Will We Start Selling to AI Buyers?
Ross Sylvester, Co-Founder & CEO, Adrata | Mar 2026 | ~12 min read
Last October, a mid-market SaaS company in the observability space lost a $380,000 deal to a competitor they had never heard of. The loss was not unusual in itself. What was unusual was the debrief. The champion — a VP of Engineering at a Fortune 500 retailer — told their AE something unsettling: "We never evaluated you. Our procurement system did."
The retailer had deployed SAP's Joule agent across its procurement workflow six months earlier. When the VP's team submitted a request for an observability platform, Joule ingested the requirements, queried vendor data across G2, Gartner Peer Insights, public API documentation, and pricing pages, then produced a ranked shortlist of four vendors with confidence scores. The observability company that lost? It was not on the list. Not because its product was inferior, but because its pricing was gated behind a "Talk to Sales" form, its API documentation required authentication to view, and its G2 profile had not been updated in eleven months. The agent could not evaluate what it could not see. So it didn't.
This is not a hypothetical scenario. It is happening right now, across industries, at an accelerating pace. And the companies on the wrong end of it do not even know they are losing.
The Machines Are Already Shopping
The conventional wisdom in B2B sales is that AI-driven procurement is a 2028 problem. Gartner's headline forecast — 90% of B2B purchases intermediated by AI agents by 2028, commanding over $15 trillion in annual spend — reinforces this framing.^1^ It feels safely distant. Three product cycles away. Something to put on next year's strategic plan.
This is a dangerous misread. The future is not arriving on schedule. It is arriving in patches, unevenly, and faster than the forecasts suggest in specific categories.
Walmart deployed an AI procurement agent in Q3 2025 for indirect spend categories — IT services, marketing technology, professional services. The system, built on a proprietary model trained on Walmart's purchasing history and vendor performance data, now handles initial vendor screening for roughly 40% of indirect procurement requests. A Walmart spokesperson told Supply Chain Dive that the average time from requisition to shortlist dropped from fourteen days to thirty-six hours.^2^
ServiceNow's Now Assist platform, launched in its Washington D.C. release, includes procurement agent capabilities that auto-evaluate software vendors against predefined criteria. The system pulls data from public sources — G2 reviews, security certifications, SOC 2 reports, pricing pages, API documentation — and generates comparative scorecards without human intervention. ServiceNow reported that early adopters reduced their vendor evaluation cycle by 60%.^3^
SAP's Joule, embedded across the SAP ecosystem, handles procurement workflows for over 300 enterprise customers as of January 2026. McKinsey's latest analysis of AI adoption in procurement estimates that 23% of Fortune 500 companies are now using some form of AI agent in their vendor evaluation process, up from 6% in early 2025.^4^ The trajectory is not linear. It is exponential.
And here is the detail that should keep every CRO awake at night: these systems are not announcing themselves. There is no notification that says "An AI agent is currently evaluating your company." The procurement agent runs silently, in the background, before any human buyer ever visits your website or responds to your outbound sequence. By the time the RFP arrives — if it arrives at all — the decision architecture is already set.
What the Agent Sees When It Looks at You
To understand why this matters practically, you need to understand how a procurement AI actually evaluates a vendor. It is nothing like how a human buyer works.
A human buyer starts with awareness. They see your ad, hear your name at a conference, get a cold email from your SDR, or read a blog post. They form an impression. They visit your website, watch a demo video, maybe download a whitepaper. The entire top of your funnel is designed for this sequential, impression-driven process.
A procurement agent does not have impressions. It has inputs.
When ServiceNow's Now Assist evaluates a vendor, it executes a structured query across multiple data sources simultaneously. Forrester's 2025 analysis of AI procurement systems identified the primary signals these agents weight, in approximate order of influence: structured product specifications and feature matrices, third-party review data (G2, Gartner Peer Insights, TrustRadius), published and machine-readable pricing, API documentation accessibility and completeness, security certifications and compliance documentation, integration ecosystem breadth, support SLA terms, and community forum activity and responsiveness.^5^
Notice what is not on that list. Your brand. Your narrative. Your customer testimonials in PDF format. Your beautifully produced customer story video. Your executive briefing center. Everything your marketing team spent the last three years building for human consumption is invisible to the machine that is now making the first cut.
