How to Use ChatGPT for Sales Follow-Ups on Open Quotes

Coding Liquids tutorial cover featuring Sagnik Bhattacharya for How to Use ChatGPT for Sales Follow-Ups on Open Quotes.
Coding Liquids tutorial cover featuring Sagnik Bhattacharya for How to Use ChatGPT for Sales Follow-Ups on Open Quotes.

Use ChatGPT each week to follow up open quotes by reviewing a sanitised list, finding the next agreed action and drafting a short message for each stage. Check contact permission, quote details and tone yourself, then send by hand and record replies, pauses and next-action dates, so silence does not trigger repeated messages.

The Complete Excel Course with AI Integration

From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.

By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner

Buy now

The best reason to follow up is a useful unresolved decision, not the age of a quote. A buyer waiting for an internal approval needs a different message from someone who asked you to check back next month. Keep those distinctions in the export, and let any contact restriction override every drafting priority.

What should a weekly ChatGPT follow-up routine do?

A weekly ChatGPT follow-up routine should find which existing quotes have an action due, prepare accurate drafts and update the sales record after a person decides. Choose one weekly slot and one owner. The routine is not meant to build a new prospect list or to send anything automatically, and your mail system and CRM stay the places where sent messages and outcomes are recorded.

OpenAI has its own Sales plugin, which it describes as preparing account plans, meeting briefs, outreach, follow-ups and pipeline summaries from available context. It runs from a chat in ChatGPT Work or Codex, can connect apps such as Slack, Zoom and Gong (Salesforce and HubSpot have plugins of their own), and its availability depends on the plugin, your plan and workspace settings. You do not need it for the export-and-draft routine here; a cleaned table and clear instructions are enough, and connections can come later once permissions and review rules are settled.

The “nothing sent automatically” rule matches how OpenAI frames this work. One of the example tasks on its ChatGPT Work page is to find the emails you need to answer, draft replies and leave them unsent for review. Keep that last step whichever tool you use.

The broader sales reply workflow covers enquiries and quotes across the whole sales process. This routine is narrower: open records, a clear reason for contact, a reviewed draft and a recorded next action. That narrowness is what lets you notice when a stale record needs closing rather than another email.

What should you export to follow up an open quote?

A follow-up export needs one row per opportunity, with the details that show what the buyer is waiting for. Include a stable ID, quote reference and revision, amount, currency, scope summary, stage, date of the last real conversation, the buyer's last question, the agreed next step, the follow-up date, the owner and any contact restriction. Use internal customer IDs instead of names and insert the recipient yourself at review. A B2B software reseller's export, cleaned, looks like this:

IDQuote (rev)AmountStageLast real contactBuyer's last question or noteAgreed next stepRestriction
Q1QT-2207 (r1)$1,200Quote sent12 days ago“Does setup include importing our customer list?”We confirm scopeNone
Q2QT-2189 (r2)$2,400Internal review9 days agoBoard meeting this weekCheck back after the meetingNone
Q5QT-2150 (r1)$1,500Quote sent20 days ago“Contact us again next month”Follow up on the 1stNot before the 1st
Q10QT-2101 (r1)$1,000Quote sent30 days agoAsked not to be contactedNoneDo not contact

Keep the last automated activity apart from the last real conversation. An internal reminder or a regenerated PDF should not reset the buyer's waiting time. If the owner cannot say when the buyer last actually spoke to you, the record goes to a review queue rather than letting ChatGPT invent a recent conversation.

The buyer's last question is the most useful field. Q1's tells you the next message should settle a scope question from the approved quote, not ask whether they have decided. A follow-up that removes the obstacle and makes replying easy is far more welcome than a reminder that the quote exists.

OpenAI's sales use cases include rewriting a follow-up email and analysing an uploaded pipeline export. Its prompts leave the facts as placeholders for you to fill, which is the right instinct: prices, scope and commitments come from your own approved records, and the model should flag a missing term rather than fill it with stock sales language.

How do you sort open quotes before writing follow-ups?

Sort open quotes into queues by what each one is waiting for before you write any message. The main worked example uses twelve invented open quotes from the reseller, worth $24,000 in total. Four, worth $9,000, have an action due for review; three, worth $4,500, are waiting until a later date the buyer agreed; two, worth $5,000, are waiting on an internal answer; two, worth $2,500, are marked do-not-contact; and one $3,000 quote needs its validity checked before any commercial message.

QueueIDsCountValueOwner action
Due for reviewQ1–Q44$9,000Check evidence, draft a relevant message
Agreed later follow-upQ5–Q73$4,500Keep the agreed date
Internal answer missingQ8–Q92$5,000Resolve the scope or delivery question first
Do not contactQ10–Q112$2,500No draft, no message
Validity checkQ121$3,000Review the quote before stating terms

The counts add to twelve and the values to $24,000. Only four records reach the drafting queue this week, and only after contact and content checks. The other $15,000 has not gone anywhere; it needs a different action. Keeping the queues apart stops a model recommending twelve messages because every record still says Open.

