A freelancer on Twitter posted about using ChatGPT to write proposals. He was three months into freelancing, struggling to close clients, and someone told him AI could fix his conversion rate. He tried it. He sent the proposal the same day. The client replied: “Thanks so much, we’ve decided to go in another direction.” He posted his question publicly: “What went wrong? The proposal looked great to me.” Nobody had a useful answer for him. This post is the answer he didn’t get.
The proposal looked great to him because it was structurally clean, professionally written, and covered all the obvious sections. It also opened with “I hope this message finds you well. I understand you’re looking for a web design solution that meets your business needs.” It included a case study about “a client in the digital space who was looking to improve their online presence and achieved significant results.” And the pricing section explained that “the investment reflects the high-quality work and value delivered.”
None of those sentences were lies. All of them were invisible. The client read that proposal and saw someone who hadn’t listened to anything they’d said.
The lesson here isn’t “don’t use AI.” It’s more precise than that: AI handles structure well and persuasion badly. The freelancers losing deals in 2026 are the ones handing AI both jobs.
What clients actually notice when a proposal is AI-written
Clients can’t reliably detect AI text. Studies confirm this, humans identify AI writing at only slightly better than chance. But clients can detect a proposal that wasn’t written for them, and that’s a different problem entirely.
The tells are specific. If you know what to look for, you can spot a first-draft AI proposal in under a minute.
The opener that sounds polite but says nothing. “I hope this message finds you well.” “Thank you for the opportunity to submit this proposal.” “I understand you’re looking for a solution that meets your needs.” Every sentence in this category is a signal that no one is home. It takes zero knowledge of the client to write any of them.
The scope section full of adjectives instead of specifics. “High-quality deliverables.” “Comprehensive strategy.” “Tailored approach.” “Best-in-class solutions.” These phrases are infinitely interpretable. When scope is described in adjectives, the client has to imagine what they’re buying, and they’ll imagine wrong. Scope should be a list of named things: pages, screens, deliverables, revision rounds, file formats. Not qualities.
The case study that doesn’t name anything. “I helped a client in the e-commerce space increase their conversions and see significant improvement in their key metrics.” That sentence could be cut-and-pasted into any proposal for any client anywhere. A real case study has a name or at minimum a clear industry, a specific problem, a specific action, and a number. “I redesigned the checkout flow for a DTC skincare brand, their cart abandonment was at 74%, and we got it to 51% in eight weeks.” That’s a case study.
The pricing rationale that justifies nothing. “This investment reflects the quality of work and the value it will bring to your business.” That sentence is filler. It doesn’t explain why $8,500 specifically. A real pricing rationale says: “This is scoped at 6 weeks because the design system alone is a 3-week build, you said you need it to work across your app and your marketing site. The testing phase is costed separately because you specifically mentioned that your last vendor launched without QA and it cost you three weeks of cleanup.”
The pattern across all of these: AI writes the version with no specific knowledge. When a client reads a proposal with no specific knowledge in it, they don’t think “this person used AI.” They think “this person didn’t listen.”
The parts of a proposal AI genuinely improves
Honest acknowledgment here: AI is genuinely useful for proposals. The freelancers who dismiss it entirely are working harder than they need to.
First draft structure. AI can take your bullet-point notes from a discovery call and return a structured proposal draft in 60 seconds. The draft will need significant editing, but “edit a draft” is faster than “write from scratch.” For freelancers who stall at the blank page, this alone is worth the tool.
Rewriting for clarity. Paste in a section you wrote and ask AI to make it more direct. Ask it to cut it by 30%. Ask it to remove passive voice. These are mechanical tasks and AI handles them well. The content stays yours; the presentation gets better.
Pricing tier naming. Naming tiers “Bronze / Silver / Gold” is generic. Naming them after outcomes (“Foundation / Growth / Complete”) is better. AI can generate 10 name options in 20 seconds. You pick the one that fits.
