Sales emails
AI prompts for sales emails that stay personal
The best sales email prompts give the AI the real facts from your call and your deal, one pain point and one ask, and an order to invent nothing about the prospect. That keeps the email short, specific and true.
macOS 13+·Windows 10+·Free for 3 days, no card
Updated
>_Templates
Why most sales email prompts give you blanks
The templates people find online hand you a prompt full of brackets, and the AI fills every gap you leave with a guess.
A template that asks you to fill in the name, the company and the problem in brackets puts the hard part back on you. You still have to find the name, the problem and the history. An AI with no facts writes what sounds right for any prospect, and that is why so many sequences read alike.
Anthropic's test for Claude: show a prompt to a colleague with minimal context, and if they would be confused, Claude will be too. A sales prompt with no deal facts fails that test. A new rep could not write the email from it either.
Patterns
Three sales prompts to copy
Each prompt below is filled in with invented details, keeps to one pain point and one ask, and names the limit on length. Swap in your own facts.
Pattern 1: a cold email to someone you have read about. Prompt: Write a cold email to Noor Sato, head of operations at Rushlight Logistics. Goal: get a 15-minute call. Facts: on 20 October she posted that her drivers lose about 40 minutes a day to paper delivery notes. We make a delivery-notes app for fleets of 20 to 200 vans. These are all the facts I have about her and about Rushlight. One pain point, one ask. Under 90 words, in plain words. Use only the facts listed here, and leave out anything you would need to guess.
Pattern 2: reply to a price objection. Prompt: Reply to Quentin Rowe at Stellar Fabrication, who wrote: "Tallowby quotes $310 a month and you are $380." Goal: keep the deal without cutting the price. Facts: our $380 plan includes onboarding for 3 sites and phone support. Quentin did not say what the Tallowby quote includes. Steps: thank him, say what our price includes, then ask what the Tallowby quote covers. Make no claim about Tallowby. Calm and curious. Under 110 words.
Pattern 3: a deal that went quiet. Prompt: Write a follow-up to Teodor Vance at Westbrook Clinics. I sent the proposal on 9 October and I have heard nothing for three weeks. On our last call he said he needs sign-off from his finance director, Nora Ward. Goal: one clear next step. Offer to send Nora a one-page summary of the proposal. Friendly, no pressure. Under 70 words, ending in one question.
Gaps
What a generic prompt misses
A prompt with no deal data forces the AI to invent a story about your buyer.
- Their words. The pain point the buyer named on the call, in the way they said it.
- The deal. Stage, price quoted and next date.
- The people. Who else signs off, and what matters to them.
- Your promises. What you said you would send and by when.
- Your limit. One ask, and a word count.
The notes from your last call and the deal record hold most of this. The AI cannot read them unless you paste them in. Scrubbr reads the call notes and the deal record when you say the line, so you skip the copying.
The rule
The rule: the AI must not invent facts about the prospect
A made-up compliment or a made-up pain point is worse than a plain email, because the prospect knows it is false.
- Tell it: "Use only the facts in this prompt about the prospect." The pattern prompts above do this.
- Allow "I don't know". Anthropic's blog lists explicit permission to be unsure as a way to cut hallucinations.
- Give it a way out when facts are thin. Perplexity's API guide tells developers to give the model an explicit way to say it found nothing, instead of speculating. It is a developer guide, and the same idea works in a sales prompt.
- Compare each line about the prospect with your call notes, and cut what you cannot trace. Microsoft gives the same advice for Copilot: check each response before you rely on it.
This page gives no reply-rate numbers. None of the guides cited here gives any. Test your emails on your own list.
Worked example
From "email the lead" to a prompt with facts
Reuben Shaw sells scheduling software, and he needs a follow-up to Ursula Voss, the day after her demo.
In ChatGPT
You say
optionemail the quillmoor lead after the demo, thank her, one pain point, ask for a call with the owner, don't make up anything about her clinic
Scrubbr finds
Granola · Quillmoor Vets demo, 27 October
HubSpot · Quillmoor Vets deal: Demo done, $4,560 a year, owner Reuben Shaw
Gmail · Ursula Voss, first email asking for a demo, 19 October- Knowledge base · Tidewell plans: cancel by text reply is on the standard plan
ChatGPT · Prompt
Goal: Write the follow-up to Ursula Voss, practice manager at Quillmoor Vets, the day after the demo. The goal is one next step: a 20-minute call with the owner, Dr. Tomasz Wren, who signs off on spending.
Facts: The demo was on 27 October. Ursula said about 12 appointments a week are missed across her 2 clinics, and the front desk calls each client by hand. She liked the text reminders and asked whether clients can cancel by replying to the text. Tidewell supports that on the standard plan. The deal in the CRM is at the stage Demo done, at $4,560 a year for 2 clinics. These are all the facts we have about Quillmoor.
Steps: 1. Thank her for her time on the 27th. 2. Name the one pain point: missed appointments and the calls by hand. 3. Answer her question: yes, clients can cancel by text. 4. Ask for a 20-minute call with Dr. Wren on 5 or 6 November.
Limits: Friendly and plain, one professional to another. Use only the facts listed above about Ursula and Quillmoor. Tell Reuben about any gap, and keep it out of the email. One pain point, one ask. Leave out the price.
Format: Plain text. Put the subject line first, then the body. Under 100 words.
A great answer: Ursula can reply in one line. Every fact in it is something she said or asked, so nothing in it can be wrong about her clinic.
Demo data. The people, company and facts are invented.
Questions
Sales email prompt questions
What are the best AI prompts for sales emails?
The best prompts carry real facts: what the prospect said, the deal stage, one pain point, one ask and a length limit. Anthropic's blog says to be specific about the audience, the constraints and the output structure.
Can ChatGPT write a cold email that does not sound like a template?
Yes, if you give it one fact that only you know, such as something the prospect wrote or said, and cap the length. Without a fact, it writes what fits any prospect.
How do I stop AI inventing facts about my prospect?
Tell it to use only the facts in the prompt, and give it permission to say when it is not sure. Then read each sentence about the prospect against your notes before you send. Microsoft's guide asks for the same care with every Copilot answer.
How do I write a follow-up email after a demo?
Send it the next day, from your demo notes. Thank them, name the one pain point they gave you, answer any question they asked and make one ask. Keep every fact to something they said.
Do AI sales emails get more replies?
We do not know, and this page quotes no reply-rate numbers. None of the official guides cited here gives any. Test two versions on your own list and compare.
Sources
Where these facts come from
Scrubbr writes this prompt for you. Hold Option and say one line.
Try Scrubbr free for 3 days
Download Scrubbr, connect your tools, and give your first prompt in two minutes.
macOS 13+·Windows 10+·Free for 3 days, no card