AI FOR SALES
ChatGPT Prompts for Sales: 10 That Hold Up
A five-part prompt structure and ten sales prompts that work in ChatGPT, Claude, Gemini and Copilot, a worked margin and discount example, and the email, LinkedIn and data rules every seller should know.
The best ChatGPT prompts for sales give the model a role, your real context (ideal customer profile, product, deal stage), one clear task, a fixed format and a check, such as "use only the facts I paste and mark anything you guessed". Below are ten prompts that hold up across the sales cycle, from account research to forecast review, and they work the same way in Claude, Gemini and Microsoft Copilot. Use them on public information or in your company's approved AI tool, never paste customer lists or deal data into a public account, and check every fact and every number before it reaches a buyer.
Checked on 11 October 2026 against the sources listed under this post. Product claims are dated October 2026.
Why most sales prompts fail
"Write a cold email for my product" produces the email every buyer deletes. The model has no idea who the buyer is, what they care about or what you can prove, so it fills the gap with generic praise and invented specifics. Salesforce's State of Sales announcement (3 February 2026, 4,050 sales professionals) found that sellers expect AI agents to cut prospect research time by 34% and email drafting by 36%, and that 74% of sales professionals were focusing on data cleansing. The time savings are real only when the input is real.
A prompt that works has five parts:
| Part | What to include | Example |
|---|---|---|
| Role | Who the model should act as | "You are a B2B account executive selling to plant managers" |
| Context | Your ICP, offer, proof and the source text | Pasted website text, your case study summary |
| Task | One job | "Draft a 4-line first email" |
| Format | Length, structure, tone | "Under 80 words, one question, no superlatives" |
| Check | How to handle gaps | "Use only facts given; list anything you assumed" |
Ten ChatGPT prompts for sales that hold up
1. Ideal customer profile from won deals. "Here are our last 10 won deals (industry, size, trigger, deal value, anonymised). Find the common pattern and write an ideal customer profile with must-have and nice-to-have criteria."
2. Account research from public text. "Using only the company text I paste, list facts relevant to our offer, each with the sentence it came from. Write 'not stated' where the text is silent."
3. Hypothesis for the first call. "From those facts, write two hypotheses about a problem this account may have that our product addresses. Label each as an assumption to test."
4. First email. "Write a first email under 80 words to the operations director: reference one verified fact, ask one question, offer a 20-minute call. No flattery, no attachments."
5. Follow-up that adds value. "Write a second email that shares one useful insight for their industry, without repeating the first email or pressuring for a reply."
6. Discovery question bank. "Write 12 discovery questions grouped by situation, problem, impact and decision process. Avoid yes/no questions."
7. Call notes to CRM. "Turn these call notes into CRM fields: pain, impact, decision makers, timeline, budget signals, next step with a date. Mark anything not said as unknown."
8. Objection role-play. "Act as a sceptical procurement manager who thinks our price is 20% too high. Push back three times, realistically. After the role-play, critique my answers."
9. Proposal outline. "Draft a proposal outline: the client's problem in their words, our solution, scope and exclusions, timeline, commercial summary, risks and next steps."
10. Pipeline review. "Here is my anonymised pipeline. Flag deals with no next step, a close date in the past or no activity in 21 days, and suggest one question for each."
Two habits make these prompts better over time. First, save the ones that work in a shared document or a custom assistant (ChatGPT Projects, Claude Projects or a Gemini Gem), with your ICP and proof points attached, so every rep starts from the same context instead of retyping it. Second, review the output against what actually happened: which emails got replies, which discovery questions opened the conversation, which forecast flags turned out right. Edit the prompt, not just the output, when something misses. A prompt library that the team improves every month is worth more than any single clever prompt.
This tutorial from Pragmatic Works AI Academy, Smarter Prospecting with AI: Research & Emails That Win Responses, covers AI-assisted research and outreach. It is one of the lessons in our free AI for Sales and Business Development course; EDWartens is not affiliated with the creator.
Worked example: the discount maths the AI gets wrong
AI assistants are good with words and unreliable with commercial arithmetic, especially margin and markup, which sound alike and are not. Suppose a buyer asks for a 15% discount on a product with a list price of $100 and a cost of $68.
| Measure | Formula | At list price | After 15% discount |
|---|---|---|---|
| Selling price | List x (1 - discount) | $100.00 | $85.00 |
| Gross profit | Price - cost | $32.00 | $17.00 |
| Gross margin | Profit / price | 32.0% | 20.0% |
| Markup | Profit / cost | 47.1% | 25.0% |
Check it: 100 x 0.85 = 85; 85 - 68 = 17; 17 / 85 = 0.20; 17 / 68 = 0.25. A 15% discount cut the profit per unit by 47% (from $32 to $17). To earn the same $32,000 gross profit you made on 1,000 units at list price, you would need to sell 32,000 / 17 = 1,883 units at the discounted price, 88% more volume.
