How to Write an AI Automation Proposal That Closes (2026 Template)
Last updated: September 2026.
Most AI agencies lose deals in the AI automation proposal, not the discovery call. The call goes well, the prospect is enthusiastic, and then a twelve-page document arrives full of platform names and hourly rates and the deal quietly dies in a group chat you will never see.
Some links below are affiliate links — if you sign up through them, Pivot 2 Thrive may earn a commission at no extra cost to you. It never changes what we recommend.
On This Page
This playbook comes from Dr Priya Jaganathan, a Go High Level Certified Admin, Certified AI Tech Stack Consultant and keynote speaker who has written, lost and won a great many of these documents for Australian automation and AI agency work.
What an AI automation proposal is actually for
An AI automation proposal is a decision document for people who were not on your call. That is the whole job. The person you spoke to already believes you; they now have to defend the spend to an owner, a finance person or a partner who has heard nothing.
Every design choice follows from that. Short, because nobody forwards twelve pages. Numbers from the client's own business, because your case studies do not move their finance person. Explicit exclusions, because the fastest way to lose a deal internally is for one stakeholder to imagine something you never offered.
Why most proposals die in the buying group
Gartner surveyed 632 B2B buyers in August and September 2024 and found 74 per cent of B2B buyer teams demonstrate "unhealthy conflict" during the buying decision process. The same research found buying groups that reach consensus are 2.5 times more likely to report a high-quality deal.
Read that as an operating instruction. Your proposal is not competing with another agency nearly as often as it is competing with disagreement inside the client's own business.
Your proposal is also the first draft of your delivery scope. Focus Digital's 2026 agency churn analysis put monthly churn at 1.6 per cent for retainer-based agencies against 4.2 per cent for project-based — 18 per cent versus 42 per cent annually.
A proposal that sells an ambiguous project buys you a client who leaves. One that sells a defined outcome with a clear ongoing role buys you a retainer.
| Proposal element | What most agencies write | What closes instead |
|---|---|---|
| Opening | About us, our team, our approach | The client's problem in the client's own numbers |
| Pricing | Hourly rate and estimated hours | Three fixed-scope options, middle one recommended |
| Scope | A feature list | In-scope, out-of-scope, and what the client must supply |
| Timeline | "4–6 weeks" | Dated milestones with client dependencies named |
| Length | 10–20 pages | 2–4 pages plus an appendix nobody has to read |
How to structure an AI automation proposal that closes
Seven sections, in this order. Anything that does not fit one of them belongs in the appendix or in the bin.
1. The problem, in their numbers. Open with what discovery told you, quantified: "You receive around 120 enquiries a month, roughly 30 arrive outside staffed hours, and your current average first response is 4 hours." No adjectives. If you cannot fill this section, your discovery was too shallow — our list of discovery call questions for Australian AI agencies exists for exactly this reason.
2. What changes, in one paragraph. The outcome in plain language: every enquiry answered within 60 seconds, seven days a week, with bookings landing in the existing calendar. Not the architecture. The state of the world afterwards.
3. Three options, not one. A single price is a yes-or-no question and buying groups are bad at those. Give a contained starting scope, a recommended scope, and a broader scope. Label the middle one "recommended". Most clients take it, and the ones who do not now have a way to say a smaller yes instead of a no.
4. What is out of scope, stated plainly. This section feels risky and does the most work. Name what you are not doing: website redesign, content writing, CRM data cleansing, third-party licence fees. Every item here is a scope-creep argument you have already won — see our guide to handling scope creep for how these clauses hold up in practice.
5. What the client must supply, with dates. Logins, brand assets, a sign-off decision-maker, and someone to answer questions within two business days. Naming these turns a future delay into a shared, pre-agreed risk.
6. Pricing, with the ongoing relationship visible. Fixed build fee plus a monthly figure for management, optimisation and reporting. Show both from the start. An ongoing fee introduced after the build reads as an upsell; one included in the original three options reads as the deal — our breakdown of how to price AI automation retainers in Australia covers the bands.
7. One clear next step. Not "let us know your thoughts". A specific action with a date: "Reply to approve Option B and we will start Monday 6 October." One action, one date, one link.
