Writing Sales Copy That Converts with AI
Without Sounding Like Every Other Rep in the Market
If you did the exercise at the end of Post 3, running call transcripts and paying attention to the exact words your prospects use to describe their problems, you have something genuinely valuable sitting in front of you right now. Most reps don’t. And that gap is exactly where this post lives. AI sales copywriting is where the magic starts to happen.
Here is the central problem with AI-generated sales copy, and it is not the one most people think. The problem is not that AI writes badly. In a technical sense, it writes quite well. The sentences are clean. The structure is logical. The tone is professional. The problem is that it writes the same way for everyone.
When every rep on every team in every industry is prompting an AI tool with ‘write me a cold email for a VP of Sales at a mid-market SaaS company,’ the output starts to converge. The same phrases. The same rhythm. The same opening gambit about ‘helping teams like yours.’ Your prospect’s inbox fills up with professionally written, competently structured, completely indistinguishable outreach, and they stop reading all of it.
The reps who are winning with AI-assisted copywriting are not using AI to write faster versions of the same email everyone else is sending. They are using AI to write in a voice their buyer actually recognizes because it sounds like the buyer’s own voice reflected back at them.
That is a different skill. And it starts well before you open any AI tool.
The Voice Mining Technique
In Post 2, we talked about buying signals — behavioral data that tells you a prospect is in the market. Voice mining is the copywriting equivalent. It is the practice of collecting the exact language your buyers use to describe their problems, and feeding that language into your AI prompts so the output sounds like them, not like you.
Here is where that language lives, and how to find it:
Your call transcripts. This is why the exercise from Post 3 matters so much. When a prospect says ‘we’re just hemorrhaging time on manual data entry’ or ‘our reps are drowning in admin,’ those phrases are gold. They are not polished marketing language. They are raw, emotional descriptions of a real problem. AI copy built around those phrases will land differently than copy built around your product’s feature set.
LinkedIn comments and posts. Find five or ten posts in your target buyer’s feed that got real engagement. Read the comments. The people leaving thoughtful comments are telling you exactly what resonates with that audience, in their own words. Screenshot them. Paste them into a document. Use them.
Review sites. G2, Capterra, Trustpilot, Amazon — wherever your buyers leave reviews of products or services in your category. They are not writing marketing copy. They are writing honest descriptions of what they needed, what worked, and what frustrated them. That language is a direct line into how your buyer thinks.
Reddit and niche communities. The subreddit where your buyer hangs out is one of the most honest language sources available. People on Reddit write the way they actually think, without the filter they use in professional settings. The vocabulary, the frustrations, the inside jokes — all of it is useful.
Your own CRM notes and email threads. Past conversations with buyers who became customers are a library of language that already converted. If a prospect said something in an email that made you think ‘that’s exactly it’, save it. That phrase works.
AI does not know how your buyer talks. You do. The quality of your AI output is directly proportional to the quality of the language you feed into it.
Prompting Frameworks That Actually Work
Most reps prompt AI the way they search Google, short, vague, and optimistic. ‘Write a cold email to a VP of Sales.’ That prompt will produce a cold email. It will not produce a good one.
The prompts that produce copy worth sending treat the AI as a skilled writer who needs a thorough brief, not a vending machine that dispenses content. Here are three frameworks that consistently produce output worth editing.
The Situation-Pain-Stakes framework. Give the AI the prospect’s situation, the specific pain they’re experiencing, and the stakes if that pain goes unaddressed. The more specific each element, the sharper the output.
Example prompt:
I’m writing a cold email to a VP of Sales at a 200-person B2B software company. Her team of 18 reps is spending roughly 40% of their time on CRM data entry instead of selling. Her Q2 number is at risk because of it, and she knows it. Write a 5-sentence cold email that acknowledges this specific pain, hints at a better outcome, and asks for 15 minutes. Use plain, direct language, no buzzwords. Do not mention our product by name.
The Buyer’s Voice frame. Feed the AI the exact phrases from your voice mining work and ask it to build the email around them. This is where call transcripts and review site language pay off directly.
Example prompt:
Here are three phrases my target buyers use to describe their problem: ‘we’re flying blind on pipeline,’ ‘I don’t know which deals are real until it’s too late,’ and ‘my forecast is a guess every single week.’ Write a cold email opener and subject line that uses this language, not paraphrases of it, but this language, to open a conversation about forecasting accuracy. Keep it under 100 words.
The Before/After/Bridge frame. Ask the AI to describe where the prospect is now (before), where they want to be (after), and hint at the bridge between the two — without pitching a product. This structure works for cold emails, LinkedIn messages, and follow-up sequences equally well.
