Straight answer: AI-written proposals read beautifully and convert badly, because clients aren't buying grammar, they're buying proof you understood their specific problem. The fix isn't to stop using AI, it's to use it for structure and speed while you keep the specifics, the numbers, and the opinions human.
Worth reading next: Your AI-Written Proposals All Sound Exactly Like Everyone Else's (Here.
The proposal that lost me a five-figure client
Two years ago I put a proposal together for a manufacturing client in the Midlands using ChatGPT to help me tighten the wording. It looked lovely. Clean headings, confident language, a nice executive summary at the top. I sent it feeling rather pleased with myself.
They went with someone else. When I asked why, the answer stung a bit: "Yours felt like it could have gone to anyone." They'd had three other proposals that week that used almost identical phrases: "streamline your operations," "drive measurable growth," "tailored solutions for your unique needs." Mine sat in that pile.
The proposal that won was two pages shorter, had one glaring typo, and quoted the client's own managing director back to him from a call three weeks earlier. It wasn't polished. It was specific. That's the whole lesson, really, and most people writing proposals with AI right now are getting it backwards.
Why this is happening to almost everyone right now
I read close to forty proposals last year, some from people I mentor, some sent to me by clients asking which freelancer to hire. I could spot the AI-heavy ones within about six seconds, before I'd read a single sentence, just from the shape of them. Three bullet points under every heading. A summary paragraph that restates the brief back at the client. The word "use" showing up twice on page one.
Here's the uncomfortable bit nobody selling AI courses wants to say out loud: the tools are getting better at sounding confident faster than they're getting better at sounding true. A proposal generated in four minutes reads like it took four hours, which used to be a signal of effort. It isn't anymore. Clients who buy from consultants, freelancers, or agencies more than once a year have started to notice, and it's making them suspicious of everything, including proposals that were written with care.
That's the trap. AI hasn't just made bad proposals worse, it's made good proposals harder to trust on sight, because the surface no longer tells you anything about the substance underneath.
What clients are scanning for
When someone opens a proposal, they're not reading it the way you wrote it. They're scanning for three things in this order:
- Did they mention something I said, not just the brief I sent?
- Is there a number in here that's specific to my business, not a generic industry stat?
- Does this person sound like they'd argue with me if I was wrong about something?
None of those come from a prompt. They come from a conversation you had, notes you took, and an opinion you're willing to put your name to. Companies that win on brand trust understand this instinctively, it's the same instinct behind the Duolingo marketing strategy, where the whole appeal is a voice that clearly belongs to one specific personality rather than a committee. A proposal is a one-to-one version of the same thing. Generic voice, generic result.
The five-step way to use AI without sounding like everyone else
I still use AI on every proposal I write. I'm not precious about it. The difference is where it sits in the process.
1. Write your own opening paragraph first, no AI involved
Before you open any tool, write two or three sentences by hand about what the client said to you. Not what the brief says, what they said, in the meeting, in their words. This becomes your opening line and it's the one paragraph in the whole document that can't be faked.
2. Use AI to build the skeleton, not the content
Ask it for a logical structure for a proposal of this type: scope, timeline, investment, next steps. Fine. That's admin. What you don't do is ask it to fill those sections in about your client, because it will reach for the same forty phrases it reaches for with everyone else.
3. Put one real number in every section
Not "significant improvement in engagement," but "your bounce rate on the pricing page is 71%, and the three clients I've fixed this for saw it drop to under 40% within six weeks." Specific numbers do something generic language can't: they prove you looked. This is the same principle behind good Hotjar marketing strategy work, where the entire pitch rests on showing people exact behaviour, not talking in vague terms about "user experience."
4. Say something they might disagree with
Every proposal I've won in the past year has included at least one sentence where I pushed back gently on something the client wanted. "You asked for a full rebrand, and I don't think you need one, I think you need to fix your homepage headline first and see what happens." AI won't write that for you because it doesn't have skin in the game and it doesn't want to risk annoying anyone. You do have skin in the game. Use it. Dale Carnegie wrote almost a century ago that people are moved far more by being taken seriously than by being flattered, and that lesson from Dale Carnegie's business principles holds up better in a proposal than any prompt engineering trick.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
5. Read it out loud before you send it
If a sentence would sound strange coming out of your mouth on a call with this client, cut it. This single step catches most of the AI residue, the "in today's fast-paced digital landscape" type filler, faster than any editing checklist.
