B2B buyers now form a shortlist in a private chat window, hand an AI assistant their procurement rules and security questionnaire, and meet a seller only to check what the assistant said. Five vendor documents published in the past week show what that does to the seller’s job: the answers that matter are the ones a seller can defend.
The shortlist forms before the first call
On October 22, Seismic published a post on where enterprise sellers fit in AI-assisted B2B buying journeys. It draws on the company’s 2026 The New Go-to-Market Standard for B2B Selling report, which we have not seen in full, so every figure below is as Seismic’s post states it. The post says answer engines now condense years of market positioning into a few paragraphs, and that this summary “replaces an organization’s homepage and a seller’s call opener as the framing for how buying councils see vendors.”
The post describes buyers feeding those tools their own context. A committee member can paste in procurement rules, a security questionnaire and last quarter’s numbers, and the recommendation then reads like an analyst briefing written for that account. Seismic also notes that each stakeholder tends to research vendors alone, in the AI system they prefer, so champions, economic buyers and technical buyers arrive with different and sometimes conflicting details.
Advertisement
Seismic cites Deloitte’s 2026 B2B commerce research for the scale: around 40% of enterprise buying groups have enlisted agentic AI to vet vendors, draft orders, comb through contracts and sanity-check pricing models. We could not check the Deloitte figure at its source, and it reaches us through a vendor’s blog post.
Where buyers still ask for a person
The same post is more useful on where buyers stop trusting the assistant. Robert Blaisdell, Sales Practice VP Analyst and Chief of Research at Gartner, is quoted in it as saying that “B2B buyers are more comfortable using digital channels and GenAI to navigate the purchase process on their own, but that does not eliminate the role of the seller.” Seismic adds a list of tasks buyers still hand to a salesperson:
- 55% rely on a seller to negotiate pricing and contract terms.
- 49% lean on a seller to evaluate return on investment.
- 45% turn to a seller to understand risk.
- 43% ask a seller to interpret how the product fits their own business and technology environment.
- 39% want industry expertise and peer references.
- 37% use a seller to navigate internal alignment.
- 34% speak with a seller to check claims found in AI tools.
These shares overlap, because one buyer can lean on a seller for several tasks, and the post does not state the sample size or when the survey ran. They are best read as a ranking of where buyers still want a person, with pricing and return on investment at the top and peer references and internal alignment further down. The ranking tells a revenue team which parts of the deal to staff with its most experienced people.
Two other figures from the post frame the problem. Seismic reports that 93% of scaled B2B organizations still go ahead with a sizable purchase while their buying group holds unresolved doubt at least a quarter of the time, and that only 7% always wait until everyone feels fully confident. Read together, the percentages say that buyers do not wait for certainty. They buy once someone has given them an answer they can repeat to a boss or an auditor. The report’s own phrase for the seller’s role is to help buyers “make sense of abundant information, reconcile competing requirements, quantify value, surface risk, and stand behind a recommendation.”
The buyer’s agent also calls you
Research is only one use of a buyer’s assistant. In a Salesforce piece on contact centers bylined to Kishan Chetan, EVP and GM of Agentforce and Service Cloud, consumers hand errands such as cancelling subscriptions to personal agents, and an agent asked to watch a delayed order “can call back every few hours until the order ships.” Because such an agent does not keep business hours, the piece says, its calls can arrive at 2 a.m. as easily as at noon. His advice is that “The best way to prepare is to meet personal agents at the front door with your own business agents.”
That article is about consumer service, and Salesforce sells the tools it recommends, so it is evidence of direction, and it measures no B2B volume. We covered it in a brief on the contact-center warning. Its relevance here is the pattern. When the counterparty is software, the seller’s side needs software of its own, and a human enters when the exchange needs a decision.
