All articles
playbook

20 ChatGPT and Claude Prompts for Sales Follow-Up, Collateral, and Pipeline Tracking

Twenty AI prompts for everything after the call: recaps that get replies, mutual action plans, proposals that survive procurement, battle cards, CRM hygiene, and forecast review. Each written to work from your notes rather than the model's memory.

August 23, 202610 min read
20 ChatGPT and Claude Prompts for Sales Follow-Up, Collateral, and Pipeline Tracking

The call went well. Then twenty-six hours pass before the follow-up goes out, the recap omits the one commitment the buyer actually made, the CRM note says "good call, sending pricing," and three weeks later nobody can reconstruct why the deal stalled.

Everything after the meeting is the least glamorous part of selling and the part where most pipeline quietly dies. It is also, unlike the meeting itself, almost entirely writing and structuring — which is what language models are genuinely good at.

Below are twenty prompts for ChatGPT or Claude covering the work that happens after the call: follow-up and recaps, collateral and proposals, internal enablement assets, CRM hygiene, and pipeline and forecast review. As with any AI-generated artifact that touches a customer, the rule is that you own what you send.

Two rules before you use any of these

Never send a recap the model wrote from memory. Every recap prompt below takes your notes as input. A model asked to "write a follow-up for my call with Acme" will invent commitments that were never made, and a buyer who reads a recap containing something they did not say will trust you less for the rest of the cycle.

Keep customer data where it belongs. Before pasting call notes, contract terms, or anything with names and numbers into a consumer chat tool, know your company's policy. Enterprise tiers with no-training guarantees exist for exactly this reason. This is a real constraint, not a formality.

Follow-up and recaps (prompts 1–5)

1. The recap that actually gets replied to

Here are my raw notes from a call with [NAME], [TITLE] at [COMPANY]:

<notes>
[PASTE]
</notes>

Write a follow-up email that:
- Opens with the problem they described, in their words
- Lists what we agreed, separated into "you said" and "I said"
- States the next step with a specific date and owner
- Ends with one question that requires a reply

Under 150 words. No "just circling back," no "I wanted to reach out,"
no adjectives about how great the conversation was.

If my notes do not contain a clear next step, say so instead of
inventing one.

The last instruction is the one that matters. A fabricated next step is worse than no next step, because it forces the buyer to correct you.

2. The multi-threaded follow-up

Call notes attached. Three people were on the call: [NAME 1, TITLE],
[NAME 2, TITLE], [NAME 3, TITLE].

<notes>
[PASTE]
</notes>

Write three short follow-ups — one per person — each addressing what
that specific person raised and what they are measured on. Do not repeat
the same paragraph three times.

Flag anyone who said nothing on the call, and suggest how to draw them
out without singling them out.

3. The mutual action plan

Deal: [PRODUCT] into [COMPANY]. Target go-live: [DATE]. Known steps:
[SECURITY REVIEW / LEGAL / PILOT / PROCUREMENT / etc].

Build a mutual action plan as a table, working backwards from go-live:
step, owner (us or them, named), date, and what unblocks the next step.

Mark every step that depends on someone who has not been on a call yet.
Then tell me which step is most likely to slip and why.

The value is the second half. Every stalled deal has a step nobody owned, and the plan makes that visible while there is still time.

4. Re-engaging a deal that went quiet

Last contact with [NAME] at [COMPANY] was [DATE]. Here is the thread:

<thread>
[PASTE]
</thread>

Since then: [ANY RELEVANT NEWS OR PRODUCT CHANGE].

Write three different re-engagement emails:
1. One that gives them something useful and asks for nothing
2. One that asks a direct question about where things stand
3. One that closes the loop honestly and offers to stop following up

Tell me which to send given the thread, and why. Do not use fake
urgency or pretend we had a meeting that did not happen.

The third option is not a gimmick. A clean close-the-loop email gets replies precisely because it is the only honest one in the inbox.

