ChatGPT for Sales: Setup, Prompts and Habits for Sales Reps

To use ChatGPT for sales, treat it like a new hire: work in your company approved account, load your seller context and approved files once, then run structured prompts for research, outreach, call prep, roleplay and post call admin. Verify every fact before anything reaches a buyer, because ChatGPT writes fluently whether it's right or wrong.

Company account vs personal account

If you already use ChatGPT at home, the biggest risk isn't a bad prompt. It's opening the wrong account. A personal login and a company workspace can look almost the same, but they're governed by different terms, settings and owners.

Paying for a personal plan doesn't change who owns the account. The full explanation is in Free vs Paid AI Tools: What's the Difference for Work?, and Is It Safe to Use AI at Work? covers what you can paste.

Set up ChatGPT for selling

AI is a brilliant new hire on day one. It knows the world, but nothing about your product, buyers or voice until you tell it. Setup is how you tell it once instead of every time.

1. Sign in to the company account

   and confirm which workspace you're in.

2. Write your seller context:

   about me, what we sell, ideal customer profile, personas, my voice and rules. Aim for one page. How to Build a Seller Context File has the template, and the glossary entry defines each section.

3. Create a reusable space for sales work

   ChatGPT offers ways to save instructions and files for reuse, such as projects or custom GPTs, depending on your plan. Paste your seller context into its instructions.

4. Upload approved reference files:

   a product one pager, battlecards, 2 or 3 approved case studies and 3 of your best emails labeled as voice examples.

5. Ask your admin about app connections

   (CRM, email, calendar, drive). If any are approved, start read only.

6. Run the setup test

   below and fix the context, not just the answer, every time it's wrong.

Project or custom GPT? In general terms, a project groups your own chats and files for ongoing work, which suits one rep's pipeline. A custom GPT is a configured assistant built for one repeatable job, which can suit a task the whole team runs, like building account briefs. Availability and sharing options depend on your plan, so check OpenAI's current documentation.

The prompts on this page use two fictional companies: Tallyline, which helps finance teams process vendor invoices without manual data entry, and Brookvale Supply, a distributor Tallyline sells to. Swap in your own.

Using only my saved instructions and the files in this workspace:
1. Describe what Tallyline does to a CFO at a 600 person distribution company in 3 sentences.
2. Name our top 3 competitors or alternatives and how we win against each. If my files list fewer than 3, say so. Don't fill the gap.
3. Write a first line for an email to a controller at a distributor that just opened a second warehouse.
Then list every fact you used that is not in my instructions or files.

That last line is the real test. Anything ChatGPT pulled from outside your files is a gap in your context or a guess.

From the course to ChatGPT

The Wicked Growth Club course teaches with Claude, but the skills transfer. Here's where each course step lives in ChatGPT, in general terms.

Every prompt below also works in any capable assistant. Only the setup steps change.

Workflow 1: Account research

A good brief answers four questions: why this account, why now, why this person and why us. When ChatGPT searches the web, ask for a link for every fact. A link in the output isn't proof until you open it. More in How to Use AI for Account Research.

Goal: a one page account brief on Brookvale Supply so I can decide whether and how to reach out.

Context: use my saved instructions for what Tallyline sells, our best fit customer profile and personas.

Research (search the web; give a source link for every fact):
1. Snapshot: what they do, size, HQ, recent growth
2. Triggers in the last 90 days, with dates
3. Likely pains in accounts payable, mapped to our top use cases, with your reasoning
4. Buying committee by title only: likely economic buyer, champion, users, blockers
5. Recommended angle and 3 talk tracks
6. Open questions I should verify or ask

Rules: mark anything you could not confirm as UNVERIFIED. Don't guess names of people. Several companies may share a similar name, so confirm you have the right one first.

Format: headings and short bullets, under 400 words, sources at the end.

Workflow 2: Outreach in your voice

Name the framework and point ChatGPT at your voice examples. Structure beats "make it compelling"; Prompting 101 explains the five parts. The full email method is in How to Write Cold Emails with AI That Don't Sound Like AI. Prospecting on LinkedIn instead? See How to Write LinkedIn Messages with AI and How to Use AI with LinkedIn Sales Navigator.

Write a first email to {name}, controller at Brookvale Supply, using Trigger, Problem, Proof, Ask.

Trigger (verified): {trigger and source link}.
Proof: [approved proof point]. Use nothing else as proof.