This is not a flaw in the agent. It is a feature. The agent is doing exactly what procurement has always wanted to do but lacked the capacity for: evaluate every viable vendor against every relevant criterion without bias, fatigue, or political pressure. The machine is a better procurement analyst than any human. And it is merciless about what it cannot parse.
The AEO Imperative
In the early 2000s, a small industry emerged around a simple insight: if Google is how buyers find you, you should optimize for Google. Search Engine Optimization became a multi-billion dollar discipline. Companies that invested early in SEO built durable competitive advantages. Companies that dismissed it as a fad spent years trying to catch up.
We are at the same inflection point with AI Engine Optimization.
AEO — the practice of structuring your digital presence so that AI systems can accurately discover, evaluate, and recommend your product — is not a marketing trend. It is the new competitive surface for revenue teams. And unlike SEO, which played out over a decade, the AEO window is compressing into months.
The reason is architectural. Google's algorithm rewarded a specific set of signals — keywords, backlinks, domain authority, page speed — and the SEO industry reverse-engineered those signals over years of experimentation. AI procurement agents reward a different set of signals, but they are far more transparent about what they want. They want structured data. They want machine-readable specifications. They want verifiable claims. They want open access.
Bing confirmed in late 2025 that its LLM-powered search heavily weights schema markup and structured data when generating AI responses.^6^ Google's AI Overviews follow the same pattern. Perplexity's retrieval system explicitly prioritizes sources with clear, hierarchical content structure. The playbook is not hidden. It is documented.
Yet most B2B companies have not adapted. A January 2026 audit by Demandbase found that only 12% of B2B SaaS companies have implemented comprehensive schema markup on their product pages. Only 8% publish machine-readable pricing. Only 15% make their API documentation accessible without authentication. The vast majority of B2B vendors are, from an AI agent's perspective, partially invisible.^7^
This creates an extraordinary window of opportunity for the companies that move first.
The Pricing Transparency Problem
Of all the signals that procurement AI evaluates, pricing is the most politically fraught for vendors — and the most consequential for agent inclusion.
The B2B SaaS industry has spent two decades perfecting the art of pricing opacity. "Contact us for a quote." "Custom pricing for enterprise." "Let's schedule a call to discuss your needs." This approach was rational when the buyer was a human who could be anchored, upsold, and negotiated with in real time. Opacity was a feature, not a bug, because it preserved the seller's information advantage.
Against an AI agent, opacity is a disqualification.
When a procurement agent at a Fortune 500 company evaluates five vendors for a CRM platform, and four of them publish transparent pricing while one says "Contact Sales," the agent has four data points and one null. It cannot evaluate what it cannot see. In most agent architectures, a null on a weighted criterion either eliminates the vendor from the shortlist entirely or assigns a penalty score that pushes them to the bottom.
OpenAI understood this early. When they launched ChatGPT Enterprise and later ChatGPT Team, they published per-seat pricing prominently on their website — a departure from the enterprise SaaS norm. Anthropic followed with Claude for Enterprise. Both companies recognized that in a world where AI agents would evaluate AI products (the recursion is worth pausing on), pricing transparency was a competitive weapon, not a vulnerability.
Kyle Poyar at OpenView Partners documented this shift in his analysis of B2B pricing trends: companies that moved to transparent, published pricing in 2025 saw a 34% increase in inbound qualified pipeline compared to companies that maintained gated pricing.^8^ The causation is direct. When an AI agent can include you in its evaluation, more humans see your name on the shortlist.
This does not mean every company should publish its enterprise pricing on its homepage. Complex, multi-product platforms with usage-based components have legitimate reasons for customized pricing. But it does mean that some version of your pricing — a starting point, a range, a per-seat baseline, a self-serve tier — needs to be machine-readable. The agent needs a number to work with. Give it one, or be invisible.
Your Website Has a New Audience
Here is a thought experiment that should reshape how your marketing team thinks about web strategy. Visit your own website. Now imagine you are not a human scanning the page — you are a language model parsing the DOM.
What do you see?
If your homepage is a beautifully designed hero section with a tagline like "Accelerate Your Revenue" and a gradient background, the agent sees nothing useful. No structured data about what your product does. No machine-readable feature specifications. No schema markup identifying your product category, target market, or key capabilities.