The pricing section is where most proposals fall apart
AI is useful here in a way people underuse. Instead of a flat number sitting on the page, build the value case around something the client can interact with rather than just read. I started adding a simple ROI breakdown to proposals over £3,000, showing what a 10% improvement in their conversion rate would be worth over twelve months using their own traffic numbers. Clients who saw actual maths tied to their own figures signed faster than clients who got a paragraph about "significant return on investment." If you want to take this further, an interactive calculator embedded on your site does the same job at scale, letting a prospect play with their own numbers before you've even had the call, which does more selling than any amount of proposal polish ever will.
What this costs you if you get it wrong
I want to put a real number on this because vague warnings don't change behaviour. Across the freelancers and small agencies I've worked with over the past eighteen months, proposal-to-client conversion rates for AI-heavy, generic proposals sit somewhere around 12 to 18%. The same businesses, once they switched to the five-step approach above, mostly landed between 30 and 45%. That's not a marginal gain. If you send ten proposals a month at an average project value of £2,500, the difference between those two ranges is somewhere close to £45,000 a year in work you're currently losing to a document that sounded fine but proved nothing.
That gap is entirely closable and it has almost nothing to do with your prices, your portfolio, or your experience. It's the document itself.
Brand voice matters more in a proposal than anywhere else
Companies that get remembered tend to have a very deliberate voice running through everything they write, including the boring commercial documents. Look at how bold and slightly blunt every piece of Revolut's marketing strategy is, or how calm and unhurried Calm's brand voice stays even in their app store copy. Neither of those voices came from a prompt. They came from someone deciding, early on, what the brand sounds like when it's annoyed, when it's excited, when it's explaining a price increase. Your proposal is one of the few pieces of writing where a prospective client is reading closely, on purpose, deciding whether to hand you money. It deserves the same deliberate voice, not the smoothest possible AI draft.
If you don't have time to fix this yourself
Most small business owners I talk to know their proposals have gone a bit flat, they just don't have the hours to rebuild the process while also running the business. That's a fair reason to bring someone in rather than a failure of discipline. If your proposals, your onboarding, and your client communication are all quietly leaning on AI in the same generic way, it's worth getting an outside eye on the whole system rather than patching one document at a time. That's precisely the kind of gap an AI consultant for small business is there to close, someone who looks at where the tools are helping and where they're flattening your voice, and fixes the second problem without losing the first.
Frequently asked questions
Is it bad to use AI for proposals at all?
No, it's bad to let AI write the parts that need to be specific to the client. Use it for structure, formatting, and tightening your grammar. Keep the opening paragraph, the numbers, and any opinion or pushback entirely your own, because those are the parts a client is reading closely.
How can I tell if my proposal sounds too generic?
Read it and ask whether you could send the exact same document to a different client in the same industry with only the company name changed. If the answer is yes, it's too generic. A strong proposal should be unusable for anyone else because it's built around something specific that client said or showed you.
Does adding numbers to a proposal improve conversion?
Yes, and specific numbers tied to the client's own business perform far better than industry averages. A line like "your current checkout has four steps and industry data shows three steps convert 22% better" is far more persuasive than "we can improve your conversion rate," because it proves you looked at their actual site rather than sending a template.
What's the single fastest fix if I only have ten minutes before sending a proposal?
Read it out loud and cut anything that sounds like it could have been said to a different client. Then add one sentence where you disagree with something they asked for. Those two changes alone remove most of the generic AI feel without touching the rest of the document.
Related reading: Why Your AI-Written Proposals Are Losing You the Job (Even Good Ones) and Why Your AI Still Writes Like Everyone Else's (And the 45-Minute Fix That Changes That).
I go much deeper on this in the AI marketing guide.