FAQ sections. The questions clients typically ask about scope, timeline, revisions, and payment terms are predictable. AI can draft a full FAQ in the first pass that you edit for accuracy.
Turning bullets into prose. If you work better with bullet points during the discovery call and need those notes to become readable paragraphs, AI handles the translation cleanly.
Suggesting headings. When you’re deep in a proposal and need to break up a dense section, AI is fast at proposing section headings and subheads.
All of this is structural and mechanical. None of it requires specific knowledge of your client. That’s the dividing line.
The parts you must write yourself
Three sections. If you outsource any of these to AI without heavy editing, you lose close rate.
The opening paragraph. This paragraph must contain something specific the client said or showed. Not a category observation, something from the actual call. “You mentioned the current site makes your agency look like a startup, not the 12-person team you’ve built.” “You said the board wants this shipped before Q3 because that’s when the Series B closes.” “You used the word ‘embarrassing’ when you described the current checkout flow, and I wrote it down.”
AI cannot write this paragraph. The information doesn’t exist anywhere except your call notes.
Here’s what the difference looks like:
AI first draft: “Thank you for taking the time to speak with me. I understand you’re looking for a web design solution that will help your business grow and better connect with your target audience.”
Human rewrite: “You mentioned the current site makes your agency look like a startup, not the 12-person team you’ve actually built. The redesign we discussed directly addresses that perception gap, before a prospective client reads a single line of copy, the site needs to communicate that you’re the grown version of the business you were three years ago.”
Same topic. The second one signals that you were in the room.
The case study. A case study with no specific number in it isn’t a case study, it’s a character reference. “I helped a client in the marketing space improve their results” tells the reader nothing and makes your credentials invisible. A real case study is three components: the problem (specific), the action (what you actually did), the outcome (a number). “A B2B SaaS client came to me with a pricing page that was converting at 1.2%. I restructured the page around outcome-first tier naming and removed four competing CTAs. Conversion went to 3.4% in 30 days.” That sentence closes deals. AI can’t write it because it doesn’t know it happened.
The pricing rationale. The price you charge is defensible when it’s tied to this specific scope, this specific client constraint, and this specific risk. Generic AI-written rationale (“this investment reflects the quality of work”) sounds exactly like a placeholder because it is one. Your rationale should reference something concrete: “The $9,500 price reflects a 7-week build, which is longer than a standard redesign, the additional three weeks are specifically for the design system you need across three properties. A single-site redesign would be $6,500.”
The rule: any sentence in your proposal that could appear, unedited, in a proposal for a different client in a different industry is a sentence that shouldn’t be there. AI first drafts are full of these sentences. Your job is to find and replace every one.
The workflow that actually works
This is the sequence. Not “blend AI and human writing”, the specific order matters.
Step 1: Take real notes during the discovery call. Write down the client’s exact phrases. “Embarrassing.” “We needed this done six months ago.” “My co-founder doesn’t trust agencies.” These words are your raw material. AI can’t invent them.
Step 2: Dump your notes into AI with a structural prompt. Something like: “Here are my notes from a discovery call with a B2B SaaS client looking for a pricing page redesign. Build a proposal draft with these sections: opening, problem statement, proposed scope, timeline, pricing tiers, and next steps.” You’ll get a usable skeleton in under two minutes.
Step 3: Replace the opening paragraph entirely. Don’t edit the AI opener, delete it. Write a new paragraph from scratch using one or two specific things the client said. This takes 10 minutes. It’s the highest-leverage 10 minutes in the proposal.
Step 4: Replace or expand the case study. Find the AI placeholder (“I helped a client achieve significant results”) and replace it with a real story from your history that is relevant to this client’s problem. Add a number.
Step 5: Rewrite the pricing rationale. Go line by line. Every generic phrase gets replaced with something tied to this specific scope and this specific client’s situation.
Step 6: Use AI to polish. Once the three human-written sections are done, paste the whole proposal back into AI and ask it to tighten the language, flag vague scope language, and check for passive voice. This is editing, not rewriting, a safe use.
