AI for Sales: ChatGPT Prompts for Sales Teams, With Output

Sales teams use AI to turn what they already know into drafts: research notes into an account brief, the brief into an email, call notes into a follow-up and CRM fields, and a deal list into a risk review. The rule that makes it work is simple. Paste the facts (about the prospect, your product and the call) and tell the AI to use only those facts. Then check every claim, price and promise before it reaches a customer.

• AI drafts from what you give it. Paste the source text, your approved product facts and your call notes.

• Never let it fill gaps with guesses. Invented results and features are the fastest way to lose trust.

• Check every promise. Prices, discounts, dates and next steps must match what you actually agreed.

• Rule-based prompts are easy to check. "Flag deals with no next step" can be verified in a minute.

• Results vary by tool. ChatGPT, Claude, Gemini and Copilot word things differently, but these prompts work in all of them.

A chat assistant writes the most likely text for the words in front of it. Give it your notes and it writes a clean recap. Give it nothing and it writes a convincing email about a product that doesn't exist.

For the general prompt method behind these examples, see the pillar guide AI prompts for work.

Each task has a prompt you can copy, the model's real answer (trimmed with "…") and what to check. The answers come from gpt-5-mini, a small OpenAI model. The seller, Cardinal (fleet maintenance software), the prospect, Ridgeview Foods, and all names and numbers are made up.

1. Build an account brief from sources you paste

Task: Write a one-page account brief on Ridgeview Foods before my first outreach.

Source text (copied from their website and a job posting):
"""
About us: Ridgeview Foods is a family-owned foodservice distributor serving restaurants, schools and hospitals across Ohio, Indiana and Kentucky since 1987. In 2026 we opened two new distribution centers, in Columbus and Louisville. Our fleet of 140 refrigerated trucks makes more than 2,000 deliveries a day.

Job posting: Fleet Maintenance Coordinator (Columbus, OH)
- Schedule preventive maintenance for 140 refrigerated trucks across three sites
- Track work orders and parts inventory; move our maintenance records off spreadsheets
- Work with drivers to make sure daily pre-trip inspections are completed and recorded
- Reduce unplanned downtime and roadside breakdowns
- Experience with Samsara telematics a plus
"""

Format:
1. What they do (2 sentences).
2. Likely priorities: 3 bullets. For each, quote the line from the source that suggests it, and label it "hypothesis".
3. Questions to confirm on a first call (3 bullets).
4. What we don't know (3 bullets).

Rules: Use only the source text. Don't add facts about the company, its people or its numbers.

…
2. Likely priorities: 3 bullets. …
- Hypothesis: Improve and professionalize maintenance recordkeeping and workflows.
  Quote: "move our maintenance records off spreadsheets"
- Hypothesis: Reduce unplanned downtime and roadside breakdowns through preventive maintenance and driver inspection compliance.
  Quote: "Reduce unplanned downtime and roadside breakdowns" and "Schedule preventive maintenance for 140 refrigerated trucks across three sites"

3. Questions to confirm on a first call (3 bullets).
- What are the current systems and formats used for maintenance records and where do you want them migrated (e.g., CMMS, cloud, database)?
…
4. What we don't know (3 bullets).
- The specific maintenance software, tools, or databases currently in use beyond "spreadsheets."
…

What to check: Every hypothesis quotes its source, so you can confirm each one in seconds. That is why the prompt asks for quotes. The first question uses jargon ("CMMS") the buyer may not use; rewrite it in their words. Don't ask an AI to research a company from memory. It may describe an older version of the company or invent details. If your tool has web search, open the sources it cites yourself.

2. Write a first-touch email

The usual prompt:

Write a cold email to a fleet manager at a food distribution company.

Subject: Cut refrigerated fleet costs and downtime by 15% — quick 15‑min chat?
…
In 15 minutes I can show how our platform combines real‑time telematics, trailer refrigeration monitoring, and route optimization to:
- Reduce fuel & idle time by up to 12–18%
…
If you’re open to a quick call, I’ll share a brief case study where we helped a similar food distributor cut refrigerated load spoilage by 20% and save $X/month on fuel and repairs.

Three statistics and a case study, none of them real. Now with the brief and approved facts:

Task: Write a first-touch email to the VP of Operations at Ridgeview Foods.

About them (from their website and a job posting): food distributor in Ohio, Indiana and Kentucky. 140 refrigerated trucks. Opened two new distribution centers in 2026. Hiring a Fleet Maintenance Coordinator to "move our maintenance records off spreadsheets" and "reduce unplanned downtime". Samsara experience is a plus in the posting.

