The short version: AI hasn’t made proposals worse, it’s made them identical, and prospects are noticing because they’re running your quote through the same tools you used to write it. The fix isn’t a cleverer prompt, it’s feeding the tool material a competitor couldn’t possibly have, then cutting every sentence that could belong to anyone.
I read about 150 proposals a year, and I can spot the AI ones in four seconds
Part of my consulting work involves sitting with small business owners while they review quotes from suppliers, freelancers, and agencies. Web designers, copywriters, PR people, marketing contractors. Over the last eighteen months something changed. Proposals used to have personality, even bad personality. Now most of them read like they came out of the same factory.
Last spring a client of mine, a Manchester-based bakery supply company, asked me to help her compare six quotes from web design freelancers. I opened all six PDFs side by side. Four of them opened with a version of “In today’s competitive digital landscape, a strong online presence is essential.” Three used the phrase “tailored solution.” Two had identical three-bullet lists of “why choose us” that started with “years of experience” and ended with “customer satisfaction.” None of it was wrong exactly. It was just interchangeable. You could have swapped the company names on any two of those documents and nobody would have noticed for a week.
That’s the real problem with AI and proposals right now. It’s not that people are using it, it’s that they’re all typing the same three-line prompt into the same chat window and shipping the first draft. “Write a proposal for a web design project for a small bakery supply business” produces roughly the same document whether it’s Sarah in Leeds or Marcus in Bristol typing it, because the tool has nothing to work with except the category of the job.
The uncomfortable bit nobody selling AI courses wants to say
Here’s what most advice on this topic won’t tell you. The problem isn’t that AI writing is detectable, though it often is. The problem is that your prospects are now doing to your proposal exactly what you did to write it. They’re pasting it into ChatGPT and asking “does this look generic compared to the other quotes I’ve had” or “what’s specific to my business in here.” I’ve watched clients do this in front of me, almost as a reflex, the same way people used to Google a company name before signing a contract. If the answer comes back “not much,” you’ve lost the deal before you’ve even had the call, and you’ll probably never know why, because they won’t tell you. They’ll just go quiet.
So the fix has nothing to do with hiding that you used AI. Everyone knows you did. The fix is making sure the output contains things a generic prompt physically could not produce.
What to feed it instead of a one-line prompt
The pattern I now use with clients rewriting their proposal process looks like this, and it takes maybe twenty extra minutes per document, not hours:
- Paste in the actual call notes or email thread. Not a summary, the raw text, including the client’s exact phrases. If they said “we lose about four hours a week chasing invoices,” that number and that phrase need to end up in the proposal, verbatim.
- Include one number from their business that’s theirs alone. Their website’s current bounce rate, their last quarter’s return customer percentage, the number of enquiries they get a month. A generic prompt cannot invent this. If it’s in the proposal, it proves someone looked.
- Name a competitor of theirs by name, accurately. Not “your competitors” but “compared to how [Rival Bakery Supplies Ltd] structures their product pages.” This single move does more to signal you’ve done research than any amount of polished prose.
- Give the AI your own past project as a reference, not a template. Feed it a short paragraph on a real job you did, with a real result, and ask it to match that tone and structure rather than write from a blank prompt. This is how you keep your voice instead of the model’s default voice, which is politely bland by design.
- Ask it to list every sentence that could apply to any business in this industry, then delete them. This is the single most useful prompt I’ve found for this problem. Paste your draft back in and ask exactly that question. It’s remarkably good at flagging its own filler.
- Add one interactive or visual element that a templated quote never has. A simple cost comparison, a rough ROI projection, even a one-screen calculator showing what the current problem is costing them per month. If you’ve never built one, the piece on using interactive calculators to convert more visitors covers the mechanics, and the same logic works inside a proposal document as well as it works on a landing page.
None of that is complicated. It’s also almost never done, because step one, pasting in the raw notes instead of a two-line summary, is the bit people skip when they’re rushing to get a quote out before lunch.
