B2B Buyers Using AI: You're the Validation Call
Your prospect already asked ChatGPT about your category, your competitors, and probably your pricing. By the time they book a call, they don't want your pitch — they want to find out whether the story they've already assembled survives contact with someone who actually does this for a living.
Gartner found that 69% of B2B buyers go to sales reps to check what AI already told them. That single number reframes the job. B2B buyers using AI aren't arriving uninformed and curious — they're arriving informed and suspicious, holding a set of conclusions they'd like you to confirm or blow up.
They didn't book you to hear what ChatGPT already said. They booked you to see if you can prove it wrong.
Key takeaways
The blank-slate discovery call is dead. Most buyers now arrive with an AI-generated view of the market, a shortlist, and a rough evaluation framework already in hand.
Your job shifted from informing to adjudicating. You confirm what's right, correct what's wrong, and add the account-specific detail no general-purpose model could know.
Generic credibility no longer works. "We're the leader in X" is exactly what the buyer's AI already told them about three other vendors.
The differentiator is evidence specific to their account — their stack, their team, their constraints, their numbers.
Prep has to change too. If your pre-call brief doesn't include "what a general-purpose AI probably said about us and our competitors," you're walking in half-prepared.
What changes when B2B buyers using AI run their own research first
The short answer: the first 15 minutes of your call are no longer yours. The buyer has already spent them, alone, with a model that summarized your category into a tidy comparison table.
That means three things are already decided before you dial in:
The frame. They've chosen the criteria they think matter — often generic ones like "integrations, pricing, ease of use."
The shortlist. You're usually one of three to five names the model surfaced, which may include vendors you don't consider real competitors.
The objections. They've read a synthesis of G2 reviews, Reddit threads, and your own website copy, and already formed a hypothesis about your weakness.
If you open with a company overview, you're re-reading a book they've already skimmed. Worse, you're confirming that talking to you adds nothing their AI didn't.
What the buyer's AI can and can't know
The buyer's AI is good at | The buyer's AI can't do |
|---|---|
Summarizing your public positioning | Knowing which of your features actually work well at their data volume |
Building a generic vendor comparison table | Predicting how their procurement team will react to your MSA |
Listing common objections in your category | Telling them which internal stakeholder will block the deal |
Reciting your pricing page | Knowing what a similar customer with their exact stack saw in month three |
Producing a plausible evaluation checklist | Knowing which criteria genuinely predict success vs. sound smart |
Everything in the right-hand column is your territory. That's the whole value of the call.
The validation call: a three-move structure
Treat the call as validation, not discovery — you're testing their conclusions rather than gathering facts from scratch. Three moves, in order.
Move 1: Surface what they already believe
Ask before you tell. You cannot correct a model you haven't read.
"Before I say anything about us — you've clearly done homework. What's your current read on where we fit versus the other options you're looking at? I'd rather correct a wrong impression now than in week six."
"When you looked into this, what came back as our weak spot? I'll tell you honestly whether it's true."
That second question is the highest-leverage question in AI-informed selling. It invites the objection, signals confidence, and gives you the exact terrain to fight on. Most buyers will answer it because it's easier than raising the concern unprompted.
Move 2: Confirm the parts that are right
Confirm at least one thing. Reps who contradict everything sound defensive and score as unreliable narrators.
"That's fair, and it's accurate. We are more expensive per seat than [Competitor]. Here's why that's the right trade for a team like yours, and here's when it isn't."
Conceding a true limitation buys you the right to be believed on the correction that follows. It also differentiates you from the model, which hedges everything and commits to nothing.
Move 3: Correct with account-specific evidence
This is where the call is won. A general-purpose model can only argue from public, averaged information. You can argue from the specific.
Weak correction: "Actually, our integration is very robust."
Strong correction: "That's true for teams on a standard setup. You're on HubSpot with a custom deal object and two years of call recordings in Gong — in that configuration, the thing that usually breaks is field mapping, not the sync. Here's how three customers with that same setup handled it, and here's the 10-day sequence we'd run for you."
The second version is unreproducible. That's the point. Specificity is the only credential a general-purpose AI can't forge.
How to prep for an AI-informed buyer in 15 minutes
Run this checklist before every first call. It takes a quarter of an hour and changes the entire shape of the conversation.
Ask the model what they asked. Prompt a general AI tool: "I'm a [their title] at a [their industry] company evaluating [category]. What are my options and what should I watch out for?" Read the answer. That's roughly the brief your buyer walked in with.
Note the three competitors it names. Two will be expected. One will surprise you — prepare for that one.
Write down the objection it hands them. Have a one-sentence honest concession and a one-sentence account-specific counter ready.
Pull three specifics about this account. Their stack, their headcount, a recent trigger event (funding, hire, launch, reorg).
Identify one comparable customer with a similar configuration, not just a similar logo size.