Step 7: Run the pre-send check (below).
Total time: 45–75 minutes for a solid proposal, down from 3–4 hours writing from scratch. The time savings come from the structural skeleton. The close rate comes from the three sections you wrote yourself.
5 prompts that improve proposals
These are specific prompts, not generic ones. Paste the relevant section into your AI tool after each prompt.
Prompt 1, Fix a bad opener: “Rewrite this proposal opening so it leads with the client’s situation instead of my credentials. The client is [describe their situation in one sentence]. Here’s the current opener: [paste section]”
Prompt 2, Name pricing tiers: “Generate 5 pricing tier name options for a brand identity project at $5,000 / $8,000 / $12,500. The client is a SaaS startup that wants to look like a funded company. Names should be outcome-focused, not feature-focused.”
Prompt 3, Expand a thin case study: “My case study is one sentence: I redesigned the checkout flow for an e-commerce client and cart abandonment dropped from 68% to 43%. Expand this into two paragraphs: one covering the problem and context, one covering the approach and outcome. Keep it specific, don’t add claims I didn’t make.”
Prompt 4, Flag vague scope: “Review this scope section. Identify every phrase that is vague enough for a client to over-imagine what’s included or misunderstand what they’re getting. Flag each one and suggest a more specific replacement. Here’s the scope: [paste section]”
Prompt 5, Write a risk reversal: “Write a 2-sentence risk reversal for a $7,500 SEO project where the client is worried about not seeing results in time. The risk reversal should reduce perceived risk without making guarantees I can’t keep.”
Prompt 4 is the most underused of the five. Vague scope is the number-one reason clients stall after reading a proposal, not price, not timeline. Running your scope section through an AI flag pass before sending catches the phrases you’ve stopped noticing because you wrote them.
The check before sending
Three questions. If the answer to any of them is no, the proposal isn’t ready.
Does the opening mention something specific about this client, not a category? “You work in marketing” is a category. “You said your current retainer structure means you’re getting paid 90 days after you deliver the work” is specific. If your opening could describe any client in the same industry, rewrite it.
Does the case study have a real number in it? Percentages, dollar amounts, time frames, volume, any real number. “Significant improvement” is not a number. If you genuinely don’t have a number for the relevant case study, pick a different case study. If you don’t have a case study with a number, ask a past client for a specific outcome and add it before you send another proposal.
Is there at least one sentence in this proposal that could only have been written for this client? One sentence. Tied to their specific situation, their specific words, their specific constraint. If you can’t find that sentence, you’re sending a template. Templates lose to proposals.
If the answer to all three is yes, send it. Send it within 48 hours of the discovery call, not because of a general rule about responsiveness, but because the client’s urgency and memory of the conversation are both highest in that window. Every day past 48 hours is close rate bleeding out.
Start using AI without outsourcing persuasion
The freelancers who are going to win with AI proposals aren’t the ones generating the fastest first drafts. They’re the ones who understand that AI is a structural tool and persuasion is still a human job.
Use AI for the skeleton. Write the muscle yourself. The skeleton is structure, prose clarity, and formatting. The muscle is the specific opening, the real case study, and the pricing rationale that only makes sense for this client.
That combination, AI handling the predictable work, you handling the irreplaceable work, is what a 60% close rate looks like in 2026.
Waco3 is built around this workflow. The structural framework is already in the tool. Your job is to drop in the three sections that only you can write. If you want to see how it works with your own proposals, start a free trial.
Related reading: Once the proposal is out, the follow-up is where most deals are actually won or lost. See How to Follow Up After Sending a Proposal for the exact sequence. If you want to track whether the client actually opened what you sent, How to Know If a Client Read Your Proposal covers four practical methods. And to let AI draft the follow-up itself, see How to Use AI to Write a Follow-Up Email After a Proposal.
Ready to send stronger proposals?
Build, send, and track proposals in one place so follow-up is easier.
Start your free trial →