Example prompt:
Write a LinkedIn connection request message for a Head of Revenue Operations at a mid-market company. Before state: she’s manually stitching together three spreadsheets every Monday morning to produce a pipeline report that’s outdated by Tuesday. After state: she has one reliable source of truth that updates automatically, and she can stop being the person everyone blames when the numbers are wrong. Write a connection request that acknowledges the before state in a way that makes her feel understood, 300 characters max.
When to Draft vs. When to Refine
There is a meaningful difference between using AI to draft your sales copy from scratch and using AI to refine copy you’ve already written. Both are valuable. They are not the same thing, and the best use of each depends on the situation.
Use AI to draft when: you need volume, when the message type is relatively standard (initial outreach, meeting confirmation, generic follow-up), or when you are staring at a blank page and need a starting point to react to. AI is faster than you at generating a first draft. Use that.
Use AI to refine when: the message is high-stakes, a re-engagement after a long silence, a proposal cover letter, or a follow-up after a difficult call. Write the first version yourself. Then ask AI to tighten it, sharpen the language, flag anything that sounds generic, or suggest a stronger opening line. Your instincts plus AI editing usually beats AI drafting alone.
The practical rule: the higher the stakes and the more specific the situation, the more your own thinking should anchor the message, with AI in a supporting role. For lower-stakes, higher-volume messages, let AI drive and edit selectively.
Before and After: What the Difference Looks Like
Abstract advice about voice and resonance is easier to absorb when you can see it side by side. Here is the same outreach scenario written two ways.
The scenario: following up with a VP of Operations at a distribution company who attended a webinar on supply chain visibility three weeks ago but hasn’t responded to two previous emails.
Generic AI output (no voice mining, basic prompt):
Hi [Name], I wanted to follow up on the webinar you attended a few weeks ago. I hope you found it valuable. At [Company], we help distribution businesses like yours gain real-time visibility into their supply chain operations, reducing delays and improving efficiency. I’d love to schedule a brief call to discuss how we might be able to help your team. Are you available this week for a 15-minute conversation?
Voice-mined, well-prompted AI output:
Hi [Name], you sat through our supply chain visibility webinar three weeks ago and then went quiet, which usually means one of two things: either it didn’t land, or it landed a little too close to home and you’re now thinking about what it would actually take to fix it. If it’s the second one, I have 15 minutes and a specific question I’d like to ask you. Worth a conversation?
Both emails were produced with AI assistance. The first used a generic prompt. The second used a voice-mining approach, fed with real language from discovery calls with distribution operations buyers, and a specific prompt about the psychology of a non-responder. Same tool. Very different outcome.
The prompt is the strategy. The AI is just the execution. Get the strategy right and the copy takes care of itself.
A Note on LinkedIn Messages and Proposals
Everything in this post applies directly to LinkedIn outreach, connection requests, InMail, and follow-up messages in the thread. The voice mining technique is especially powerful here because LinkedIn is where your buyers are already publishing their own language. Read their posts before you write to them. Use what they’ve told you they care about.
For proposals, AI is most useful in two places: the cover letter and the executive summary. These are the sections most buyers actually read, and they are the sections most reps write last, in a hurry, after spending most of their energy on the pricing table. Flip that. Use AI to draft a sharp, buyer-specific cover letter first, built around the pain language from your discovery call, and let that set the tone for everything that follows.
What’s Coming in Post 5
The copy is working. Conversations are happening. Deals are moving into your pipeline. Now the question is whether your CRM is helping you close them or just recording that you’re trying.
Post 5 covers AI and CRM, what’s actually available now inside the platforms most teams are already paying for, how AI is eliminating the data entry problem that kills adoption, and how deal intelligence is changing what pipeline management looks like for both reps and managers.
Subscribe below if you want it delivered directly. It goes out next week.
One Thing to Do Before Post 5
Start a voice mining document. Open a blank note, your phone, your laptop, wherever, and title it ‘Buyer Language.’ Every time a prospect says something that captures their problem in a way that feels raw and real, write it down. Exact words, no editing.
Give it two weeks. You’ll have more useful copy material than any AI tool can generate on its own.
About This Series
The AI Sales Stack is a 7-part series on FillTheFunnel.com covering the tools, tactics, and workflows that are rebuilding the modern sales process. Miles Austin has been covering sales technology since 2006 and works with sales leaders and entrepreneurs to navigate the intersection of sales strategy and emerging technology.