The four candidates are Q1 at $1,200, Q2 at $2,400, Q3 at $1,800 and Q4 at $3,600. Q1 needs an inclusion clarified, Q2 has an agreed review point, Q3 has gone quiet after the quote and Q4 is an old enquiry whose relevance needs checking. Four different reasons, so four different messages; the amount alone should never decide who hears from you.

Classify this anonymised open-quotes table before drafting anything.
Respect do-not-contact flags, agreed later dates and unresolved internal questions.
Return one row per opportunity: source ID, queue, reason, next action.
Reconcile the record count and total value across queues.
Do not infer permission, quote validity, buying intent or probability of closing.
Do not draft or send messages, and do not update the source system.

A bad classification shows up quickly. Say the output ranks Q10 first because its value is attractive, despite the do-not-contact flag. Fix the rule and rerun before any drafting: a restriction is a gate, not a lower score. The same goes for an agreed later date, which is not overridden because the quote has been open longer than the others.

If sorting the export is the slow part, try The Complete Excel Course with AI Integration on Coding Liquids.

How do you check a follow-up is allowed before sending?

Check the relationship, the original enquiry, the buyer's stated preferences and the rules before any follow-up message is sent. Look at any agreed timing and any objection or pause. If you have customers in the EU, two sets of rules matter. The GDPR covers the personal data sitting in your pipeline export. And the ePrivacy Directive's Article 13 says email for direct marketing to individuals needs their prior consent, with a narrow exception for marketing your own similar products to existing customers whose details you got in the context of a sale, provided every message offers a free, easy way to object; it also bans hiding the sender or omitting a valid address for opting out.

Whether a follow-up on a quote the buyer asked for counts as direct marketing, and how the rules apply to business contacts, texts and calls, are exactly the questions to put to a qualified adviser for the places where your buyers are. Rules elsewhere differ in detail. Practically, record for each contact why you are entitled to write to them (they requested the quote, they are an existing customer, or something else), and process every opt-out the day it arrives.

Check restrictions outside the drafting model as well as in its instructions. The sender looks at the live customer record immediately before sending, because a reply or opt-out may have arrived after the export. If you cannot tell whether contact is appropriate, hold the draft and ask the responsible person; never test it by sending.

When a buyer writes “We have paused this project; please leave it with us”, record the pause and take the opportunity out of the weekly queue. Do not ask ChatGPT for a clever objection-handling reply. The next action may be an internal pipeline review, not another email, and the reason stays visible so a colleague does not restart the sequence next week.

Where the business gives regulated advice, as a financial adviser or mortgage broker does, the responsible professional reviews any follow-up before it goes. The message stays on the agreed service question and never introduces a personal recommendation, an approval promise or a predicted outcome.

Write Q1 and Q2 around the actual next decision

Q1's $1,200 quote covers setup, and the buyer asked whether importing their existing customer list is included. The approved scope says the import needs reviewing before it can be confirmed, so the draft answers that honestly and asks for what is needed. It must not imply the import is included because that would read more persuasively.

Hi [first name], you asked whether the customer-list import is included in the setup quote. We need to look at the source file and the fields you use before we can confirm that. Would a short preparation checklist for that review help? [your name]

One reason for contact, one next step. At review, confirm the quote reference and check that the checklist actually exists and is approved; if it does not, replace that line with an accurate request for the file. A helpful offer in a draft becomes a business commitment the moment it is sent.

Q2's $2,400 quote has an agreed review point after the buyer's board meeting. Refer to the agreement without pretending to know the result: “Hi [first name], we agreed I'd check back after your board looked at quote [quote reference]. Is there anything in the scope you'd like me to clarify before you decide next steps? [your name]”

Check that review point against the latest conversation before sending. If the buyer has since moved it, update the record and hold the message; a well-written draft is useless if its timing ignores what the buyer asked. An HR consultancy chasing a proposal that awaits a partner's sign-off works the same way: acknowledge the process and do not manufacture urgency.

What do you write when an open quote has gone quiet?

When a quote has gone quiet, ask one useful question using only what you know, and avoid a run of “just checking in” emails that add nothing. Q3 is an $1,800 quote with no reply since the expected response date. Silence does not say whether the buyer is busy, uninterested, confused or talking to a competitor.

Hi [first name], I'm following up on quote [quote reference] for the setup work we discussed. Is there anything in the scope you'd like clarified, or would you prefer me to close this for now? [your name]

Only offer to close if your team will act on the answer. If the buyer says close it, close the follow-up task and update the opportunity through your normal process. A polite “not now” is not a challenge for a more forceful message next week.

Q4 is a $3,600 enquiry that has gone stale. Confirm the basis for contacting them and whether the project is still relevant before drafting. A restrained draft asks whether the original project is still active and offers to update the record. It must not claim the price still stands, that capacity is reserved or that the buyer promised a decision, unless those facts are confirmed.

Here is a stale draft that needed fixing: “Your reserved implementation slot expires tomorrow.” No slot exists in the record, so the line goes and the false urgency is flagged. The better version: “Hi [first name], is the project in quote [quote reference] still under consideration? If plans have changed, let me know and I'll update our records. [your name]”

For longer sequences, the quote follow-up sequence tutorial goes into cadence. This routine responds to each record's evidence instead. A rule such as “owner review after two unanswered follow-ups” is fine, labelled as your own operating rule and applied after the contact checks; it is not a vendor limit or a legal cadence.

