Problem-first selling: what buyers still need from sellers when AI handles the information

A buyer can now ask an AI assistant to compare five vendors, summarize their reviews and draft questions for the demo. The product facts that sellers used to carry are a prompt away. So what do buyers still need from a seller? The answer is help with their problem, not more about your catalog. This guide sets out the six jobs that remain, shows what changes in a first meeting, and gives a worked example.

What do buyers still need from sellers when AI handles the information?

Buyers still need sellers for judgment about their own situation: what the problem is, what it costs, and whether a solution fits. A Gartner survey of 632 B2B buyers, published in June 2025, found 61% prefer a rep-free buying experience overall. The same research found buyers prefer seller input for tasks that need contextual judgment, such as whether a product fits their company's needs. A Gartner analyst advised that sellers should offer unique guidance and act as a sounding board, rather than repeat generic information.

Buyers also arrive later and better informed. 6sense's 2025 Buyer Experience Report, based on nearly 4,000 buyers, found first contact came when buyers were about 61% through their buying process. Buyers initiated that contact 79% of the time. They bought from the vendor they ranked first 77% of the time.

Put together: the information job has moved to the buyer's own research and their AI. The decision job is still open, and that is where a seller earns a place.

What are the six jobs of a problem-first seller?

A problem-first seller does six jobs, and every one is about the buyer's problem:

The job most teams skip is number 4. Sellers optimize for the champion and forget the other people who can quietly say no. Gartner found buying groups that reach consensus are 2.5 times more likely to report a high-quality deal. See buying groups and consensus.

What changes in a problem-first first meeting?

The first meeting changes from the vendor's world to the buyer's problem. Many first meetings still open with who we are, our logos, our modules and our best case study. The buyer sits through it, waiting for it to become about them. The difference shows in five places:

The realism line: you cannot name their number if discovery did not find one. Problem-first selling starts in call one, not on slide one.

Where does the pivot from discovery to the problem happen?

The pivot happens in the first five minutes of call two. A common pattern among newer reps: call one goes well, with good questions and a curious buyer. Then call two opens with "let me show you the platform", and everything learned disappears into a demo. Use a fixed sequence:

1. Call one is discovery. No pitch. Find the problem and who owns it.
2. Open call two with it. "Here is what I heard last time. Tell me what I got wrong."
3. Put a number on it. What does this cost per quarter, in their numbers?
4. Show the path in three steps, not the full catalog.
5. Confirm the sponsor. Who signs, and do they see the problem the same way?
6. Only then, propose.

The moment a buyer corrects your problem statement is the most valuable minute in the deal. You learn what they care about, and they start to own the story. Some buyers want the demo first. Give it, then still state the problem back before you discuss price.

How should AI-written follow-up change?

AI-written follow-up should carry the buyer's words and the next decision, not a default template. Gartner found 73% of B2B buyers actively avoid suppliers who send irrelevant outreach. If every seller uses the same assistant with the same default prompt, every email reads the same: a polite opening, a paragraph about the vendor, a soft ask. The buyer's own AI may summarize it as "vendor follow-up, no action needed".

Emails that get replies differ in five ways:

• The opening repeats something the buyer said, not a pleasantry.
• The ask is specific: two time slots, not "would be great to connect".
• There is one decision to make, not zero and not three.
• It is about five lines, not nine.
• It sounds like the conversation you actually had.

AI drafts are fine. The fix is the input: give the model the buyer's words, their number and the next decision. A great email cannot save a weak call. It can only carry the problem forward.

How should targeting change?

Targeting should start from why customers changed, not from what they look like. AI agents can now take one good customer and find 20 lookalikes in minutes. But most lookalike methods copy the profile, and buyers do not buy on profile:

Start the search from "why they changed", then filter on firmographics. At the top of the funnel, profile lookalikes still beat nothing. But they fill a list, not a pipeline. If your CRM has no field for why the customer changed, add one.

Example: a problem statement with a number

Example: in call one, a seller learns that the buyer's team creates 400 stage-2 opportunities a quarter. Conversion from stage 2 to stage 3 has fallen from 40% to 30%. Average deal size is $25,000, and about a quarter of stage-3 deals close.

The seller opens call two like this: "Last time you said stage-2 deals stall after the first demo. Conversion dropped from 40% to 30%. That is 40 fewer stage-3 deals a quarter. At your close rate and deal size, about $250,000 a quarter in bookings. Did I get that right?"

The buyer corrects one number: the close rate is closer to 20%. The cost becomes $200,000 a quarter. That correction is the point. The buyer now owns the number, and it goes on the one-page case the champion forwards to finance.

How do you train problem-first selling?

Train it as a skill with practice, scoring and manager follow-up, not as a workshop. Three parts work well:

• A roleplay of the call-two opening, practiced weekly for six weeks. See AI roleplay for sales training.
• Three scored behaviors: states the problem back, quantifies the cost, confirms who else must agree. Use one talk track and one skills scorecard.
• One manager question in each one-to-one: "What number did the buyer give you, and did they correct it?"

Start in onboarding, where habits form fastest.

Common mistakes

• Opening with the company story. The buyer has read it, or asked their AI.
• Running discovery and never saying the problem out loud. The pivot belongs in the first five minutes of call two.
• Using the seller's numbers instead of the buyer's. A cost the buyer did not supply is a claim, not a problem statement.
• Bringing ten insights. Bring one. A number, not a deck.
• Sending default AI-written follow-ups. Feed the model the buyer's words and the next decision.

If you want to know whether your enablement team has the hours to build and coach this, a free Capacity Check takes 20 minutes.