Demandbase announced three new AI agents on 20 October 2026, including an AEO Agent that it says can examine how AI assistants describe and recommend a brand for specific B2B buying-group members. The announcement is not evidence that AI visibility has become a proven pipeline lever. It is evidence that the measurement question is changing: a brand can be present in an answer yet still be unconvincing, incomplete or absent for the person who must approve the next step.
The shift matters because enterprise buying rarely happens through one query or one stakeholder. A CFO, CTO, CMO, security lead and procurement director can ask different questions about the same supplier. If an AI assistant provides different evidence to each person, reporting one combined AI visibility score can obscure the actual commercial weakness.
What Demandbase announced
Demandbase's 20 October announcement introduced the AEO Agent, Campaign Agent and Audience Agent. The company says the AEO Agent is designed to help teams compare how major AI assistants describe and recommend a company, assess question-level visibility and sources, identify information gaps, activate campaigns and identify anonymous AI-driven traffic at account and contact level. These are product-positioning claims, not independently demonstrated outcomes.
Demandbase describes the Campaign Agent as a route for connecting campaign activity to revenue goals and recommending adjustments. It describes the Audience Agent as a way to build and refresh audiences from ideal-customer profiles, buying groups and account activity, including missing buying-group members. Its earlier Mojo product is not one of the three new agents.
The company calls this B2A, or business-to-agent, marketing. Its definition is useful as a product lens, but it should not be treated as settled industry taxonomy. Demandbase also says its platform includes more than 342 million contacts across 13 million companies. The announcement does not publish the methodology, freshness, uniqueness, legal basis or performance evidence behind that data-scale claim.
Why generic AI share of voice can mislead a B2B team
A generic mention can be a useful discovery signal. It is not sufficient evidence that a buying committee can make a confident decision. Consider an enterprise cybersecurity provider:
| Buyer role | Typical question | Evidence an answer should surface | Risk if evidence is missing |
|---|---|---|---|
| CFO | What does this solve and what is the financial exposure? | Cost logic, risk boundaries, implementation assumptions | The brand sounds interesting but financially vague |
| CTO or CISO | Can this work safely in our environment? | Technical scope, integrations, governance and limitations | The brand is excluded before a commercial conversation |
| Procurement | Is this supplier credible and contract-ready? | Delivery model, security evidence, commercial clarity | The shortlist stalls at comparison |
| CMO or growth lead | Will this support the required market motion? | Audience fit, message evidence and measurement method | The solution is seen as disconnected from demand generation |
The same supplier can therefore be visible to one role and invisible or unconvincing to another. That is why the next useful AEO question is not merely, "Were we mentioned?" It is, "Was the appropriate evidence available to the relevant buyer when the decision moved forward?"
The AI Buying-Committee Visibility Matrix
I use a simple matrix with four checks. It is a planning method, not a claim that an AI platform can control recommendations.
- Buyer role: Name the stakeholder and the decision they own.
- Decision stage: Separate problem identification, category evaluation, supplier comparison and final selection.
- AI representation: Record whether the brand is present, accurately described, supported by credible sources and relevant to the specific question.
- Commercial evidence: Connect the observation to account engagement, sales-qualified opportunities and pipeline only where the evidence supports that connection.
This approach pushes a team to validate an answer against a real buyer task. A landing page, case study, service description, source citation and product explanation may all need to work together. AI-powered ABM and AI visibility benchmarking become more useful when they are tied to that role-and-stage context.
Do not confuse a forecast with a market fact
Demandbase cites a Gartner prediction that, by 2028, 90% of B2B buying will be AI-agent intermediated and more than $15 trillion of B2B spend will move through AI-agent exchanges. Gartner presents that statement as a forecast, not a description of all current B2B buying.
The actionable takeaway is not to assume that conventional research has disappeared. It is to prepare information that can withstand more forms of mediated evaluation. Buyers still need useful, accurate, source-backed pages. Sales teams still need to validate fit. Procurement still needs evidence. AI systems may change the route into that work, but they do not remove the need for it.
A practical B2B AEO measurement model
A commercially useful programme should combine four evidence layers:
- Answer observations: A documented set of role-specific questions, answers and cited sources.
- Content and entity evidence: The visible pages, authorship, proof, structured data and internal routes that support a claim.
- Account activity: Relevant account engagement and sales context, handled with appropriate consent and attribution limits.
- Commercial progression: Sales-qualified opportunities, stage movement and pipeline reviewed alongside other active channels.
This preserves a distinction that is often lost in AI search conversations. A citation is an observation. A qualified opportunity is a sales event. Pipeline is a commercial measure. They can inform one another without becoming the same metric. Use the AEO measurement guide to set baselines and decision rules before adding more dashboards.
What a CMO should do next
Begin with a finite buying group, not every conceivable prompt. Choose a priority account segment. List the five roles that shape the decision. Document the questions each role asks at each stage. Then audit the visible evidence each answer could use: product information, technical boundaries, commercial proof, governance, customer outcomes and supplier reassurance.
The best outcome is not a larger collection of screenshots. It is a short, testable backlog: which claim needs a stronger source, which page needs a clearer answer, which service path is missing, and which sales handoff must be measured. That is where SEO, AEO and GEO work can become a real commercial system instead of a claim about influencing an opaque answer engine.
Sources
- Demandbase: B2A AI agents announcement
- Gartner: top predictions for IT organizations and users
- Business Wire distribution of the Demandbase announcement
About the Author
Modi Elnadi is the founder of Integrated.Social. He works with B2B teams on evidence-led AI search, account-based marketing and commercial measurement. His approach is practical: define the buyer question, make the evidence visible, validate the released page and separate observed AI visibility from claims about pipeline or revenue.
