Use ChatGPT for sales by giving it a short account brief and asking for a specific output: a call plan, a recap or practice responding to an objection. Start with a task whose answer you can check against your notes.

Choose a task for the stage of the deal

What you have Ask ChatGPT to produce What you still decide
An inquiry and a product description Questions for a discovery call Which unknown matters most
Notes from a completed call A recap with owners and next actions Whether the recap matches the conversation
A buyer’s stated objection A practice conversation What your offer can actually address
An approved proposal A plain-language explanation Scope, pricing and commercial commitments

Use your CRM for the account record and agreed actions. This exercise uses a regular chat with pasted text; it does not require a CRM integration or automated sending.

Prepare a discovery call from a short brief

The following account, product and notes are fictional. They give you a complete practice case without using customer records.

ACCOUNT: Cedar Desk, an office-furniture distributor.
INQUIRY: Maya, operations manager, requested a reporting demo.
HER NOTE: "We combine weekly order updates from four branches.
I want to see how your reporting view handles branch-level detail."
OUR PRODUCT: BranchView combines uploaded CSV order reports into
a branch comparison. This exercise assumes CSV upload only.
UNKNOWNS: Current reporting time, budget, decision process, purchase date.
CALL GOAL: Understand how the current report is made and what a useful
demo would need to show.

Open a new ChatGPT conversation and paste the brief with this request:

Prepare a discovery call using only this brief.
List confirmed facts separately from unanswered questions.
Write four questions in the order I should ask them.
For each question, explain what its answer would change in the demo.
Do not assume that Maya has budget, purchasing authority or a deadline.

A useful call plan should look like this. It is an example answer to compare with your result, not a recorded ChatGPT response.

Question How the answer helps
How do the four branch updates reach you today? Identifies the source files and handoff.
What do you change before combining them? Shows whether a demo needs to explain column cleanup.
Which branch-level question is hard to answer from the current report? Gives the demo a specific task to demonstrate.
Who else would need to evaluate a different reporting process? Identifies participants without assuming Maya is the buyer.

Read the questions aloud before the call. “How much revenue are you losing to inefficient reporting?” would be a poor addition because the brief establishes neither a revenue loss nor inefficiency. Ask about the current process first; a problem may emerge, or the buyer may simply want another way to view the data.

If ChatGPT produces ten generic questions, follow up: “Keep the four questions that could change this demo. Remove anything the brief already answers.”

Turn the call notes into a useful follow-up

Continue the fictional exercise with these notes:

CALL NOTES
Maya manually combines four CSV files each Friday.
The same order-status field has different names across branches.
Maya wants a demo focused on comparing open orders by branch.
Maya will send a sample with customer details removed.
Alex will check whether the sample columns work for the demo.
No deadline, budget or purchase decision was agreed.

Ask ChatGPT to draft an email of no more than 120 words, preserving who agreed to do what. Tell it to omit commercial terms and deadlines absent from the notes.

An example body:

Subject: Branch reporting demo: sample CSV

Hi Maya,

Thanks for explaining how you combine the four branch reports each Friday.
For the demo, we'll focus on comparing open orders by branch, including
the different names used for order status.

You'll send a sample with customer details removed. I'll check the
columns and confirm what we can show with that sample.

Have I captured the demo focus correctly?

Alex

The note about different column names belongs in the recap. A promise that the product automatically fixes them does not: the product brief only establishes CSV upload and comparison. If your draft says, “We’ll have you running by Friday,” ask:

Remove the Friday launch commitment. Friday describes their reporting
routine, not our delivery date. Replace it with Alex's actual next action:
check the sample columns and confirm the demo scope.

Compare the edited draft with both the call notes and product brief. Checking only the notes could miss an invented feature. For more message examples, use the AI for sales emails guide.

Practice an objection without inventing the answer

Suppose Maya says, “We already do this in a spreadsheet.” Don’t ask ChatGPT for a forceful rebuttal. Practice finding out whether there is a reason to change:

Play Maya using the account brief and call notes.
Begin with: "We already do this in a spreadsheet."
Wait for my response, then answer as Maya.
Do not invent new account facts. If I ask about something absent from
the notes, say that the exercise has not established it.
After two exchanges, identify any unsupported claim I made.

A reasonable opening from the salesperson is: “Which parts of the spreadsheet process work well, and where would you want the demo to show a difference?” If Maya reports no relevant difficulty or unmet need, a demo may not be useful. Practice recognizing that outcome as well.

Fix the output before it enters the account record

Use a small worksheet while reviewing:

Draft statement Basis Keep, change or ask?
Four branches provide CSV reports Call notes Keep
Maya controls the budget Not established Ask about the buying process
A sample is the next input Call notes Keep, with Maya as owner
Automatic column cleanup is included Not in product brief Remove and verify the feature separately

This is especially useful when the draft sounds polished. The error is often a small change in certainty: interest becomes intent to buy, a possible date becomes a commitment, or a requested feature becomes an available feature.

Use only the account context your team permits

For practice, use the fictional brief above. For real work, check your organization’s workspace policy and remove unnecessary personal information. OpenAI’s Data Controls FAQ explains how to turn off model-training use for chats through Data Controls. That setting does not decide which customer records your organization permits you to enter.

Move the reviewed recap into your approved sales system and check the recipient and attachments there. This workflow ends with a draft; it does not send messages.

Questions about ChatGPT for sales

Can ChatGPT qualify leads?

It can organize supplied evidence against criteria you provide. For example, it can show whether the notes establish a use case, a decision process or a next meeting. Keep missing fields as unknown; a complete-looking score is not evidence that the buyer fits.

What if I have almost no account information?

Ask for questions to learn the missing information. A short inquiry can support a call agenda, but it cannot support a detailed account strategy or claims about budget.

Should I start a new chat for each account?

For this manual workflow, separate chats make it easier to avoid mixing account notes. Label each brief clearly and recheck the account name before reusing a draft. A separate chat does not replace your organization’s data policy.

How can I tell whether the prompts are helping?

Compare a few drafts with your normal preparation: count factual corrections, check whether questions refer to the account and note the time you spend editing. Keep prompts that produce usable work without adding more review than they save.

Practice with the discovery-call preparation workflow when you have a real brief ready to review.