales proposals with AI: fast draft, human review
AI writes the first draft of a proposal in minutes when it receives the discovery summary, the company template and the pricing rules. The salesperson and the manager review scope, prices, timelines and promises before anything goes out. The workflow, the prompt and the review checklist.
Generative AI writes the first draft of a B2B sales proposal in minutes when it receives four inputs: the validated summary of the discovery call, the company’s proposal template, the price list with discount rules and two or three similar deals already sold. The salesperson and, above a certain amount, the manager review what AI cannot decide on its own: scope, prices, timelines, assumptions and any sentence that becomes a commitment. If you run sales, your decision comes down to three points: which template the AI follows, where it can take information from and who reviews each part before it goes out. Leaders who settle this before rolling out the tool gain speed without losing control over what the company promises.
The picture I find in sales teams comes in two versions. In the first, the salesperson spends an afternoon opening another customer’s proposal, swapping names and numbers, and sends a document with the wrong industry in one paragraph and six pages of company background before the customer’s problem. In the second, more recent version, the salesperson asks ChatGPT for “a proposal for a logistics company” and gets a well-written, generic text with benefits the company does not deliver and a timeline the delivery team cannot meet. Both versions share a cause: the proposal is born without the customer’s diagnosis and without rules about what can be promised.
How do you use AI to write a sales proposal?
The workflow I recommend has six steps, and AI takes part in three of them.
- Validate the diagnosis with the customer. The written summary of the discovery call, with problem, example, estimated cost and people involved, comes back corrected by the customer before any proposal.
- Gather the inputs. The salesperson collects the validated summary, the meeting transcript, the proposal template, the price list and two similar cases.
- Generate the draft with AI. The model writes the text following the company structure, with the problem in the customer’s words and the scope built from approved modules.
- Review with the checklist. The salesperson checks numbers, names, scope, timelines and promises. Above the agreed limit, the manager reviews price and terms.
- Ask the AI for a critical read. With the reviewed text, the AI plays the customer’s CFO and points out likely questions and objections.
- Present it live. The proposal is presented in a meeting with the decision-makers, and the file follows afterwards.
The time savings sit in step 3, and the quality gains sit in steps 1, 4 and 5. The structure of a proposal a buying committee approves is in B2B sales proposal.
What does AI do well and what stays with the salesperson?
AI writes fluently, organizes scattered information and adapts language to the reader. It fails when it has to decide something that depends on information it did not receive, such as the team’s capacity or the minimum margin on the contract. The split below is the one I use to train teams.
| Proposal section | Does AI draft it? | Who reviews | Risk if nobody reviews |
|---|---|---|---|
| Context and customer problem | Yes, from the validated summary | Salesperson | Problem described in the seller’s words, which the committee does not recognize |
| Goals and success metrics | Yes, from the diagnosis | Salesperson | A metric the customer never mentioned |
| Scope and deliverables | Yes, only with approved modules | Salesperson and delivery team | A deliverable the team cannot do |
| Timeline | Draft based on similar cases | Delivery team | A deadline that clashes with the team’s schedule |
| Investment and terms | No; the salesperson fills it from the price list | Salesperson, and manager above the limit | Discount outside policy, unapproved payment terms |
| Assumptions and exclusions | Yes, from the standard list | Salesperson, and legal when applicable | A missing exclusion that turns into an extra request during the project |
| Cases and proof | Yes, only with authorized cases | Marketing | A customer named without permission |
| Executive summary | Yes, last | Salesperson | A promised result the company does not control |
Price and terms stay out of generation for a practical reason: the model can invent a plausible discount or round a number. The salesperson fills in these fields from the price list and the discount policy, and the AI works on the text around them. Brazil’s tax reform and its effect on sales also changes how companies build prices and present proposals, one more reason to keep these fields with the salesperson.
What information does AI need to write a good proposal?
AI writes with what you give it. A request for “a proposal for a food manufacturer” produces generic text, and buyers spot generic text in two lines. The inputs that make the difference:
- Validated discovery summary, with the problem in the customer’s words and the cost they estimated themselves.
- Meeting transcript or notes, so the AI captures expressions, priorities and objections. Recording and automatic summaries are covered in meeting summaries with AI.
- The company’s proposal template, with sections in the right order, the expected tone and length.
- Catalog of scope modules, with description, deliverables and effort for each, approved by the delivery team.
- Standard list of assumptions and exclusions, reviewed by legal.
- Two or three similar cases, with permission to use them and results the customer confirmed.
Items 3 to 6 are prepared once and maintained by leadership. They form the base that keeps the AI from improvising, and building them often reveals that the company was selling the same service with four different descriptions.
Prompt to generate the proposal draft
Use a company-approved AI account, such as the business versions of Claude, ChatGPT or Gemini, or the assistant built into your CRM. Attach the inputs as files and keep the prompt fixed for the whole team, so proposals come out with the same structure.
The instruction to flag what is missing changes the result. Without it, the model fills gaps with plausible text, the kind of error described in AI hallucination. With it, the salesperson gets a list of what still needs to be asked.
How do you review a proposal written by AI?
The salesperson reviews with a fixed checklist, and the manager reviews proposals above the agreed amount. Review takes 20 to 40 minutes for a mid-sized proposal, and it is the moment when the company decides what it signs.
- Names and numbers. Customer name, people, industry, amounts, quantities and dates checked against the CRM and the price list.
- A problem they recognize. The first paragraph uses the words the customer used. If the contact reads it and does not recognize themselves, the proposal goes back.