5. Handling a "send me some information" brush-off

[NAME] ended our call with "send me some information and I'll take a
look." Here is what we discussed:

<notes>
[PASTE]
</notes>

Write a response that:
- Sends exactly one thing, chosen to match what they actually cared about
- Explains in one line why I chose that one thing
- Asks for a specific, small next step

Then tell me what "send me information" most likely means here, and the
question I should have asked before the call ended.

Collateral and proposals (prompts 6–10)

6. The one-pager built from their language

Here is what [COMPANY] told me about their situation:

<discovery>
[PASTE NOTES]
</discovery>

And here is what [PRODUCT] does:

<product>
[PASTE CAPABILITIES]
</product>

Write a one-page document my champion can forward internally. Structure:
their problem, cost of doing nothing, what changes, what it takes to get
there, what it costs.

Use their vocabulary throughout. Do not use any capability from the
product list that does not connect to something in the discovery notes.

7. The proposal that survives procurement

Draft a proposal for [COMPANY]: [PRODUCT], [SCOPE], [PRICE STRUCTURE],
[TERM].

Sections: scope, what is explicitly out of scope, success criteria,
implementation timeline with owners, commercial terms, assumptions.

The out-of-scope and assumptions sections should be the most specific
parts of the document. Write them as if the goal is to prevent a dispute
in month four.

Most proposals are optimistic documents that become liabilities. Asking for precision in the boring sections is how you avoid that.

8. The security and compliance response

[COMPANY] sent these security questions:

<questions>
[PASTE]
</questions>

Here are the facts about how [PRODUCT] handles data:

<facts>
[PASTE YOUR ACTUAL SECURITY DOCUMENTATION]
</facts>

Draft answers using only the facts provided. Where the facts do not
cover a question, write "NEEDS INPUT FROM [TEAM]" instead of an answer.

Do not soften a limitation into a strength.

The refusal to fill gaps is the entire safety mechanism here. A confidently wrong security answer is a problem that surfaces during legal review, at the worst possible moment.

9. The business case for the economic buyer

Champion: [TITLE]. Economic buyer: [TITLE]. Deal size: [AMOUNT].

What the buyer's organization is trying to achieve: [FROM DISCOVERY].

Write a business case in the economic buyer's language:
- The problem in financial terms
- Three scenarios: do nothing, build internally, buy
- What we are asking for and what they get
- Risk and how it is mitigated

Every number must be a placeholder for me to fill in with figures the
customer supplied. Do not generate benchmark statistics or ROI
percentages.

10. Turning a won deal into a reference asset

We closed [COMPANY]. Here is what happened, from first call to go-live:

<history>
[PASTE]
</history>

Draft:
- A three-sentence internal win summary for the team
- A customer-facing case study outline with the questions I would need
  to ask them to fill it in
- Two sentences I could use in a cold email to a similar company

Mark anything I would need the customer's approval to say publicly.

Internal enablement (prompts 11–14)

11. The battle card

Competitor: [NAME]. Here is what I know from deals where we met them:

<field_notes>
[PASTE]
</field_notes>

Build a battle card:
- Where we win and the evidence
- Where they win and how to handle it honestly
- Three questions that expose the difference without naming them
- Language to avoid because it sounds defensive

Use only my field notes. Do not add public claims about the competitor.

12. The objection library

Here are objections our team logged this quarter:

<objections>
[PASTE]
</objections>

Cluster them into themes. For each theme:
- The underlying concern
- The clarifying question to ask first
- A response that works
- A common response that backfires

Rank themes by how often they appear and how often the deal was lost
after them.

13. Onboarding a new rep

Our sales motion: [DESCRIBE]. ICP: [DESCRIBE]. Product: [DESCRIBE].
Sales cycle: [LENGTH]. Common failure mode for new reps: [DESCRIBE].

Write a 30-day ramp plan with a specific, checkable output for each week.