Make tone rules concrete enough to review

Draft only the approved candidates, one message per ID.
Use the buyer's last question and the approved quote terms.
One reason for contact, one simple next step, under 90 words.
No invented familiarity, discounts, deadlines, availability, outcomes
or testimonials. Use [first name], [quote reference] and [your name].
List missing or unsupported facts under REVIEW NEEDED. Do not send.

Each stage gets its own purpose, which is what keeps the messages from sounding like one template with the name swapped:

StagePurpose of the messageOpening that worksNever write
Question outstandingAnswer it or ask for what you need“You asked whether…”“Just checking in”
Agreed review pointHonour the agreement“We agreed I'd check back after…”“Have you decided yet?”
Silent after quoteOne clarifying question or an offer to close“Is there anything in the scope…”“I haven't heard back” three times
Stale enquiryCheck relevance, offer to update records“Is the project still under consideration?”Deadlines or reserved slots that do not exist
Revised quote issuedPoint to what changed“Revision 2 changes…”The old amount

Aim for drafts short enough to check in a minute; sixty to ninety words is a sensible internal target, and fewer when one question will do. Cut stock compliments and long recaps. For a before-and-after, replace “I hope this message finds you well and that you're excited to take your business to the next level” with “You asked whether the import is included in the setup scope.” The second sentence points to the real conversation, and that, not enthusiasm, is what makes a message read as human.

The sales follow-up automation tutorial is the next step once these review rules are stable. Automation does not fix weak source notes: if the export cannot tell a pause from a due reply, an automated sequence repeats the same mistake more often.

What should you check in a ChatGPT draft before sending?

Check the recipient, quote reference, revision, scope, currency, amount and validity against the current record before sending any ChatGPT draft. If Q2's quote has been revised from $2,400 to $2,650, an old draft must not repeat $2,400. Do not ask ChatGPT to work out which of several similarly named files is current; the owner names the approved revision and updates the table.

Watch for invented discounts. A draft that offers Q2 10% off to encourage a decision is giving away $240 and quoting $2,160. Unless someone with authority approved that, delete it. A correct calculation does not create commercial authority.

Open the attachment you plan to send and confirm it belongs to this buyer and matches the quoted revision. A draft can cite the right reference while the sender attaches an older file or another customer's proposal, and no review of the text catches that. Insert recipient details by hand and check that no placeholder survives.

Several contacts on one opportunity need one decision. If two people appear against the same quote, do not send two similar follow-ups because the export has two rows. The opportunity owner decides who receives the message and who is copied, and the conversation is logged once so counts and reminders stay coherent.

What should you record after a follow-up is sent?

Record the time, channel, quote revision, purpose, next-action date and owner of every follow-up after it is sent. When a reply arrives, update the state and any promised action. Keep Drafted, Approved and Sent as separate states: a finished email in a chat window proves neither that someone reviewed it nor that the buyer received it.

In a sample week, four approved messages go out and two buyers reply: one asks for a scope call and one asks to be contacted later. Two replies from four messages is 50%. There are no confirmed wins in that, so do not call the scope call a sale or credit the emails with the full $9,000.

Track corrections, pauses and complaints alongside replies. If one of four drafts needed a factual fix, record it; if many records need clarifying before a message is safe, the fix is better quote notes and clearer ownership, not faster drafting.

Pipeline value is not forecast revenue. The $24,000 export includes do-not-contact, paused and unchecked quotes, so never multiply it by a probability the model made up. Report queue counts and actual outcomes, and let the sales owner supply any forecasting assumptions separately.

How long should the weekly follow-up review take?

A small, tidy pipeline needs about thirty minutes a week to start: five to export and check restrictions, five to classify, ten to draft and review, ten to send and record. Complex scope questions need separate time, and your own first few weeks will show the real figure.

Finish with every record in exactly one next state: sent and awaiting reply, scheduled for an agreed date, awaiting an internal answer, held for a permission or validity check, do not contact, or closed. Each state has a reason and an owner. A record with no next state is next week's data problem, not an invitation to generate more follow-up text.

Keep the prompts and the stage table, and update them whenever a factual error or a buyer's reply shows a gap. The aim is a small routine that respects the conversation already under way: every message has a true reason, a checked commercial basis and a recorded next step, so no buyer gets the same generic chase twice.

If you want help setting up this weekly routine for your own pipeline, I offer one-to-one AI consultations for small businesses.

Further reads

Sources: OpenAI's Sales plugin page; OpenAI Academy, ChatGPT for sales; ChatGPT Work page; OpenAI help on file uploads; Directive 2002/58/EC (ePrivacy), Article 13, consolidated text. Checked 18 November 2026.

Rather not work this out alone?

An AI implementation consultation is a one-to-one call with me, not an agency or a sales team. You leave knowing what AI you can use in your business and where to start.

Book a 1:1 call with me