- Deliverable scope. Every scope item exists in the catalog and has effort estimated by the delivery team.
- Promises. No sentence guarantees a result the company does not control, such as a sales increase or cost reduction with a percentage.
- Assumptions and exclusions. Exclusions cover what the customer might assume is included.
- Price and terms. Amount and discount within policy, with recorded approval when the limit is exceeded.
- Missing markers. No [MISSING] left in the text that goes out.
- Tone. The text sounds like the company, without advertising adjectives and without paragraphs that could sit in any competitor’s proposal.
How much time does AI save on proposals?
The gain depends on the starting point. In a team where salespeople write each proposal from scratch, a complete draft in minutes frees up hours per week. Salesforce, in its 2024 State of Sales report, with 5,500 sales professionals in 27 countries, found that salespeople say they spend 70% of their time on tasks that are not selling, and writing documents is part of that group. A Gartner survey from 2023, with 501 sellers, found that 77% struggle to complete their assigned tasks efficiently.
Time saved only turns into revenue if the salesperson spends it with customers. I recommend measuring two things before and after adoption: the time between the discovery call and the proposal presentation, and the share of proposals presented live. An illustrative example: if the team takes eight business days to present today and that drops to three, the customer receives the proposal while the problem is still fresh, and the salesperson has time to book the presentation with the decision-maker.
How should you handle customer data?
A proposal gathers sensitive information: the customer’s internal problems, numbers shared in confidence, names and titles. Three rules cover most of the risk.
- A business account with a contract. Use enterprise versions of the tools, where the vendor commits not to train models on your data, and forbid personal accounts for this use. The rule goes in writing in the AI usage policy.
- The minimum necessary. Send the AI what the proposal needs. Personal data that does not appear in the final document stays out.
- Legal basis and transparency. Processing personal data follows Brazil’s data protection law (LGPD) and its equivalents, as I explain in AI and LGPD, and customers may ask how the company uses AI. The salesperson needs to know how to answer.
How do you present the proposal once it is ready?
AI speeds up writing, and the presentation is still the moment the deal moves. A proposal sent by email without a conversation leaves the contact alone to explain it to the committee, and the salesperson never learns which part raised doubts. I recommend booking the presentation before the proposal is finished, with the decision-maker present, and showing the document on screen. The reasons a proposal sent by phone tends to disappear are in proposals over WhatsApp. After the presentation, following up on the proposal tracks the decision date agreed in the meeting.
For deals that need financial approval, AI also helps prepare the page the CFO reads: the cost of the problem, the investment and the expected return with explicit assumptions. How to build that page is in a business case for finance, and the conversation about value and return at the negotiating table is in B2B negotiation with ROI.
A 30-day plan for leadership
- Week 1, foundation. Review the proposal template, build the scope module catalog with the delivery team and the list of assumptions and exclusions with legal.
- Week 2, pilot. Two salespeople generate the week’s proposals with the prompt above, and the manager reviews all of them with the checklist.
- Week 3, adjustment. Adjust the prompt and the template based on the errors found in review, and record the cases authorized for use.
- Week 4, whole team. Open the workflow to everyone, with the mandatory checklist and manager review above the agreed amount.
- Day 30, review. Compare the time between discovery and presentation and the share of proposals presented live with the previous month.
If your leadership wants the team to use AI to write better, faster proposals without losing control over price and scope, that is the work of the in-company negotiation training. I run the program on your company’s real proposals, with the prompt and checklist adapted to your template, and prepare managers to review and coach the team on presenting. The company leaves with a revised proposal template, the scope catalog and the review workflow in place.
Frequently asked questions
What is the best AI for writing sales proposals?
The business versions of Claude, ChatGPT and Gemini all write proposals well. The difference in results comes from the inputs and the prompt, and the choice should follow the tool the company has already approved in its AI usage policy.
Can AI set the proposal price?
I do not recommend it. The salesperson fills in amount and terms from the price list and discount policy, and the AI works on the text. Language models can invent plausible numbers.
Can I put customer data into ChatGPT?
Only in a business account with a contract that prevents the data from being used to train models, and only what the proposal needs. Personal accounts should be off-limits for this use.
How long does it take to review an AI-written proposal?
Twenty to 40 minutes for a mid-sized proposal, with a fixed checklist. Proposals above the agreed amount also go through the manager.
Do AI-written proposals sound generic?
They do when the AI receives only the name of the industry. With the validated discovery summary, the transcript and the scope catalog, the text uses the customer’s words and answers the problem they described.
Want to bring this to your team?
The topics in this article also become in-company training: designed from a diagnosis of your operation, with real cases from your team and a deliverable that stays with the company.
More writing on the same subject.

Meeting summaries with AI: from conversation to CRM
Logging the meeting is the task salespeople put off the most and the one managers need the most. How to turn the transcript into CRM fields with AI, with the prompt, the two-minute review and the care the recording requires.
12 min readRead more
Weekly sales report with AI: the Monday summary
A manager who builds the weekly report on Sunday night arrives at the meeting tired and with numbers that do not match other teams'. How to set up a weekly summary that the CRM calculates, AI writes and the manager reviews in minutes.
11 min readRead more
B2B price increases: how to announce one and keep the client
An increase that arrives as a generic email from finance pushes the client toward a competitor or toward a request for an exception. How to segment the customer base before announcing, when to give notice, what to say in the conversation, which trade-offs to offer and what to measure in the following 90 days.
14 min readRead more