How to Write a Sales Follow-Up Email Sequence Using ChatGPT

What a Sales Follow-Up Email Sequence Does

A sales follow-up email sequence is a small set of pre-planned messages sent after a real conversation, each with a distinct job, spaced days apart. It is not cold outreach. The prospect already knows who you are — which is exactly why repeating your pitch reads as pressure rather than persistence.

The three stages get progressively smaller on purpose. As the deal goes quiet, the polite move is to take up less of the prospect’s attention, not more of it. That descent — 90 words, then 60, then 45 — is the part ChatGPT will not do unless you tell it to.

If you have not had a real conversation yet, this is the wrong page — that job belongs to a first cold outreach email, which follows different rules entirely. The sequence below assumes a call happened and notes exist.

Load Your Context Before You Prompt

Context is the whole game. A prompt without your call notes produces a brochure; the same prompt with four specifics produces something that reads like you wrote it on the train. Most sales reps skip this step and then blame the model.

You need exactly four things, and you can pull all of them off a call summary in under a minute:

• Their words. A phrase the prospect actually used about their problem.
• Their goal. What they said they were trying to achieve, in their framing.
• Your asset. One real resource, benchmark, or example you can hand over.
• The next step. The smallest thing you want them to do.

Paste your messy notes in raw. There is no need to tidy them first — bullet fragments from a post-meeting summary work better than polished sentences, because fragments carry the prospect’s own vocabulary. If your notes are thin, that is a signal to fix your discovery call questions rather than to ask the model to invent detail.

Deciding what to paste

Prospect notes are someone else’s business information sitting in your account, so treat them accordingly. The line that matters is not free versus paid — it is a consumer account versus a company-managed one. On Free, Go, Plus and Pro, training on your content is on by default with an opt-out available in settings; on Business and Enterprise workspaces, no training on business data is the default.

Swap real names for stand-ins like [Prospect] and [Company] before pasting, then put the real names back when you move the draft into your email client. It costs five seconds, and it means no sales follow-up email sequence you write ever puts a client’s name somewhere it does not belong. Our fuller guide on using ChatGPT safely at work covers the settings side, and you can check the current tier boundaries on OpenAI’s official plan comparison.

The Three Prompts, One Per Email

Each prompt below produces one email. Run them in order in the same chat so the model keeps your context, and change only the bracketed slots. The structure underneath is the Logic System — role, goal, context, format — which is worth understanding if you want to adapt these prompts to other tasks.

Start every sequence with this. It gives the model its role and its guardrails before you ask for a single word of copy.

You are an experienced B2B salesperson writing short follow-up emails after a discovery call. You write the way a busy person actually types: plain, specific, no marketing vocabulary.

Here is the situation.

Prospect: [role, e.g. Head of Operations] at [Company]
What they said their problem was, in their words: [paste the phrase they used]
What they said they wanted to achieve: [their goal]
The resource I can send them: [one specific thing you actually have]
The smallest next step I want: [e.g. a reply telling me whether this is still live]

Raw call notes:
[paste your bullet points exactly as they are, no tidying]

Rules for everything you write from here:
- Use only facts from these notes. If I have not given you a number, statistic, customer name, or result, do not invent one.
- Never write "I hope this email finds you well", "just checking in", "touching base", "reaching out", "circling back", or "wanted to see".
- No adjectives about our product. Describe what it does, not how good it is.
- Open with their situation, never with me or my company.

Confirm you have this, then wait. Do not write an email yet.

The first email hands over the asset and asks an open question. It is the longest of the three and still under ninety words.

Write email 1 of the sequence, to send 3 days after the call.

Its job: give them something useful and ask one open question.

Format:
- Subject line: 4 to 7 words, lowercase, no colon, references their situation not my product
- Paragraph 1: one sentence naming the problem they described, in their words
- Paragraph 2: two sentences maximum, offering [the resource] and saying in plain terms what is in it
- Paragraph 3: one open question about their situation. Not a meeting request.
- Hard limit: 90 words total, excluding the subject line and sign-off

Count the words before you answer. If it is over 90, cut and try again.

