Aisha is the sales manager at a facilities services company. Her team started using AI to speed up proposals, and it worked. Drafts that took an afternoon took an hour.
Then one draft promised delivery within five working days. The approved service document only supports ten. It nearly went to the customer.
A proposal can sound persuasive and still promise something you can't deliver. Here's an AI sales proposal workflow that keeps the speed, without the risky commitments.
I. Build an approved proposal pack
Give every part of the offer a clear source. A practical pack contains:
The customer's confirmed requirements, with open questions listed separately
Current product or service specifications, with version dates
The approved quotation, scope, exclusions and payment terms
Delivery assumptions confirmed by operations
Approved evidence, such as a published case study or a customer quote cleared for reuse
Old proposals can show the right tone, but they shouldn't be the source for today's prices, features or terms. Use only an AI tool your company has approved, and leave out unrelated customer records.
II. Map the buyer's needs before writing
Before drafting, ask AI for a short requirements map. Match each customer need with an approved response and its source. Any requirement without an answer stays visible.
That's what would have caught Aisha's problem. The customer asked for five working days, and the service document says ten. The draft should flag the mismatch for the salesperson, not quietly shorten the lead time.
III. Give AI clear boundaries
Try this prompt with your approved pack:
"Draft a proposal for the customer requirements provided. Use only the approved source pack. Keep quoted prices, scope and terms unchanged. In review notes, give the source for each factual claim. List missing or conflicting information separately. Don't invent delivery dates, guarantees, certifications or customer results. Mark the output as a draft requiring approval."
No prompt guarantees AI will follow every rule, so the review step stays.
IV. Check the commitments before the wording
Review in this order, so a polished introduction doesn't hide a costly mistake:
Offer: Does the product or service actually fit the confirmed requirements?
Commercials: Do prices, quantities, discounts and payment terms match the approved quotation?
Delivery: Has the right colleague confirmed dates, dependencies and capacity?
Evidence: Can every claim and testimonial be traced to an approved source?
Scope: Are exclusions and customer responsibilities clear?
Next step: Does it ask for the right action, without implying approval already exists?
Do calculations in your quotation system, not the AI draft. If the proposal and the quotation disagree, fix that before sending either.
V. Keep approval and sending separate
Name who approves the commercial and delivery details. Save the sources, the final proposal and the approval record together. If the price or scope changes, the proposal goes through the checks again.
For a first pilot, the salesperson sends the approved proposal the usual way. Connecting AI straight to email is a separate decision you don't need to make yet.
VI. Measure what actually matters
Track the time from a complete brief to an approved proposal, including review and rework. Note common corrections, such as missing exclusions, so the pack improves each time.
Keep turnaround time separate from win rate. Faster drafts free time for better customer conversations, but wins also depend on fit, pricing and competition.
VII. Common questions
Can AI write sales proposals? Yes, it can draft them quickly from approved information. A person should still check prices, scope and delivery promises.
What should AI never change? Prices, discounts, delivery dates and contract terms. These must come from your approved sources.
Still deciding which task to start with? See our guide to business processes to automate with AI. To build a proposal workflow around your own materials, explore our AI for Sales Teams programme or book a discovery call.
Aisha is an illustrative example based on a typical SME.