Gen AI Is Rewriting B2B Buying Before Vendors Even Know It

The dark funnel problem

A pharmaceutical company invests $2 billion developing a cancer drug and $200 million annually on physician engagement. But when oncologists query AI clinical assistants about treatment options, a competitor's drug dominates the synthesized answers—not because it's superior, but because its content was engineered for algorithmic retrieval.

This isn't speculation. OpenEvidence, a clinical AI assistant, now serves more than 40 percent of U.S. physicians daily and handled over 20 million queries in January 2026, up sevenfold from December 2024, according to research first reported by Harvard Business Review. OpenAI and Anthropic both launched healthcare-specific LLMs in early 2026, positioning AI at the center of clinical decision-making.

The shift extends far beyond healthcare. B2B buyers across industries increasingly use AI tools to discover vendors, compare specifications, and evaluate fit before ever contacting sales teams. International Data Corporation predicts 62 percent of traditional B2B demand generation will be AI-led by 2028. Yet most of this traffic remains invisible to sellers—a phenomenon researchers call the "dark funnel."

Why it matters

Traditional B2B go-to-market investments—sales force training, channel partnerships, controlled media—are being bypassed by AI-mediated discovery that companies neither own nor fully understand. Among U.S. B2B technology buyers, 75 percent now complete purchases in 12 weeks or less, down from 11 months in 2024. The buying funnel widens at discovery but narrows far earlier, and vendors lack visibility into how their content influences these compressed decisions.

Where disruption hits hardest

Industrial manufacturers face similar pressure. AI-powered procurement platforms like Arkestro and Keelvar analyze technical specs and recommend sourcing strategies automatically. IMI, a UK-based engineering firm specializing in fluid control, discovered HVAC installers had stopped using Google search in favor of ChatGPT and Gemini for product research. "Today it's much easier to find alternatives, compare features and prices, and equip yourself to push the conversation with B2B suppliers even if you don't have the knowledge," IMI's marketing director Marcella Colombino told HBR.

Banking shows early signs of the same pattern. While JPMorgan Chase reports AI-driven payment validation reduced fraud and account rejection rates by roughly 20 percent, these remain internal deployments. The Hong Kong Institute for Monetary and Financial Research found 75 percent of financial institutions view gen AI primarily as a productivity tool. Meanwhile, commercial banking customers increasingly use AI to analyze cash management options and compare treasury structures, eroding the relationship-led models banks have relied on.

The 4C response framework

Researchers propose four disciplines for what they call "generative readiness":

Coordination across silos. Marketing, research, product, legal, and communications teams must align narratives for accuracy, consistency, and governance. AI systems treat press releases, technical docs, and Reddit posts as equally weighted sources.

Citability for machine readers. Pharma companies are rethinking journal selection, prioritizing open-access publications over paywalled prestige journals. IMI restructured product knowledge with schema markup, comparison tables, and "AI-snackable" micro-answers that LLMs can parse without ambiguity.

Credibility signals. Authority varies by industry. In pharma, medical guidelines and peer-reviewed journals carry weight. A Digitas UK pilot study found over 80 percent of fintech payment solution sources cited by LLMs came directly from brand sites like Stripe and PayPal—leaders in R&D and technical content. For manufacturers, credibility means ensuring specs appear in industry databases and certifications are registered with standards bodies.

Calibration through listening. GSK audited one COPD treatment brand by testing 6,000 prompts across nine decision nodes in the healthcare practitioner journey, monitoring how often and in what context the brand appeared in AI-synthesized answers.

McKinsey's 2025 B2B Pulse Survey found only 19 percent of respondents are implementing gen AI use cases for buying and selling. The gap between AI adoption by buyers and seller readiness is widening.

The details were first reported by Harvard Business Review in research by Amit Joshi, Ivy Buche, and Caroline Schwaer.