The brands that never sound like anyone else, and why that’s the actual model here
When I’m working with a business owner on tone, I often point them at how established brands refuse to be interchangeable even in categories full of competitors. Look at how Red Bull built a marketing strategy around a voice so specific it survived being copied by every energy drink that followed it. Or how Miro built its brand around one very specific promise rather than a list of generic features. That’s the standard a proposal needs to hit too. Not “we offer tailored solutions with years of experience,” but a specific claim only you could make, backed by a specific number only your business has.
A small business owner doesn’t need a brand agency to do this. You need to stop letting AI write from nothing and start making it write from your actual material.
A quick before and after, because this is easier to show than explain
Before, generic prompt output: “We understand the challenges facing small businesses in the retail sector today and have designed a tailored proposal to help you achieve your goals through our years of expertise.”
After, same job, fed with real notes: “You told me on the call that Saturday footfall drops off after 2pm and your current Instagram posts get most engagement before 9am, which suggests your posting schedule is fighting your actual customer pattern. Here’s a four-week test moving your main post to 7am with a cost breakdown below.”
Same tool, same underlying model, wildly different result, because the second one contains a fact the first one couldn’t possibly have.
Where this connects to a bigger problem: AI reads your whole site, not just your proposal
There’s a wider version of this issue that a lot of small business owners haven’t clocked yet. The same flattening effect happening in proposals is happening on websites, because AI search tools like ChatGPT and Perplexity now summarise your business for prospects before they ever land on your page. If your About page, your case studies, and your service pages read as generically as those bakery web design quotes did, the AI summarising you to a potential client will produce the same bland version of your business that a lazy proposal would. It’s worth running the 15-minute AEO audit on your own site to see whether the specific proof points that would make a prospect choose you are even visible to the tools doing the summarising, because if they’re not, you’re losing the same battle before the proposal stage even starts.
When it’s worth bringing someone else in
If you’re a one or two person business, you can fix your own proposal process with the six steps above in an afternoon. Where I usually get called in is when a business has five or more people writing quotes, all using AI slightly differently, all producing wildly inconsistent results, some brilliant and specific, some the same generic template three of your competitors are also sending out. That’s a process problem, not a writing problem, and it’s the kind of thing worth getting outside eyes on. If that’s where you are, it’s the sort of gap an AI consultant for small business would normally sort out in a couple of sessions, building a shared template of what raw material has to go in before AI touches it, rather than leaving six different people to prompt six different ways.
One more thing worth saying plainly. None of this is about writing longer proposals. The bakery supply quotes I mentioned earlier ranged from two pages to eleven, and length had almost no relationship to which ones felt specific. The eleven-page one was the most generic of the lot, just more of it. Specificity beats length every time a prospect is comparing more than one quote, which, if you’re doing well enough to have competition for the job, is most of the time.
Frequently asked questions
Can prospects tell if a proposal was written by AI?
Most of the time yes, though not because of clunky phrasing. It’s the absence of anything specific to their business that gives it away. A proposal with their actual numbers, their actual competitor names, and their own words fed back to them reads as researched regardless of which tool wrote the sentences.
Is it bad to use AI to write proposals at all?
No, and trying to hide that you used it wastes energy that’s better spent on the input. The problem was never the tool, it’s typing a one-line prompt with nothing in it and sending the first draft out unedited.
How much time should a proper proposal process take?
For most small jobs, twenty to thirty minutes on top of whatever you already spend, mostly pasting in real call notes and one competitor name, then five minutes asking the AI to flag and cut anything that could apply to any business. That’s a small time cost for a document that’s meant to win you the work.
What’s the single fastest fix if I only have five minutes?
Take your draft, paste it back into the same chat, and ask it to list every sentence that could apply to a different business in the same industry. Delete those sentences. It won’t fix everything but it removes the worst of the generic filler in under a minute.
Related reading: The Best AI Tools for Small Business Owners (What I Pay For) and Why Your AI-Written Proposals Are Losing You the Job (Even Good Ones).
For the bigger picture, see my full guide to AI marketing.
Useful references
Related: drop shipping take business next level.