If your product page is an interactive demo that requires JavaScript to render, the agent may not be able to access the content at all. Many procurement agents use lightweight rendering engines or direct API calls to evaluate web content. A React-rendered single-page application with no server-side rendering and no structured data in the HTML is, to a significant class of AI agents, a blank page.
If your case studies are PDFs behind a form fill, they do not exist in the agent's evaluation. If your integration partners are listed as logos without structured data identifying the specific integration capabilities, the agent cannot match them against a buyer's requirements. If your security certifications are mentioned in marketing copy but not tagged with schema markup, they carry less weight than a competitor's certs that are properly structured.
The fix is not to rebuild your website for machines at the expense of humans. It is to build for both simultaneously. This is the same lesson companies learned with mobile: you do not create a separate mobile site. You create a responsive experience that serves all contexts. AEO is the same discipline applied to a new consumption layer.
Practically, this means three things. First, implement comprehensive schema markup — Product, SoftwareApplication, Organization, Review, FAQ, and HowTo schemas at minimum. Second, ensure that critical product information is available in the raw HTML, not just rendered via client-side JavaScript. Third, create a machine-readable product specification page — a structured document that lists your capabilities, integrations, pricing tiers, security certifications, and SLA terms in a format an agent can parse without ambiguity.
Some companies are going further. Notion published a dedicated /for-ai page on their website in late 2025 — a structured, machine-readable summary of their product designed explicitly for AI consumption. It includes product specifications, pricing, integration capabilities, security certifications, and competitive positioning, all formatted with schema markup and clean semantic HTML. Within three months, Notion's citation rate in AI-generated vendor comparisons increased by 47%.^9^
The Documentation Moat
There is a specific category of company that is disproportionately advantaged in an AI-mediated buying world: companies with exceptional documentation.
Stripe is the canonical example. Stripe's API documentation has been widely regarded as the gold standard in developer tools for over a decade. It is comprehensive, well-structured, consistently formatted, and publicly accessible. Every endpoint is documented with request/response examples, error codes, and edge case handling. When a procurement agent evaluates payment processing vendors, Stripe's documentation provides a rich, structured dataset that the agent can parse and compare against requirements with high confidence.
Twilio is another. HashiCorp. Cloudflare. DataDog. The companies that invested heavily in documentation as a product — not as a support afterthought — are discovering that their documentation is now a sales asset in ways they never anticipated.
This is because procurement agents treat documentation as a proxy signal for product quality and vendor maturity. Forrester's analysis found a 0.73 correlation between documentation quality scores (as assessed by AI systems) and vendor shortlist inclusion rates.^5^ The logic is intuitive: a company that documents its product thoroughly is more likely to have a well-engineered product, a mature support operation, and a customer-centric culture. The documentation is evidence.
For companies whose documentation is thin, outdated, or gated behind authentication, this is an urgent problem. Not a Q3 initiative. Not a "nice to have when we get to it." An urgent, pipeline-affecting problem that is costing you deals you do not know you are losing.
The Review Economy Gets Real
G2 has 2.5 million reviews across 175,000 software products. Gartner Peer Insights has over 500,000 verified reviews. TrustRadius, Capterra, and Software Advice add millions more. For years, B2B companies treated these platforms as marketing vanity metrics — nice badges for the website, useful for the occasional prospect who asked for references.
Procurement AI has turned these platforms into pipeline infrastructure.
When an agent evaluates vendors, review data is one of the highest-weighted inputs. But the agent does not read reviews the way a human does. A human might skim a few reviews, anchor on a particularly compelling or alarming one, and form a general impression. The agent ingests every review, applies sentiment analysis, weights by recency and reviewer credibility, and produces granular scores across specific dimensions — implementation quality, support responsiveness, product reliability, feature completeness, value for price.
This means that your aggregate review profile is now a quantitative input to deal outcomes. A single unanswered negative review about onboarding quality does not just look bad to a human browsing G2. It mathematically reduces your score on the "implementation quality" dimension in every agent evaluation that includes that data.
G2's own research shows that vendors who actively manage their review profiles — soliciting reviews from successful customers, responding substantively to negative reviews, and maintaining a cadence of recent reviews — see 2.3x higher inclusion rates in AI-generated vendor recommendations compared to vendors with stale or unmanaged profiles.^10^
The implication is clear: review management is no longer a marketing function. It is a revenue function. It belongs in the CRO's operating cadence, measured with the same rigor as pipeline coverage and win rates.