Rules: under 90 words. Match the style of the voice example emails in this workspace. Short sentences. Lead with their situation, not our product. One low effort question as the ask. Never use "I hope this finds you well" or "quick question."

Format: 3 subject lines under 5 words, then 2 versions: one direct, one curious.

Before you send, run the 60 second edit: verify every fact, cut a warm up first line, read it aloud, check there's one clear ask and ask yourself whether you'd reply. How to Tell If Something Was Written by AI lists the tells to cut.

Workflow 3: Call prep

A focused five minute card beats scrolling the CRM. Paste the verified brief, or reference it in the same project. For meetings further down the funnel, see How to Prep for a Sales Meeting in 10 Minutes with AI.

Using the verified Brookvale Supply brief below and my saved instructions, create a call prep card for {name}, controller:
1. Goal of the call (one line)
2. Why now: the trigger in one sentence
3. My hypothesis: the problem they likely have and why
4. Opener: 2 options, under 15 seconds each when spoken
5. 3 discovery questions that test the hypothesis
6. The 3 objections most likely from a controller, each with a short response
7. The ask: exactly what I'll propose
Keep it to one screen.

<account_brief>
{paste verified brief}
</account_brief>

Workflow 4: Roleplay with voice

AI gives you judgment free call practice whenever you want it. ChatGPT's voice conversation option makes it feel like a real phone call. Tell the AI buyer to keep replies short, the way a busy prospect would, and ask for the scorecard when you stop so you can read it, not just hear it.

You're roleplaying a prospect so I can practice cold calling. Stay in character until I say END.

Character: {name}, controller at Brookvale Supply, a 600 person distributor. Busy, mildly annoyed by cold calls. Hidden situation (reveal only if I ask good questions): month end close slips by several days because invoices are keyed in by hand.

Difficulty: hard. Raise at least 3 objections. If I pitch before asking a question, end the call.

Rules: answer in one or two short sentences, like a real phone call. Don't help me. Don't agree to a meeting unless I earn it.

When I say END, step out of character and score me 1 to 5 on opener, relevance, questions, objection handling, clarity of ask and talk ratio. Quote my strongest and weakest lines and give 3 specific improvements.

Practice the calls you fear most, and handle pushback with ACRA: acknowledge, clarify, respond, advance. See How to Handle Objections with AI and AI Sales Roleplay.

Workflow 5: Post call admin

The hour after a meeting is full of work AI can draft for you. First check two things: the recording followed consent rules, and you're in the company account. A transcript is yellow data.

Here is the transcript of my discovery call with Brookvale Supply, in the labeled section below. Produce:
1. Summary in 5 bullets
2. Pains in the buyer's exact words, with quotes
3. Numbers they mentioned
4. Stakeholders named and their role in the decision
5. Risks and red flags
6. Agreed next steps with owners and dates
7. CRM field updates, formatted to paste
8. A recap email to the buyer: under 150 words, their pains in their words, next steps, no new pitch
Use only what is in the transcript. Mark anything uncertain.

<transcript>
{paste transcript}
</transcript>

Update CRM fields yourself, or approve each change one at a time if your admin has approved a CRM connection. Walkthroughs: AI for Sales Call Notes, How to Write Follow Up Emails with AI for the recap, and the CRM guides for HubSpot and Salesforce.

Workflow 6: The verification pass

A confident, plausible, wrong answer is a hallucination. Web search and files reduce them but don't remove them. Run this before anything customer facing leaves your hands.

Review the draft below before I send it.
1. List every factual claim about Brookvale Supply or its people. For each, give the source you used, or write UNVERIFIED.
2. List every claim about Tallyline. Flag any that isn't in my instructions or files.
3. What in this draft are you least sure about, and why?
Don't rewrite the draft. Give me the list.

<draft>
{paste draft}
</draft>

Where ChatGPT is strongest and weakest for sales

Until those notes are in, test it yourself. Run Prompts 1, 2 and 7 on three accounts you know well, then score each output 1 to 5 on accuracy, specificity and usability with no edits. Long threads drift, so start a new chat per prospect and keep the context window for what matters.

When another assistant may suit your company better

The course is tool agnostic. ChatGPT is a sound choice when your company has approved it. Another assistant may fit better depending on how your company works.

This isn't a ranking. Capabilities change every few months. The skills on this page, from seller context to verification, work in all of them.

Frequently asked questions