Proposal Generator

What this generates, and who it is for

You give it the client, the problem in their words, the work you propose to do and the price. It returns a structured draft: the problem restated, the outcome you are selling rather than the tasks you will perform, the deliverables, the timeline, the commercial terms and one clear next step.

It is for freelancers and small agencies who lose deals to response time rather than to price, and who are still writing every proposal from a blank page. If a decent proposal takes you two hours, you will send fewer of them, and you will send them late.

What it is not is a document you send unread. A generated proposal is a first draft with your structure and the client's facts in the right order. The numbers in it are the numbers you typed. The numbers you did not type are the ones the model will quietly fill in for you, and that is where proposals go wrong.

If what you want is the reusable skeleton rather than a drafted document, use the Proposal Template Generator instead. That page covers the clauses, the statutory payment position and who owns the work when nobody wrote it down. This page is about generating a draft and then reviewing it hard enough to send.

How to read the output: a worked example

Say you are proposing an automation build. You typed one number into the tool: £5,000. The draft comes back reading well, with a section headed "Investment" that says £5,000, and a section headed "The cost of doing nothing" that says the client is currently losing around twenty hours a week to manual work.

Both of those sentences are defects. Here is the arithmetic that shows why.

Defect one: the price with no VAT position

The UK standard rate of VAT is 20%. It has been 20% since 4 January 2011, when it rose from 17.5%, and it applies to most goods and services. A reduced rate of 5% and a zero rate of 0% apply to specific categories, and some things are exempt altogether, including postage stamps and financial and property transactions. Consultancy and software work sits at the standard rate. All of that is set out on the GOV.UK VAT rates page.

Your proposal says £5,000. It does not say whether that is before or after VAT. If you are VAT-registered, you invoice £5,000 plus £1,000, and the client, who approved a £5,000 budget, receives a bill for £6,000. One of two things then happens.

Either you have an awkward conversation in week one of a project you have just won, or you absorb it. Absorbing it means treating the £5,000 as the gross figure, so the net is £5,000 ÷ 1.2 = £4,166.67, and the VAT you hand over is £833.33. You have given away £833.33 on a £5,000 job. That is 16.7% of the fee, and it happened because a single line was missing from a document that was otherwise fine.

The fix is four words in the price section: all fees exclude VAT. Or, if you are not VAT-registered, say so and say why the figure will not move: this figure is final and no VAT is chargeable. A generator will not add either sentence unless the field exists in your input, because it has no way of knowing your registration status.

Defect two: the cost of inaction with no source

"Around twenty hours a week" is the kind of number that reads persuasively and collapses the first time the client tests it against their own experience. Worse, if it is wrong in your favour, you have opened your proposal by overstating the problem, which is exactly the impression you were trying to avoid.

Replace it with a figure you can defend. Ask the client one question on the call: how many hours a week does this take? Then price those hours with an accredited official statistic rather than an invented rate. The Office for National Statistics Annual Survey of Hours and Earnings, released on 23 October 2025 and drawn from a sample of around 174,000 employee records, puts median hourly earnings excluding overtime for full-time employees at £19.67 in April 2025, up 5.4% from £18.66 the year before. The same release puts median gross annual pay for full-time employees at £39,039 and median weekly earnings at £766.60.

So if the client says six hours a week:

• 6 × £19.67 = £118.02 a week
• £118.02 × 52 = £6,137.04 a year

That number has a source, a date and a method behind it. It is also honest about what it is not: the ONS figure is the all-occupations median, so it is a proxy rather than a measurement of the specific person doing this specific task, and it is gross employee pay, which excludes employer National Insurance and pension contributions. The true cost to the business is higher. Say that in the proposal. A floor that you can defend beats an estimate that you cannot.

What the two corrected numbers do to the deal

Now the proposal has £5,000 plus VAT against £6,137.04 a year, and payback becomes a division rather than a claim:

• If the client is VAT-registered and reclaims the VAT, the real cost is £5,000. £5,000 ÷ £118.02 = 42.4 weeks.
• If the client cannot reclaim VAT, the real cost is £6,000. £6,000 ÷ £118.02 = 50.8 weeks.

That is a difference of eight and a half weeks, roughly two months, created entirely by the client's VAT status. It is the single most decision-relevant fact in the document, and no proposal generator will ever surface it, because it depends on something about the client that lives in your notes and not in your prompt.

The method: five numbers, five sources

The review pass is not a proofread. It is a check that every number in the document came from somewhere real. There are only five of them, which makes this quick once you know the list.

Notice the pattern. The generator is good at the parts that are structure and prose, which is most of the document. It is unreliable on precisely the parts that are quantities, because a language model producing a number in an empty field is producing something that looks like a number rather than a measurement.

Benchmark: when the arithmetic actually supports the sale

This is the table worth having open during the call, because it tells you whether there is a deal here before you spend an hour drafting. All of it is the same division applied to the ONS median of £19.67 an hour.

The line to notice sits between two and four hours a week. Below roughly four hours, a £5,000 build does not pay for itself inside a year on labour savings alone, and most buyers will not sign a business case with a payback longer than their own planning horizon.

That does not mean the project is worthless. It means labour saving is the wrong argument for it, and you should be selling something else: error rate, response time, a compliance requirement, capacity the business cannot hire for. What it definitely means is that pasting the client's two hours a week into a generator and letting it write a savings narrative will produce a document that argues against you the moment the client does the sum.