Draft one question only you could ask — something that proves you read their context, not their industry.
That last item is the tell. If every question you ask could have been asked of any company in their SIC code, you've confirmed the buyer's suspicion that the call was optional.
This is the layer that's hard to do manually at volume. Keeping a live read on each account's stack, stakeholders and risk flags across 30 open deals is exactly the work Ruby's automatic pre-call briefs are built to handle — so the 15 minutes go into strategy rather than reconstruction.
Rewrite your discovery questions for buyers who already researched
Old questions assume ignorance. New questions assume a draft answer exists and test it.
Old discovery question | Validation-call version |
|---|---|
"What are you using today?" | "Your AI probably told you to look at replacing versus layering. Which way are you leaning, and why?" |
"What's your budget?" | "What number did your research suggest this should cost, and who told you that's the number?" |
"Who else is involved in the decision?" | "When you share your summary internally, who's most likely to push back on it?" |
"What are your requirements?" | "Which of your requirements did you write yourself, and which came out of a template or a model?" |
"What's important to you in a vendor?" | "If two vendors look identical on paper, what would actually break the tie for you?" |
That fourth question is quietly devastating in a good way. Buyers frequently discover mid-sentence that half their evaluation criteria came from a generic checklist and don't reflect what their team actually needs. Helping them rewrite the criteria is the most valuable thing you can do on a first call — and it puts you in the room while the scorecard is being written.
What to track so you know this is working
Change the call, then check whether it moved anything. Four measurable signals:
Time-to-first-objection. Should drop. If real objections surface in the first 10 minutes instead of week four, the validation framing is working.
Criteria influence rate. What percentage of deals include at least one evaluation criterion you introduced? Aim to make this a tracked field.
Second-call conversion. Validation calls that add nothing don't get sequels. This is the cleanest scoreboard.
Surprise-competitor rate. How often does a name you didn't expect appear? High rates mean your prep isn't mirroring the buyer's research.
For managers, this is also a coaching lens. Listening for "did the rep ask what the buyer already believed?" is a faster review than scoring a 40-minute recording line by line — and it's the kind of pattern sales managers can hold the whole team to.
The uncomfortable conclusion
B2B buyers using AI have outsourced the informational part of the sales conversation. Everything your website says, your competitors say, and your category says is now available to them in 30 seconds, pre-summarized, free of charge.
What's left is judgment, specificity, and the willingness to say "that's wrong, and here's why, for you." That's a higher bar than the old discovery call. It's also a much better job.
The reps who lose in this environment aren't the ones with weaker products. They're the ones still delivering the overview.
If you want your reps walking into those calls already knowing the account's stack, stakeholders and likely objections, see how Ruby builds that context automatically from the calls and CRM data you already have.
Frequently asked questions
What does it mean that 69% of B2B buyers use sales reps to validate AI research?
Gartner's finding is that most B2B buyers now treat the sales conversation as a verification step rather than an information-gathering one. They've already formed a view using AI tools and want a human to confirm or correct it. Practically, it means opening with a company overview wastes your best 10 minutes — the buyer is waiting to test their conclusions, not receive new ones.
How should I open a discovery call with a buyer who already used AI?
Ask what they already believe before you present anything. A reliable opener: "You've clearly done homework — what's your current read on where we fit versus the alternatives?" Follow it with "what came back as our weak spot?" Those two questions reveal the frame, the shortlist and the objection in under three minutes, and let you spend the rest of the call on evidence that matters.
What if the buyer's AI got facts about my product wrong?
Concede one true limitation first, then correct the error with specifics. Generic denials ("that's not accurate, we're very robust") sound defensive and mirror the hedging tone the buyer already got from their model. Instead, name the exact conditions under which the claim is true or false, and reference a comparable customer with a similar stack. Specificity is what a general-purpose AI can't produce.
Does AI-assisted buyer research shorten or lengthen sales cycles?
Both, depending on the deal. Buyers arrive further along, so early stages compress. But they also arrive with more confident wrong assumptions and larger buying committees who each ran their own research. Cycles often stall in the middle rather than the beginning, which is why surfacing objections in the first call — not the fourth — matters more than it used to.
How do I prepare for a validation call without hours of research?
Run the buyer's likely prompt yourself, note the competitors and objections it returns, then pull three account-specific facts: stack, headcount, recent trigger event. That's a 15-minute routine. Tooling that already tracks call history, CRM context and stakeholder maps removes most of the reconstruction work, so prep time goes toward deciding what to prove rather than finding out what's true.
Should I ask buyers directly whether they used ChatGPT?
Yes, and neutrally. "What did your research turn up?" or "what's the current read internally?" gets you the same information without making the buyer feel caught out. Most will tell you readily, because the AI summary is the artifact they're circulating internally — and knowing what that document says lets you influence the version that reaches the decision-maker.