B2B Marketing is Driving Forward, Looking Backward

Ninety-four percent of B2B buyers now use AI somewhere in the buying process. Forrester and 6sense surveyed buyers separately and landed on the exact same number. Eighty-nine percent of B2B marketers use AI to create content Same technology. Opposite jobs. Buyers use it to decide. It does the research, narrows the field, and builds the shortlist, and 95% of purchases still come from that initial shortlist. Marketers mostly use it to produce: more content, more sequences, more campaigns, aimed at a buyer whose AI is now filtering most of it out. That’s an industry driving forward while looking in the rearview mirror. And the market has started pricing the difference. The questions were about leadership. Now they’re about value. In January I published 3 Questions B2B Marketing Still Hasn’t Answered: \* **Ownership.** Who actually owns demand creation and is accountable for revenue? \* **Demand quality.** What does quality mean when buyers are anonymous and AI-driven? \* **Outcomes.** Can the industry move from fragmented marketing tools and systems to trusted, outcome-based systems? It’s still a leadership problem. Nine months later it’s a valuation problem as well. ZoomInfo took a $650.5 million goodwill impairment in its second quarter, on revenue growth of 1.2%, and guided to a full-year revenue decline. In September, Clay, which sells signal and workflow to the same go-to-market teams, raised money at $7.1 billion, more than double the valuation it carried thirteen months earlier, after quadrupling revenue in 2025. Same buyers. Same category. Opposite valuations. It shows up inside single companies too. Salesforce stock was down 26% by early February, as software shed roughly $1 trillion of market value. By late August, its Agentforce and Data 360 recurring revenue had reached nearly $3.9 billion, up more than 210% in a year. One P&L, moving in two directions at once. Agentic AI doesn't lift or sink a sector. It splits product lines on two variables: what you get paid for, and whose AI you're building for. Most of the industry is getting the second one wrong Question #1: The buyer's AI now owns the front door of demand creation. The January question asked whether marketing, sales, or ops owns demand creation. The answer came from a direction none of them were watching. The buyer owns it, and the buyer's machine does more of the work every quarter. Here is the counter-intuitive part. Buyers are reaching out to sellers earlier than they used to: about 61% of the way through their journey, down from 69% a year before, according to 6sense. They still start that conversation themselves roughly 80% of the time. And 95% of purchases still come from the shortlist they built before the first call. Earlier contact hasn’t reopened the field. Buyers are calling to check a decision they’ve mostly made. Gartner expects AI agents to intermediate more than $15 trillion of B2B spending by 2028. Even if Gartner is off by a year, the direction isn't in question. So the ownership question has changed shape. The issue now is who owns what the buyer's AI knows, believes, and repeats about you before the buyer ever raises a hand. In most marketing organizations, nobody does. A Scribewise and Scrunch survey of 602 marketers this spring found 91% claim a documented AI search strategy, while 62% are still measured on traditional SEO metrics. The job moved. The scoreboard didn’t. Question #2: Quality is whatever the buyer’s AI trusts. In January I wrote that the industry still sells leads while buyers behave like a buying group. The buying group just picked up a member nobody counts: the agent doing the research. Forrester's 2026 research puts the typical decision at 13 internal stakeholders and nine external influencers. When the product itself has generative AI features, the group doubles, from seven people to 14. A lead is one person. An agent doesn't fill out a form. For twenty years, the seller defined quality: a scoring model, an MQL threshold, a handoff rule. The buyer's AI defines it now, and it isn't grading on machine logic alone. It rewards what's verified across independent sources, specific enough to answer a real question, and solid enough that someone in the buying group will stand behind it. It discounts volume. Volume is where most seller-side AI went. A 10Fold survey of 400 B2B tech marketers found 46% producing three to five times more content than the year before, on budgets up 10% or less. The Content Marketing Institute found 87% of marketers using AI for content report productivity gains, and 39% report better content performance. Output multiplied. Results didn’t. Compare what happens when quality measurement follows the buyer. Palo Alto Networks dropped MQLs for buying groups and, per a Forrester case study, saw a 17% higher win rate, with multi-stakeholder opportunities eight times more likely to advance. The value math follows. Anything sold by a unit the buyer’s AI makes redundant (contact records, gated-content leads, undifferentiated intent, outbound volume) is being repriced down. Verified signal that a buyer’s agent needs and can’t scrape is being repriced up. The market is paying for signal and marking down records. Question #3: Trust moved from the tool to the result. The third question asked whether the industry could move from fragmented tools to outcome-based systems. The software market answered for it this year. As I argued in The Stack That Broke Marketing, the fix for sprawl was never more technology. It was a different application of it. Buyers have stopped trusting claims and started testing them. More than 60% now run a trial before they buy, according to Forrester, and only about a third of those end up buying from the vendor whose trial they ran. The trial has become the place where shortlists flip. And buyers trust the machine about as much as they trust you. Gartner found 51% of buyers worry about misleading information from generative AI, and 49% worry about misleading information from sales reps. So they check one against the other: 69% want a sales rep to validate what the AI told them. Trust sits with whoever is accountable for the result and can prove it. Features, attribution models, and dashboards still don’t answer the CEO’s question from January. Evidence does. Both org charts are being redrawn. Only one side is doing it on purpose. **The buyer’s side added people to check the machine.