Email two exists to make replying trivially easy. One question, no new pitch, nothing attached. If the prospect can answer it in four words, you built it right.

Write email 2, to send 7 days after the call.

Its job: make replying take under ten seconds.

Format:
- Subject line: reply to the same thread, so write "same subject" and nothing else
- One short paragraph, maximum three sentences
- Reference the specific thing I sent in email 1, in four words or fewer
- End with a single closed question they can answer in one line, for example whether this is still a priority this quarter
- Do not re-explain the offer. Do not attach anything new.
- Hard limit: 60 words

Count the words. If it is over 60, cut it.

The break-up email is the one people get wrong, and the failure mode is always the same: it turns passive-aggressive. Written properly it reads as respect for their time, and in practice it is the email that most often gets a reply.

Write email 3, to send 14 days after the call.

Its job: close the file gracefully and remove all pressure.

Tone rules, these matter most:
- Assume they are busy, not rude. Never imply they owe me a reply.
- No guilt, no sarcasm, no "I guess this isn't a priority", no counting how many times I have written.
- State plainly that I am closing this for now and will not keep emailing.
- Leave one clean door open for them to reopen it later.

Format:
- Same thread, no new subject line
- Maximum three sentences
- Hard limit: 45 words

Once these work for you, save the whole set. Storing them in a ChatGPT Project keeps your buyer-persona context loaded so you skip the setup step on every future deal, and there are more sales-side templates in our free AI work templates library.

Using the context block above, write all three emails in one response, separated by a line of dashes.

Email 1 (day 3): give them the resource, ask one open question. Under 90 words.
Email 2 (day 7): one closed question, no new pitch. Under 60 words.
Email 3 (day 14): close the file graciously, no guilt. Under 45 words.

Apply every rule from the context block. Put the word count in brackets after each email so I can check them.

Why Constraints Beat Persuasion Instructions

Asking ChatGPT to “be persuasive” produces adjectives. Asking it to fit ninety words while opening on the prospect’s own phrasing produces something a human would actually send. Persuasion in a follow-up comes from specificity, and specificity comes from your notes plus a hard limit.

Notice what the second version does not contain: no percentage, no company name, no meeting request. Every number in the bad version was invented by the model, which is the default behaviour when you leave a gap where a fact should be — the same mechanism behind every other kind of AI hallucination.

The banned-phrase list in the context block is doing more work than it looks like it is. Telling a model what not to write is consistently more reliable than telling it what to write, a technique worth reading up on properly in our guide to negative constraints in prompting.

When ChatGPT Gets Your Emails Wrong

Four failures account for nearly everything that goes wrong here, and each has a specific fix rather than a general “try again”.

The word-count failure is the most common by far, and the reason is structural: models are poor at counting while generating. Give it a shape instead of a number and it complies.

What This Sequence Cannot Do

Three things stay with you, and outsourcing any of them is how AI-assisted selling goes badly.

Worth saying plainly: nobody can tell an AI-assisted email from a human one when it carries real context from a real conversation. What people can spot instantly is a generic email — and that is a context failure, not an AI failure. The same email written badly by hand would land just as flat. For a broader look at how experienced reps combine research with outreach, Zapier’s guide to using ChatGPT for sales is a solid, current read.

Which Sequence Fits Your Prospect

Three emails after a discovery call is the default, not the only shape. The prospect’s last known status decides which variant you run.

For the sixty-day reactivation, the Give email needs a genuine trigger — a new quarter, a regulation change, a hire they just made. Pull that from a quick prospect research pass before you prompt, because a reactivation email without a reason to exist is just a nudge in disguise.

Run this sales follow-up email sequence once this week on a deal that has already gone quiet. You will spend ten minutes setting up the context block and about two minutes per prospect after that — and the emails you have been avoiding will stop sitting on your to-do list.