About us (approved facts):
- Cardinal is fleet maintenance software. It schedules preventive maintenance by mileage or engine hours, tracks work orders and parts, and has a mobile app for drivers' pre-trip inspections.
- It pulls mileage automatically from Samsara and Geotab.
- Approved reference: Hollis Beverage (90 trucks) cut unplanned breakdowns by 31% in their first year.

Audience and tone: A busy operations executive. Plain, specific, no flattery, no hype.

Format: Subject line, then the email. Under 100 words. One ask: a 20-minute call.

Rules: Use only the facts above. Don't mention pricing. Don't claim we know their problems; say what we noticed.

Subject: 20 minutes to discuss maintenance records and uptime?

Hi [Name],

I noticed Ridgeview runs ~140 refrigerated trucks, opened two DCs in 2026, and is hiring a Fleet Maintenance Coordinator to move records off spreadsheets and reduce unplanned downtime (Samsara experience is a plus).

Cardinal schedules PMs by mileage/engine hours, tracks work orders/parts, has a driver pre‑trip app, and pulls mileage automatically from Samsara and Geotab. Hollis Beverage (90 trucks) cut unplanned breakdowns 31% in year one.

Can we schedule 20 minutes to see if this fits?

What to check: Every claim is from the approved list. The first paragraph, though, recites four facts about the prospect. One shows you did your homework. Four can feel like surveillance, so keep the job posting and cut the rest. Abbreviations ("DCs", "PMs") are fine for some buyers and not others. The guide to grounding explains why "use only the facts above" works so well.

3. Prepare discovery questions

Task: Prepare questions for a 30-minute discovery call with the VP of Operations at Ridgeview Foods.

What we know: 140 refrigerated trucks across three sites. Two new distribution centers opened in 2026. They are hiring a Fleet Maintenance Coordinator to move maintenance records off spreadsheets and reduce unplanned downtime. They may use Samsara.

What we need to learn: how they track maintenance today, what downtime costs them, who else is involved in a decision, their timeline, and how they would judge success.

Format: 8 open questions in the order I should ask them, grouped under short headings. After each question, one line on what to listen for.

Rules: No yes/no questions. Don't pitch our product in the questions.

Current maintenance process
1) Can you walk me through how a typical maintenance event is reported, scheduled, performed, and recorded across your three sites today?
Listen for: manual steps, paper/spreadsheet use, site-to-site variations, bottlenecks, and information gaps.
…
Unplanned downtime impact
3) How do you calculate or estimate the operational and financial impact when a truck is unexpectedly out of service, and what examples can you give from recent months?
Listen for: cost drivers (lost revenue, detention, driver hours, rerouting), frequency of events, concrete dollar/time estimates or anecdotes.
…
Decision process and stakeholders
6) Who else in the organization will influence or sign off on changing maintenance processes or selecting a new maintenance/telematics solution, and how do they typically evaluate vendors?
…

What to check: The order and the "listen for" lines are good. Many questions ask two or three things at once, though, and a buyer will answer only the last part. Split them. Eight questions won't fit in 30 minutes with real answers, so pick five.

4. Write the follow-up after a call

Task: Write a follow-up email to send after today's discovery call.

My call notes:
"""
Call w/ Ridgeview Foods, Tue. Angela Brooks (VP Ops), Marcus Lee (fleet manager).
- maintenance tracked in Excel per site, each site does it differently
- 6 roadside breakdowns last quarter, Marcus says each one ~$3k (tow + spoiled product + overtime)
- they use Samsara on all trucks
- Angela wants one system before the new Louisville site is fully staffed (Jan)
- IT (Priya) has to review security, they'll send their questionnaire
- asked about price for 140 trucks. I said I'd send pricing.
- asked if we can waive onboarding fee. I said I'd check with Marta (my director). NO promise.
- next: demo for Angela + Marcus + Priya, they'll propose times
"""

Format: Subject line. Thank them in one sentence. Recap what we heard (3 bullets). Next steps with owners (bullets). Under 150 words.

Rules: Use only my notes. Don't promise a discount or anything I didn't commit to. Don't include the breakdown cost estimate in the email.

Subject: Thanks — follow-up and next steps from today’s call

Thanks for your time today.

Recap of what we heard:
- Maintenance is tracked in Excel at each site with different processes.
- Six roadside breakdowns last quarter; you currently use Samsara on all trucks.
- Angela wants one system in place before the new Louisville site is fully staffed (target: Jan); IT (Priya) will review security.