What to Do Monday Morning
The strategic picture is clear enough. Procurement AI is here, it is evaluating you right now, and most companies are not optimized for it. The question is what to do about it, concretely, starting this week.
First, run the audit. Open ChatGPT, Claude, Perplexity, and Gemini. Ask each one: "What are the best [your category] platforms for [your target segment]?" Do this for five different phrasings of the query. Document where you appear, how you are described, what data is cited, and where your competitors show up instead. This takes an hour and it will tell you more about your competitive position in the agent-mediated world than any analyst report.
Second, publish a machine-readable product specification page. Not a marketing page. A structured document with schema markup that lists your product capabilities, pricing tiers (even if approximate), integrations, security certifications, SLA commitments, and deployment options. Make it accessible without authentication. Make it indexable. This is your new storefront for a buyer that never blinks.
Third, open your API documentation. If your docs require a login, a significant class of procurement agents cannot evaluate your technical capabilities. The security risk of public API documentation is minimal — these are reference docs, not access credentials. The revenue risk of gated documentation is now measurable and growing.
Fourth, assign an owner to your review profile across G2, Gartner Peer Insights, and TrustRadius. Not a junior marketer who checks it quarterly. A senior person with a monthly cadence, a review solicitation workflow tied to customer success milestones, and a mandate to respond to every negative review within 48 hours. This person's impact on pipeline will exceed most SDRs within six months.
Fifth, talk to your buyers. Ask them directly: "Are you using any AI tools in your vendor evaluation process?" The answers will calibrate your urgency. If even 20% say yes, you are already behind.
The Window
There is a pattern in technology adoption that Clayton Christensen documented decades ago: incumbents dismiss disruptions that initially serve only the low end of the market. By the time the disruption moves upmarket, the window for response has closed.
AI procurement is following this pattern precisely. Today, it is concentrated in indirect spend categories — IT services, marketing technology, SaaS tools. The deals are mid-market. The agent's evaluation is one input among several, not the sole decision-maker. It is easy to dismiss.
But the trajectory is unmistakable. McKinsey projects that by 2027, AI agents will be involved in 60% of enterprise software purchases above $100,000 ACV.^4^ Gartner forecasts that AI agents will outnumber sellers by 10x by 2028.^11^ The window to optimize for this reality — to make your product, your pricing, your documentation, your review profile machine-readable and agent-evaluable — is open right now. It will not stay open long.
The companies that move in the next twelve months will build structural advantages that compound over time. Better agent visibility leads to more shortlist inclusions, which leads to more pipeline, which leads to more customers, which leads to more reviews, which leads to better agent visibility. The flywheel is real, and it favors first movers.
The companies that wait will discover, like the observability vendor that opened this article, that they are losing deals to competitors they have never heard of — filtered out by a machine that could not see them.
The question is not whether AI will start buying. It already is. The question is whether it can find you.
Footnotes
^1^ Gartner, "AI Agents Will Command $15 Trillion in B2B Purchases by 2028," Digital Commerce 360, November 2025.
^2^ Supply Chain Dive, "Walmart Deploys AI Procurement Agent for Indirect Spend Categories," September 2025.
^3^ ServiceNow, "Now Assist for Procurement: Washington D.C. Release Overview," ServiceNow Documentation, 2025.
^4^ McKinsey & Company, "The State of AI in Procurement: 2026 Update," McKinsey Digital, January 2026.
^5^ Forrester Research, "The AI Procurement Agent Landscape: What Vendors Need to Know," Forrester Report, Q4 2025.
^6^ BOL Agency, "What Is GEO and AEO? How AI Is Changing B2B SEO in 2026," 2026; Bing and Google structured data confirmations.
^7^ Demandbase, "B2B SaaS AI Readiness Audit: January 2026 Findings," Demandbase Research, January 2026.
^8^ Kyle Poyar, "The End of Pricing Opacity: Why Transparent Pricing Drives More Pipeline," OpenView Partners, 2025.
^9^ Notion, "/for-ai: Our Machine-Readable Product Page," Notion Blog, November 2025.
^10^ G2, "AI-Generated Vendor Recommendations: How Review Profiles Drive Inclusion," G2 Research, 2025.
^11^ Gartner, "By 2028 AI Agents Will Outnumber Sellers by 10X," Gartner Newsroom, November 2025.