Ask a chat assistant "what is my margin after a 15% discount on a 32% margin product?" and you will sometimes get "17%" (32 minus 15), which is wrong. Use the AI to explain the logic and draft the reply to the buyer ("we can offer 5% for a 3-year term"), and do the numbers in a spreadsheet. Our course has a full module on margin, markup and discount maths for exactly this reason.
The rules: email law, platforms and data
AI makes it easy to send more. These rules decide what you may send.
Email. The FTC's CAN-SPAM compliance guide says the law "makes no exception for business-to-business email", requires accurate headers, non-deceptive subject lines, a valid physical postal address and a working opt-out, and says opt-out requests must be honoured within 10 business days. It lists penalties of up to $53,088 for each separate email in violation. Other regions, including those under the GDPR, add consent and data protection rules on top, so check where your recipients are.
LinkedIn. Section 8.2 of the LinkedIn User Agreement prohibits using software, scripts, robots or other means "to scrape or copy the Services, including profiles and other data". AI browser extensions that bulk-copy profiles fall into that.
Customer data. Never paste your CRM export, customer list, price list or a client's documents into a public AI tool. Use your CRM's built-in AI or the enterprise tool your company approves, under its policy.
| Do | Don't |
|---|---|
| Research one account from public sources | Bulk-scrape profiles with an extension |
| Verify every fact before sending | Send AI claims you have not checked |
| Keep emails short with a clear opt-out | Send hundreds of near-identical "personal" emails |
| Use approved tools for CRM data | Paste customer lists into a public chatbot |
| Do margin maths in a spreadsheet | Trust the chat's arithmetic on discounts |
Where the CRM's own AI fits
General assistants are best for thinking and drafting. The AI inside HubSpot, Salesforce and Zoho CRM, as of October 2026, is better for work that needs your live data: summarising call recordings into fields, suggesting next steps, scoring leads and drafting sequences. Because it works under your company's data agreement, it is usually the right place for anything involving real customer records. Treat its outputs the same way: check the amount, the stage and the next step before you save.
For more ways to structure prompts, see our comparison of free generative AI and prompt engineering courses and our guide to the best free AI courses with certificate.
A free course for sellers
The free AI for Sales and Business Development course follows the B2B sales cycle: where AI fits, prompting for sales, ideal customer profiles and account research, outreach, discovery, call notes to CRM, objection role-play, proposals and quotations, margin maths, pipeline review and forecasting, AI in CRMs, and email law and data rules. It is a free course with a verifiable certificate of completion, and anyone can check a certificate on our verification page. Start the AI for Sales and Business Development course whenever you are ready.
Take the free course
Questions
What are good ChatGPT prompts for sales?
Prompts with five parts: a role, your real context (ICP, offer, proof, pasted source text), one task, a fixed format and a check such as 'use only facts given and list anything you assumed'. Useful ones cover ICP building, account research, first emails, discovery questions, CRM notes, objection role-play and pipeline review.
Can ChatGPT write cold emails that work?
It can draft them well if you supply verified facts about the account and a tight brief: under 80 words, one question, one clear ask. Check every fact before sending, send from your own business account and include a working opt-out, as email law requires.
Does CAN-SPAM apply to B2B emails?
Yes. The FTC's compliance guide says the law makes no exception for business-to-business email. Messages need accurate headers, honest subject lines, a valid physical postal address and an opt-out honoured within 10 business days, with penalties of up to $53,088 per violating email.
Is it safe to paste CRM data into ChatGPT?
Not into a public or personal account. Customer lists, contact details, prices and deal data are confidential and often personal data. Use your CRM's built-in AI or the enterprise AI tool your company approves, under its policy, and use anonymised data for practice.
Why does AI get discount maths wrong?
Language models predict text rather than calculate, and margin and markup are easy to confuse. On a $100 product costing $68, a 15% discount drops margin from 32% to 20%, not 17%. Do the numbers in a spreadsheet and use AI to explain them and draft the reply.
Is there a free course on AI for sales?
Yes. AI for Sales and Business Development on EDWartens is a free course with a verifiable certificate of completion, covering prompting for sales, account research, outreach, discovery, CRM updates, objection role-play, proposals, margin maths, forecasting and email and data rules.
Sources
- Salesforce: State of Sales report announcement (3 February 2026, read 11 October 2026)
- FTC: CAN-SPAM Act, A Compliance Guide for Business (read 11 October 2026)
- LinkedIn: User Agreement, section 8.2 (read 11 October 2026)
Written by the EDWartens engineering team for general education. Product names are trademarks of their owners; mentioning them does not imply endorsement. Prices and terms of other providers were checked on the date shown and can change.
1Z0-1122 Study Guide: OCI AI Foundations, Topic by Topic
ChatGPT for Procurement: Prompts, Maths and Limits
Generative AI in Banking: Use Cases, Limits, Rules
How to Use AI in Project Management, Stage by Stage
MQTT Sparkplug B Explained: Topics, Birth/Death and UNS (Video)
EU AI Act After the Digital Omnibus: New Deadlines for Industrial AI