Week 4 should end with something a manager can evaluate objectively, not
"shadow calls" or "get familiar with the product."

14. Turning call recordings into coaching

Here is a transcript excerpt from one of my calls:

<transcript>
[PASTE]
</transcript>

Analyze:
- My talk-to-listen ratio, roughly
- Questions I asked that were actually closed-ended
- Moments where the prospect signaled something and I moved on
- The single highest-leverage thing to change next call

Be specific and quote me. Do not be encouraging.

CRM hygiene and tracking (prompts 15–17)

15. Notes to structured CRM fields

Raw notes:

<notes>
[PASTE]
</notes>

Convert to CRM fields:
- Next step (with date and owner)
- Identified pain, in the buyer's words
- Metrics they named
- Decision process and timeline
- Buying committee members mentioned
- Risks
- Confidence in this deal (and why)

Leave any field the notes do not support blank. Do not infer.

The blank fields are the useful output. An empty "decision process" field is an accurate picture of what you do not yet know.

16. Auditing your own pipeline

Here is my pipeline export:

<pipeline>
[PASTE: opportunity name, stage, amount, close date, last activity date,
next step]
</pipeline>

Flag:
- Deals with a close date but no next step
- Deals where last activity is older than half the sales cycle
- Deals whose stage does not match the described next step
- Close dates that cluster suspiciously at quarter end

Rank by how much of my forecast is at risk. Ask me the one question you
would need answered to assess each flagged deal.

17. Preparing for a forecast review

My committed number this quarter: [AMOUNT], from these deals:

<deals>
[PASTE: name, amount, stage, next step, known risks]
</deals>

For each deal, tell me:
- What my manager will ask
- The weakest part of my case
- What evidence would make it credible
- Whether it belongs in commit, best case, or pipeline

Be the skeptical manager, not the supportive one.

Territory and prioritization (prompts 18–20)

18. Deciding where to spend the week

Here are my open opportunities:

<opportunities>
[PASTE: company, size, stage, days in stage, engagement level,
known champion yes/no]
</opportunities>

I have roughly [N] hours of selling time this week.

Rank these by expected return on my time, not by deal size. Explain the
ranking. Tell me which deals I should stop working entirely, and how to
close them out cleanly.

Most reps under-invest in the disqualification decision. Asking for it explicitly makes it a decision rather than a drift.

19. Building a target account list

Our best customers share these traits:

<customers>
[PASTE: industry, size, tech stack, trigger event, use case]
</customers>

Derive an ICP definition from these, then give me:
- The observable signals that would identify a similar company
- Where each signal can actually be found
- Three disqualifying signals

Do not name specific companies. I want criteria I can apply to a list,
not a list you generated.

The final constraint prevents the most common failure mode — a confident list of prospect companies, some of which do not exist or were acquired two years ago.

20. The weekly self-review

Here is my week:

<week>
[PASTE: calls held, emails sent, deals advanced, deals stalled,
what I avoided]
</week>

Tell me:
- What actually moved and what only felt productive
- The pattern in what I avoided
- One thing to do differently next week, stated as a specific behavior
  I can check on Friday

One thing. Not a list.

What none of this replaces

Every prompt on this page operates on something that already happened. They take your notes, your transcripts, your pipeline export, and make the work around the conversation faster and more disciplined.

None of them help while a prospect is actually talking. The moment a buyer raises an objection you have not rehearsed, mentions a competitor mid-demo, or asks a pricing question that decides the deal, the follow-up workflow is irrelevant — that gets resolved live or not at all.

MagicScreen sits in that gap: real-time coaching, objection handling, and battle cards on your Mac during Zoom, Google Meet, and Microsoft Teams calls, without joining as a bot or recording. Use the prompts above for everything around the call, and have something with you inside it.

Download MagicScreen for Mac — free to start.

Try it free

See MagicScreen in action on your next call.

Real-time intelligence. No bot. No recording. Just you, your prospect, and the right words at the right moment.