Your sales team is spending real selling time on email admin. That is the thing worth fixing first.
Not the CRM integration. Not the AI strategy deck. The follow-up email that someone on your team writes, deletes half of, rewrites, second-guesses, and sends twenty minutes after they should have — every single day, for every single prospect.
Here is one approach worth trying.
The problem is not laziness. It is cognitive switching.
Writing a follow-up email requires a salesperson to stop thinking like a salesperson — "what does this person need to hear next?" — and start thinking like a copywriter — "how do I phrase this so it doesn't sound like a template?" That switch costs more than the time it takes — the salesperson re-enters drafting mode mid-afternoon, and the call context they were holding starts to fade.
The objection I'd expect here: surely a generic AI email sounds worse than a personalised one the salesperson wrote themselves? Yes. Which is why the workflow below is not "let AI write the email." It is "let the salesperson think like a salesperson and let AI do the drafting."
The workflow, step by step
What you need: A note-taking habit of thirty seconds, and access to ChatGPT, Claude, or Gemini — any of the major models will do this well.
Step 1: Capture the call context immediately after the call ends.
This is the step most people skip, and it's why their AI output is generic. Before you close the call window, type — in bullet points, not sentences — the three things that were actually said:
- What the prospect mentioned as their main frustration
- Any specific detail they gave you (a number, a deadline, a name, a competitor they mentioned)
- What you said you'd send them
That's it. Thirty seconds. You don't need full notes. You need signal.
Step 2: Feed the signal into your AI tool with a clear prompt.
Here is a prompt structure that works:
"Write a follow-up email to [prospect name] at [company]. We spoke today. They mentioned [specific frustration or detail]. I said I'd send [specific thing]. The goal of this email is [one sentence: e.g., to share the resource I promised and propose a next call]. Keep it under 150 words. No corporate language. Sound like a real person."
The specific details you captured in Step 1 go directly into the brackets. The AI drafts from signal, not from nothing.
Step 3: Edit for the one thing AI consistently gets wrong.
AI follow-up emails tend to over-explain the value proposition. They add a sentence — sometimes two — re-pitching your product when what the prospect needs is just the next small step. Read your draft and ask: is there a sentence here that's trying to resell them? If yes, delete it. The follow-up email's job is to move the conversation forward, not to close the deal on its own.
This edit takes sixty seconds if you know what to look for.
Step 4: Save your best prompts, not your best emails.
After two weeks of this, you will have written some prompts that produce good drafts reliably and some that don't. Keep the good ones in a shared document — a simple Google Doc is fine. This is your team's prompt library. A prompt library beats a template library because the output changes with every conversation — a template doesn't.
What this actually saves
I won't give you a made-up number. The real cost is cognitive switching, not just clock time. A salesperson still mentally in the conversation writes better bullet notes, which produce a better prompt, which produce a draft that needs less fixing.
The whole process — notes, prompt, edit — should take under four minutes. Composing a thoughtful follow-up from scratch, for most people, takes longer.
Consistency matters more than speed here. Imagine you run a small sales team of four people. In that team, salespeople follow up at different speeds, with different quality, based on energy levels and writing ability. The strongest follow-up style on the team is hard for everyone else to replicate. A shared prompt library gives every salesperson on the team access to the same drafting quality — without flattening their individual voice.
The limits you should know before you start
AI will hallucinate details it doesn't have. If you give it a vague prompt, it will invent specificity — wrong dates, wrong names, wrong product features. The more real signal you put in, the less room the model has to fill gaps with fiction.
AI also cannot read the emotional register of a call. It doesn't know that the prospect sounded hesitant, or rushed, or unusually warm. That calibration still belongs to the salesperson. A follow-up email where the tone is wrong — too pushy after a cautious conversation, too casual after a formal one — will undo the work of a good call. Read the draft with that in mind before you send it.
And if your prospect is someone who will recognise AI-written prose immediately — a writer, a marketer, a founder who thinks carefully about language — add more of your own voice in the edit.
The one thing to do today
Open ChatGPT, Claude, or Gemini. Find the last follow-up email you sent or received that felt generic. Copy it in. Then type this:
"Rewrite this follow-up email so it sounds like a specific person wrote it after a real conversation. Make it under 100 words. Remove any sentence that re-explains the product's value. Keep the next step clear."
Run it. Compare the two. You are not adopting a new system today — you are checking whether the output is worth your time.
Next time you finish a real call, try Step 1 before you do anything else.