Common mistakes

Sending the first draft. The draft is 80% of the document and none of the risk. Every defect that costs you money is in the remaining 20%, and all of it is numbers.

Letting the tool set the price. Price is a commercial decision that depends on your capacity, your pipeline and how much you want this particular client. It is not an output.

Listing tasks instead of outcomes, then over-correcting. Generated proposals lean heavily on outcome language, which is right, but a proposal with no concrete deliverables gives the client nothing to approve and gives you nothing to point at when scope moves. You need both: the outcome in the summary, the deliverables in a list underneath it.

Leaving the timeline vague to stay safe. "Approximately four to six weeks" reads as caution to you and as six weeks to the client. Give dates, tie them to a start condition such as the deposit clearing or access being granted, and you have moved the risk to the thing that actually causes delay.

Attaching terms that contradict the proposal body. If the body says 50% up front and the terms say net 30 on completion, you have created an argument for later. Generate the proposal and the terms from the same set of facts, and read them side by side once before sending. The Proposal Template Generator exists so the terms half stays consistent between deals.

No follow-up plan. A proposal that goes out with no scheduled follow-up depends on the client being organised. Most of the deals lost after a good proposal are lost to silence, not to a no. There is a walkthrough of building that follow-up as a system rather than a habit in AI follow-up systems for sales.

Treating a signature as the end. It is the start of the part the client judges you on. Automating client onboarding covers what should fire the moment the document comes back signed.

Frequently Asked Questions

Can I send a generated proposal without editing it?

No. The prose will usually be fine, but every number needs checking against the list above, and the VAT position almost certainly needs adding because the tool has no way to know your registration status. Budget five minutes for the review. It is the cheapest five minutes in the whole sales process.

Should the price include VAT or exclude it?

State which, explicitly, either way. Business-to-business proposals conventionally quote excluding VAT, because a VAT-registered client reclaims it and cares about the net figure. What matters is not which convention you pick but that the document says so, so the invoice cannot surprise anyone. See GOV.UK on VAT rates for which rate applies to what.

Where should the cost-of-inaction figure come from?

Hours from the client, rate from something citable. The ONS median hourly earnings excluding overtime for full-time employees was £19.67 in April 2025, which is a defensible proxy and a conservative one, since it excludes employer National Insurance and pension costs. If the client will tell you what the person actually earns, use that instead and say where it came from.

Is £19.67 an hour the right rate for any task?

It is the all-occupations median for full-time employees, so it is a starting point rather than an answer. A task done by a senior specialist is worth considerably more per hour than that; a task done by an apprentice is worth less. The reason to use the median is that it is published, dated and independent, which makes it much harder to argue with than a number you chose. Label it as the median and the objection disappears.

How long should a proposal be?

Long enough to answer the questions the client will otherwise have to ask, and no longer. For most freelance and small-agency work that is two to four pages: problem, outcome, deliverables, timeline, price, next step. If the generated draft runs to eight pages, most of the excess will be restated context the client already knows.

Does a longer proposal win more often?

There is no reliable public figure on this that is worth quoting, and the vendor-published win-rate statistics that circulate on this topic come from the companies selling proposal software, on samples they define themselves. Treat all of them as marketing. What is safe to say is mechanical rather than statistical: a proposal cannot be approved if it does not contain a price, a scope and a date, and length has nothing to do with whether those are present.

Can it write the contract too?

It can produce something contract-shaped, which is not the same thing. Payment terms, interest on late payment and intellectual property ownership all have statutory defaults in the UK that are worth understanding before you overwrite them, and on payment timing the default is usually better than what a template will give you. That is covered on the Proposal Template Generator page. For anything with real money or real liability attached, have a solicitor read the terms once. You reuse them for years, so it is a single cost.

What if I do not know the client's budget?

Then the proposal is doing a job it cannot do. Ask before you draft. If you genuinely cannot find out, present two scoped options at different prices rather than one figure, and let the response tell you where the budget sits. Do not let a generator invent a mid-range number for you.

Can I use this for scoping the work as well as pricing it?

For a rough shape, yes, but sizing a build is a different problem from writing about it. The Project Cost Calculator handles the estimate itself, including what to do about the fact that estimates are systematically optimistic.

Where this tool stops

A proposal generator can write the document. It cannot know what was said on the call.

That is not a limitation of this particular tool, it is the shape of the problem. The facts that decide the deal, whether the client is VAT-registered, how many hours the task really takes, who has to sign it off, what the last agency did that annoyed them, all arrive in conversation and then live in your head for about a day. By the time you sit down to draft, half of it is gone, and the blank fields in the form get filled with plausible substitutes.

Fixing that is a CRM problem rather than a writing problem. Call notes and transcripts land in fields on the deal record; those fields populate the proposal draft; the draft goes out with the client's own numbers in it because nobody had to remember them. The signature event then moves the pipeline stage and triggers onboarding, so the document is not something you chase separately from the deal.

Building that is most of what I do: GoHighLevel and CRM implementations where the pipeline, the proposal and the follow-up are one system rather than three tools and a habit. If you are writing proposals from memory and losing the details between the call and the draft, that is the fixable part.

You can see how I work on Upwork or Fiverr, or book a consultation and bring one recent proposal with you. Fifteen minutes on where the numbers came from usually tells you what to automate first.