Vendors answer with agents of their own
The vendor documents from the past week fit that pattern. On October 21, SAP published its Customer Experience Q3 2026 release notes. Balaji Balasubramanian, president and chief product officer for SAP Customer Experience and Consumer Industries, writes in the release post that “The next stage of customer experience requires more than intelligent recommendations.” The sales section introduces a Deal Qualification Assistant that automates lead qualification, enrichment, routing and follow-up, with separate agents for enrichment, qualification and conversion. SAP says the capabilities are available to select customers. We covered the sales section in an earlier article on agents between lead and opportunity, and the release publishes no conversion data for the automated steps.
ASPR AI took a different route on October 23. Its Private Sales Agent Studio release runs the company’s sales agents inside a customer’s own cloud on a model the customer’s security team has already approved. Ganesh Iyer, co-founder and CEO of ASPR AI, said that “Generic sales AI failed enterprises because it shipped their most sensitive conversations to someone else’s cloud.” The release says sales AI often stalls in security review. That is a vendor’s description of its own market, and the release offers no count of stalled deals.
The two launches answer different halves of the buyer’s behavior. SAP’s agents speed the seller’s side of qualification. ASPR’s studio is built for the buyer’s security gate, the point in a deal where the questionnaire the buyer pasted into an assistant gets answered by a real person.
The security answer is the evidence a seller carries
Seismic’s post names the audience for this evidence. It says proof should be matched to each role: payback math for finance leaders, audit trails for IT and security directors, and rollout evidence for operations. The security director’s piece is the hardest to improvise, because it depends on facts about systems that selling skill cannot supply.
Zoho’s October 22 notice on two PHP vulnerabilities shows what a defensible answer looks like. Zoho states that its products do not use PHP and that none is affected. It then lists the surfaces that were: some blogs and public static pages, a product called Gofrugal SAM, and internal services that are not reachable from the public internet. Each is marked as updated, and the notice reports no evidence of exploitation. A buyer’s reviewer, or the buyer’s assistant, can compare that structure against its own checklist line by line.
We make no claim about how often buyers’ assistants ask for such notices. A scoped, dated, surface-by-surface statement is the form a seller can forward on the day a flaw is disclosed. A general assurance of strong security is the form a reviewer sends back with more questions. We argued earlier that agentic AI vendors owe buyers security receipts, and the same logic applies to sellers who bring AI into a deal.
What it means for the sales leader
The five documents point to a short list of changes in how a revenue team prepares. None of them is proven to raise win rates, and the sources do not claim it. They are the reasonable consequences of what the sources say buyers do.
- Ask each committee member what their assistant said. Seismic recommends asking the executive, the architect and the budget holder which chatbot they consulted and how it answered, then correcting stale claims. The answers show where the buyer’s picture of your product has drifted.
- Own the corrections file. If three buyers quote three assistants, the seller has to reconcile them against current facts. That needs one maintained source for pricing, packaging, integrations and security posture, and a person accountable for keeping it current.
- Prepare the security packet before the questionnaire arrives. A list of every system that touches prospect or customer data, with owners and patch status, turns a week of email into an afternoon. Zoho’s notice is a template for the form.
- Show the math. Seismic’s finding that 49% of buyers lean on sellers for return on investment argues for ROI models whose assumptions finance can inspect and change in the meeting.
- Decide what your own agents may say. Whatever SAP, ASPR or a competitor puts between lead and opportunity will speak for your company. In our reading, a revenue leader should know which statements those agents can make about price, security and roadmap, and who reviews them.
What the evidence does not show
Most of the numbers in this piece come from one vendor’s survey, relayed in that vendor’s own post. Seismic sells software that helps sellers prepare for these meetings, so it has an interest in the conclusion that sellers matter. The Gartner quote supports the same conclusion from an independent seat but is a single statement. The Salesforce article is about consumers. SAP and ASPR describe products that are launched or available to select customers, with no published results.
What the documents do establish is narrower and still useful. Buyers research with assistants before they talk to a seller. They still rely on sellers for pricing, return on investment, risk and security. Vendors are building agents for the seller’s side, and at least one is built for the security gate. A sales team that treats the buyer’s assistant as a stakeholder, and keeps its own facts in order, is working with the same information its buyers have.
Source: Seismic (where enterprise sellers fit in AI-assisted B2B buying journeys)