Next steps:
- I will send pricing for 140 trucks (ACTION: me).
- I will check with my director Marta about the onboarding fee (ACTION: me; no commitment).
- Please propose demo times for Angela, Marcus, and Priya so we can schedule (ACTION: Ridgeview / Angela).
- Priya to send the security questionnaire for IT review (ACTION: Priya).

What to check: The commitments are right. It sends pricing, checks on the fee, and promises no discount. The cost estimate stayed out, as told. But "(ACTION: me; no commitment)" reads like an internal note, not an email to a VP. Rewrite those lines. Writing about Angela in the third person to Angela is odd too. Add a greeting and name the people you're writing to.

5. Turn call notes into CRM fields

Task: Turn my call notes into CRM fields.

Fields: Company | Contacts and roles | Current process | Pain | Impact | Decision process | Timeline | Budget | Next step | Risks

Rules: Use only my notes. If a field isn't covered, write "not stated". Keep each field under 25 words.

My call notes:
"""
[the same call notes as above]
"""

Company: Ridgeview Foods
…
Impact: Marcus estimates ~ $3k cost per breakdown (tow + spoiled product + overtime).

Decision process: IT to review security questionnaire; team (Angela, Marcus, Priya) will attend demo and decide.

Timeline: Wants one system before new Louisville site fully staffed (January).

Budget: not stated
…
Risks: Onboarding fee waiver undecided; IT security review may delay decision.

What to check: "Budget: not stated" is exactly right. That's why the rule is there. But "will attend demo and decide" is a guess. Nobody said those three people decide. Wrong CRM fields spread into forecasts, so fix guesses before you save. A fixed list of fields works the same way as a form. The guide to structured outputs shows more ways to get consistent fields.

6. Prepare for objections

Task: Help me prepare for two objections I expect from a prospect.

Objection 1: "We track maintenance in spreadsheets and it works fine."
Objection 2: "This is too expensive for us right now."

About us (approved facts only):
- Cardinal schedules preventive maintenance by mileage or engine hours, tracks work orders and parts, and has a mobile app for drivers' pre-trip inspections.
- It pulls mileage automatically from Samsara and Geotab.
- Price: $6 per vehicle per month, annual contract, plus a one-time $2,500 onboarding fee. Discounts need my director's approval.
- Approved reference: Hollis Beverage (90 trucks) cut unplanned breakdowns by 31% in their first year.

Format: For each objection: (1) a question I can ask to understand it, (2) a short spoken response, under 60 words, (3) what not to say.

Rules: Use only the facts above. Don't offer a discount. Don't argue with the prospect.

…
Objection 2: "This is too expensive for us right now."

1) Question to understand
- Can you tell me what budget you’ve allocated for maintenance tools and what trade-offs you’d have to make to free up funds?

2) Short spoken response (under 60 words)
- I understand. Cardinal is $6 per vehicle per month with a $2,500 one-time onboarding fee. Many customers find the automation and centralized tracking reduce unplanned breakdowns — Hollis Beverage cut unplanned breakdowns by 31% in year one — which can offset ongoing costs over time.

3) What not to say
- Don’t offer a discount. Don’t argue that their budget is wrong or that they “can’t afford” not to buy.

What to check: The price is right and there's no discount. But "Many customers find" is not in the facts: you have one approved reference, not many. Lines like that slip into a call easily and are hard to take back. Cut it. The opening question is also pushy ("trade-offs you'd have to make"). Ask what they compare the price with instead.

7. Review your pipeline against clear rules

Task: Review my open deals and flag the ones at risk.

Today is November 3.

Rules for "at risk" (use only these):
- The close date has passed, or
- There is no next step, or
- The last activity was more than 21 days ago.

Deals:
"""
Deal,Stage,Amount,Close date,Last activity,Next step
Ridgeview Foods,Discovery,10080,Dec 15,Oct 28,Demo with VP Ops and IT
Hollis Beverage expansion,Proposal,4320,Oct 31,Oct 20,Send revised quote
Lakeshore Transit,Negotiation,21600,Nov 30,Oct 2,Legal review of contract
Brightway Movers,Demo,3600,Dec 1,Oct 30,
Summit Rentals,Proposal,7200,Nov 20,Oct 29,Pricing call Nov 6
Greenfield Waste,Discovery,18000,Jan 15,Sep 30,Follow up after budget
"""

Format: A table: Deal | At risk? | Which rule | Suggested next action (one line). Then the total amount at risk, with the numbers you added.

Rules: Don't invent information about the deals.

