Two domains serve the same public Notion page titled "Implementing Upselling Strategies", and one vendor serves one article on two domains. That is page one: a product list reorganised into prompts by people who never carried an expansion number.
Hand a model your product catalogue and ask where the expansion opportunity is, and it finds one in every account: a catalogue is a list of things to sell, and nothing in it argues against selling them. So every prompt here starts from how the customer already uses what they bought. The strategy belongs to the account expansion playbook and the wider post-sale motions to AI prompts for customer success managers. This page owns the evidence and the conversation.
Expansion starts with evidence of use, not a product list
The first prompt deliberately withholds the catalogue, because handing it over is how you end up pitching a module nobody needs.
Here is how this customer uses what they already bought: adoption by feature, usage trend, who logs in, support tickets, and what they told me in the last two reviews. Do not suggest products. Tell me only what they are doing more of, what they have abandoned, and where they have built a workaround.
Workarounds are the finding: a customer exporting to a spreadsheet every Monday has told you about a gap without filing a request. Usage patterns sit in Account Management & Customer Growth.
Prompts that find the opportunity in what they already own
Only now does the catalogue come in, second, so the evidence anchors it rather than the reverse.
Here is the usage picture you just produced and here is our product list. For each thing you suggest, name the specific observed behaviour that justifies it, and rank by strength of evidence. Where the evidence is thin, say so rather than finding a reason.
Now tell me which items on that list solve a problem the customer has not raised with me, and which simply sell them more of what they already have. Label each one.
The second split is the honest one. More seats for a growing headcount is expansion; a module fixing something they never raised is a pitch waiting to be ignored. Framing the fit is in Solution Framing, Demo & Proposal.
Separating what the customer needs from what you are measured on
Your compensation plan is not evidence. Writing the conflict down is the cheapest way to stop it steering the call.
Here is what I am compensated to sell this quarter and here is the evidence-ranked list for this account. Mark every place where the two agree and every place where my incentive points somewhere the evidence does not support. Be blunt.
Run it and you usually find one item sitting high for no reason beyond being the quarter's push. Knowing that beforehand beats any script.
Prompts for the timing question, which kills more expansion than price
Most expansion deals die on timing, not fit. The customer agrees it is a good idea, then does nothing for nine months, because you raised it in the wrong week.
Given this account's budget cycle, renewal date, recent change in their business, and the state of their current rollout, tell me whether this expansion conversation should happen now, at renewal, or later. Argue the case against raising it now first.
If the answer is later, name the trigger or date that should bring it back and the one question I should ask in the meantime to establish it.
Arguing against first is a correction, not politeness: ask a model whether to raise something and the default answer is yes. Where this meets the renewal clock, ChatGPT prompts for renewal conversations owns the renewal side, and Post-Sale & Renewal Micro-Work holds the smaller tasks around it.
Take one prompt free, no account needed. Pro is $12 per month or $89 per year, and lifetime access is $149 — See pricing.
Raising it: prompts for a conversation that is not a pitch
The difference between an expansion conversation and a pitch is who talks for the first five minutes.
Write the two questions that open this expansion conversation. They must be about what the customer is trying to do next, must be answerable without reference to our product, and must not set up a pitch. Then write the one sentence I say after their answer that connects it to what I observed in their usage.
Here is what they said. Tell me whether the opportunity I assumed is the one they described, and where my assumption was wrong.
The second prompt is the one most reps skip, and it is where the expansion becomes real or quietly should not happen. Question design sits in the QBR playbook, which is often where this conversation actually lands.
Prompts for the internal case, because expansion needs a new budget holder
Expansion usually means money outside the current budget line, so somebody your contact does not control has to agree.
My contact is in this role and the budget for this expansion sits with this other person, who cares about these things. Write the half-page my contact can forward: the problem in their words, what changes, what it costs, and what happens if they wait. No product names in the first paragraph.
Withholding the product name forces the document to lead on the problem, the only version a budget holder reads twice. The longer form is how to build a business case your buyer can defend, and multi-stakeholder sequencing is in Deal Strategy & Stakeholder Management.
When the honest answer is that they should not buy more
Sometimes the account is not ready: the first rollout is half-finished, the champion has left, adoption is falling. Expanding into that is how a renewal becomes a churn.
Make the case that this account should not expand right now. Use only the evidence I have given you. Then tell me what would need to be true in ninety days for the answer to change.
Run it on every account you are about to expand. It costs two minutes and occasionally saves a renewal worth more than the expansion.
Why a model given your catalogue will find an opportunity in every account
This is the failure mode to design around. A catalogue contains no evidence against selling anything in it, so a model asked to match an account to one always matches. Give it twenty accounts and it returns twenty opportunities, plausibly ranked, with reasoning that reads well and rests on nothing.
It also cannot see what decides expansion: their headcount freeze, procurement's new approval threshold, a champion interviewing elsewhere, a disputed invoice. None of that is in the usage report.
So the order matters more than the wording. Evidence first, catalogue second, every suggestion tied to an observed behaviour, the case against argued first, and "the evidence is thin" allowed as an answer. Reverse it and you have built a machine for pitching.
The rest of the catalogue, by role and by stage, is mapped in AI prompts for sales. Browse the library → or read the full taxonomy.
Frequently asked questions
What are good ChatGPT prompts for upselling existing customers?
The ones that withhold your product list until step two. Start with a usage prompt naming only what the customer does more of, what they abandoned and where they built a workaround. Then bring in the catalogue, requiring an observed behaviour behind every suggestion. Then argue against raising it now.
How do I find cross-sell opportunities with AI?
From evidence of use rather than from a product match. Adoption by feature, usage trend, who logs in, tickets and what they said in the last two reviews. The strongest finding is usually a workaround: a customer exporting to a spreadsheet every Monday has described a gap without filing a request.
When should I raise an expansion conversation?
Judge it against their budget cycle, renewal date, recent business change and the state of the current rollout, and make the model argue against raising it now first. Most expansion deals die on timing, not fit. If the answer is later, fix the trigger that brings it back.
How do I upsell without damaging the relationship?
Open with two questions about what the customer is trying to do next, answerable without reference to your product, then check whether the opportunity you assumed is the one they described. Label every suggestion as either solving a problem they raised or selling more of what they already own. The second kind is a pitch.
What data does AI need to spot an expansion opportunity?
Adoption by feature, usage trend, login patterns, support tickets and your notes from the last two reviews — plus your compensation plan, so the conflict between evidence and incentive is written down rather than acting on you quietly. What it will not have is their headcount freeze, a disputed invoice, or a champion interviewing elsewhere.