Then send it with a two-line email and a short recorded walkthrough. The video is for the people who were not on the call, and it is the single cheapest thing you can add.
Want a second set of eyes on a proposal before it goes out? Book a strategy call and we will pull apart your current template against the structure above.
Not on HighLevel yet? Start with a free 30-day trial — enough time to build everything you are proposing before you pay a cent.
What this looked like on a real Australian deal
An Australian automation consultant we work with was sending 14-page proposals with an hourly rate and a range of estimated hours. The losses were almost never to a competitor — they were to "we'll revisit next quarter".
We cut the document to three pages. Page one was the client's own enquiry and response-time numbers from discovery. Page two was three fixed-scope options with the middle one recommended, each showing a build fee and a monthly management fee. Page three was in-scope, out-of-scope, client dependencies and a dated next step.
The change that mattered most was removing the hourly rate. Hours invite negotiation about hours. Fixed scopes move the conversation to which outcome the client wants.
Two other things shifted. Roughly a third of clients took the smaller option first and upgraded within a quarter — previously those were lost deals. And because the monthly fee sat in the original options, almost every won project came with a retainer attached.
The delivery side got easier too — the out-of-scope section became the first draft of the statement of work, which ties into contracts and SLAs for Australian AI agencies.
Proposal mistakes that cost AI agencies deals
Selling the platform instead of the outcome. No buyer outside your industry cares which tools you are using. Name the stack once, in the appendix, and spend the document on what changes for them.
Pricing by the hour. Hourly pricing caps your margin at your speed and punishes you for getting better. It also invites the one question you cannot win: "why will it take that long?"
Writing only for your champion. A document that assumes the context of your discovery call is unreadable to the finance person who gets it third-hand.
Leaving out-of-scope blank to seem accommodating. Ambiguity does not make you easier to buy from. It makes you riskier, and it guarantees a delivery argument in week three.
Following up with "just checking in". Follow up with something useful: a short video answering the objection you expect, a reference from a comparable business, or a revised option. Every follow-up should give the champion new ammunition for their internal argument — the same logic that governs how to report AI automation ROI to clients once you have won the work.
Frequently Asked Questions
How long should an AI automation proposal be?
Two to four pages for the decision document, with anything technical in an appendix. The proposal has to survive being forwarded to someone who was not on your call, and long documents do not get forwarded or read in full.
Should I include pricing in the proposal?
Yes, always, and give three fixed-scope options rather than one number. Withholding price to force another call adds a week and signals you are not sure what the work is worth. Name the middle option as recommended.
Should I price AI automation projects hourly or fixed?
Fixed scope, nearly always. Hourly pricing caps your margin at your delivery speed, invites negotiation about estimates rather than outcomes, and penalises you for building reusable systems. Quote a fixed build fee plus a monthly management fee.
How do I handle a client who wants to cut the monthly fee?
Reduce the scope rather than the rate. Offer a smaller build with a smaller monthly fee, or a lighter service level with slower response times. Discounting the same scope teaches the client the first number was not real.
How many follow-ups should I send after a proposal?
Three or four over about three weeks, each carrying something new — a walkthrough video, a comparable reference, or a revised option. Then ask directly whether it is a no for now, and put it in a dated nurture sequence rather than chasing it.
What should always be in the out-of-scope section?
Anything adjacent that a reasonable person might assume is included: website changes, content writing, data cleansing, third-party licence and usage fees, and support for systems you did not build. Being explicit here prevents most delivery disputes.
If your close rate is the constraint rather than your lead flow, the document is usually the problem. Book a strategy call to review your proposal template, or start at Pivot 2 Thrive to see how we build and hand over these systems. If you are building the delivery side on HighLevel, start your free 30-day trial first.
Related Articles
Services
More on running an AI agency
- AI Agency Discovery Call Questions
- How to Price AI Automation Retainers in Australia
- How to Handle Scope Creep in an AI Agency
- AI Agency Contracts and SLAs
- GoHighLevel vs HubSpot for Australian Agencies
Everything else