Deal | At risk? | Which rule | Suggested next action (one line)
--- | ---: | --- | ---
Ridgeview Foods | No | — | Proceed with scheduled demo with VP Ops and IT.
Hollis Beverage expansion | Yes | Close date has passed (Oct 31) | Send the revised quote and immediately confirm an updated close date and next steps.
Lakeshore Transit | Yes | Last activity >21 days ago (last activity Oct 2) | Check status of the legal review, get a timeline from legal, and confirm outstanding items to move to close.
Brightway Movers | Yes | No next step | Schedule a follow-up (call or meeting) to capture feedback from the demo and agree next steps and timeline.
Summit Rentals | No | — | Prepare for the pricing call on Nov 6 and capture agreed adjustments to the proposal.
Greenfield Waste | Yes | Last activity >21 days ago (last activity Sep 30) | Reach out to confirm the customer's budget timing and schedule the follow-up meeting.

Total amount at risk: 47,520 (Hollis 4,320 + Lakeshore 21,600 + Brightway 3,600 + Greenfield 18,000)

What to check: All six calls and the total are right; we checked them by hand. Clear rules make this easy to verify: the same filter in a spreadsheet gives the same answer. Always give the date. A model doesn't reliably know what day it is. On a long deal list, check the flagged rows and the total before you bring them to a forecast call.

Before I send this:
[ ] Every product claim is on our approved facts list
[ ] Every number (price, %, customer result) matches its source
[ ] No discount, date or promise that I haven't agreed to
[ ] Personal details about the prospect are relevant, not creepy
[ ] Names, titles and company details are spelled right
[ ] It sounds like me, not like a template
[ ] Nothing from another customer's deal leaked in

• Follow your company's data rules. Use the AI tool your company approved, with your work account. Some CRM and email tools have AI built in that is already covered by your company's agreements. Ask.

• Don't paste: CRM exports, contact lists, call recordings or transcripts, contracts, deal terms under negotiation, or documents a prospect shared under confidentiality, such as a security questionnaire, unless your approved tool and your company's rules allow it.

• Use the minimum. "VP of Operations at a food distributor with 140 trucks" is usually enough. The AI rarely needs a personal email or phone number.

• Claims and prices come from approved sources only. Discounts follow your normal approval process, whatever the AI suggests.

• Some uses of AI are regulated in some places. Ask your legal and compliance team what applies to you before you roll this out.

For what each vendor says it keeps, see is it safe to use ChatGPT at work?. For why models invent statistics, see hallucination basics.

When reps run the same prompts after every call, the next step is usually a shared prompt library, then the AI features in your CRM. Some teams want more: software that updates the CRM after each call, drafts the follow-up for the rep to approve, or flags stale deals every morning. Software that takes steps in your systems is an AI agent. It needs limits and approvals, and it should never send customer messages or change prices without a person. Read what is an AI agent? and the overview of AI agents for business leaders before you buy or build one.

Can AI write cold emails that get replies?

AI can write a clear, short first draft in seconds. Replies come from relevance: a real observation about the prospect, a real result from a similar customer, and one easy ask. Give the AI those facts and a word limit, then edit the draft so it sounds like you.

What are the best ChatGPT prompts for sales?

The best sales prompts paste the facts (prospect research, approved product facts, call notes), name the reader, set a format and a word limit, and add "use only the facts above". Good starting tasks are account briefs, first-touch emails, discovery questions, follow-ups, CRM notes, objection prep and pipeline reviews.

What is prompt engineering for sales?

Prompt engineering for sales means writing instructions that get a usable sales draft from an AI tool: the task, the context, the audience, the format and the rules. It isn't technical. Most of the skill is knowing which facts to paste. See what is prompt engineering? for the basics.

Can I paste CRM data or call transcripts into ChatGPT?

Only if your company has approved that tool for customer data and you use your work account. Otherwise, summarize the notes yourself and remove personal details. Many CRMs now include AI features under your company's existing agreement, which may be the safer route. Ask IT.

Can AI practice sales calls with me?

Yes. Ask it to play a skeptical buyer, give it the buyer's role and concerns, and ask for one question at a time. It's useful for practicing objections. Remember it plays a generic buyer, not your real prospect. The guide to role prompting shows how to set it up.

Should AI send sales emails automatically?

Not without a person approving them. Automatic sending multiplies any mistake, from a wrong price to a wrong name. If you want automation, keep a person in the approval step and start with low-risk messages.

• Learn the method behind these prompts: AI prompts for work

• Catch wrong facts before you send: how to check AI answers

• Know what you can paste: is it safe to use ChatGPT at work?

• Write clearer instructions: clear asks, roles, constraints and format

• The other role playbooks: AI for marketing, AI for HR and AI for finance