** Buying groups double when AI is part of the purchase. Procurement is a decision-maker in 53% of buying cycles, per Forrester, and is increasingly involved from the earliest stages. 6sense found 58% of buyers engaged sellers earlier specifically to evaluate the vendor’s AI. Buyers built a verification layer, and it is getting larger. **The seller’s side added machines and subtracted titles.** Spencer Stuart found that 31% of S&P 500 companies no longer have an enterprise CMO, and the ones who remain average 4.1 years in the job. Postings for GTM engineers, the people who wire data, AI, and workflow into the revenue engine, grew 205% year over year to roughly 3,000 open roles in early 2026, according to one job-market analysis, and 91% of the companies that have hired one employ exactly one. Gartner expects AI agents to outnumber sellers ten to one by 2028, while fewer than 40% of sellers will say the agents made them more productive. The money follows the same pattern. Gartner’s 2026 CMO Spend Survey puts marketing budgets at 7.8% of revenue, with 15.3% of that going to AI. Only 30% of those CMOs say their organizations are ready to use it, and 56% say they lack the budget for their own AI strategy. Both sides are running into the same wall. Gartner’s procurement research calls it a productivity paradox: AI makes individuals faster, and the gain shrinks at the team level. The buyer’s side is answering with more judgment. The seller’s side is answering with more output. One side is building a windshield. The other is polishing the mirror. Five predictions Here is what I expect to happen from here, and what I’m watching for. Each one cuts against the consensus that AI makes B2B simpler, cheaper, and less human. \* **AI is making B2B buying more human, not less.** Groups double when AI is involved. Two-thirds of buyers say they prefer a rep-free experience, and more than two-thirds want a rep to check what the AI told them. Both are true at once. The seller’s job is moving from persuasion to verification, and the best-paid sellers of 2028 will be the ones buyers call to check the machine. \* **Ten agents per seller will make outbound worth less.** When every seller has ten agents writing to buyers whose own agents read the mail, unsolicited volume gets filtered at machine speed. The cost of a message falls toward zero. The cost of a real meeting with a real buying group goes up. Anyone pricing on outbound volume is pricing the wrong side of that trade. \* **More content will make you harder to find.** Teams producing three to five times more content are not seeing three to five times the results. Buyers’ AI rewards sources that are specific, verified, and consistent. I expect the strongest marketing teams to publish less, prove more, and show up on more shortlists because of it. \* **Procurement, not marketing, will set the new unit of value.** Procurement already holds a decision vote in more than half of buying cycles, and it will be the first function to hand comparison work to agents. Activity pricing (per lead, per impression, per seat) is the easiest thing for an agent to line up in a spreadsheet and squeeze. Outcome pricing is the hardest. That asymmetry will reprice whole categories faster than any analyst model expects. \* **The most important marketing hire of 2027 won’t carry a marketing title.** The CMO role is already splitting. One half is becoming the owner of what machines know and say about the company. The other half is becoming the owner of the outcome number. The leaders who hold both will keep the job and the budget. The ones holding neither will be reporting to someone who does. Where your business sits on the Windshield Test Here is an Uphoff on Media framework you can download and use as a diagnostic tool in your business. It plots each product line, rather than the company as a whole, on the two variables that decide whether agentic AI is adding value or taking it away. The vertical axis is what you get paid for, from activity (leads, impressions, content, seats, hours) to outcomes (pipeline, revenue, decisions made). Questions one and three set your height: who is accountable for the result, and whether the customer trusts you for a result or for a tool. The horizontal axis is whose AI you’re building for, from your own (more output, lower cost) to your buyer’s (being what their AI finds, trusts, and recommends). Question two sets how far right you sit: whether your idea of quality matches what the buyer’s machine rewards. **The Rearview Mirror** (activity, your AI). AI makes more of what you already sell, faster and cheaper. The buyer’s AI filters the volume and the unit price follows it down. Most content syndication, contact data, and outbound-volume businesses sit here today. Value destroyed. **Margin Without a Moat** (outcomes, your AI). You’re paid on results and AI cut your cost to deliver them. The margin is real. The risk is that the shortlist is built before you see the buyer, and efficiency doesn’t get you on it. Value enhanced, for now. **Right Road, Wrong Meter** (activity, your buyer’s AI). You show up where buyers’ agents look: verified data, cited research, trusted reviews. You still bill by the lead, the impression, or the seat, so someone else captures the value you create. Value created, not captured. **The Windshield** (outcomes, your buyer’s AI). Buyers’ agents rely on your data, proof, and judgment, and you’re paid on the result. Valuation premiums are moving here. Value compounding. Most businesses will find product lines in more than one quadrant. The same company can be creating and destroying value at the same time, and the P&L will blend the two until the market separates them for you. What to do with the next AI dollar **If you’re running one of these businesses,** split your AI budget into two columns: money that makes your own team more productive, and money that makes you findable, credible, and verifiable to the buyer’s AI. If the first column is most of it, you’re looking in the mirror. Then put a name next to the second column, and a shortlist number next to the name. **If you’re funding one,** ask two questions before you look at the AI slide. What unit is on the invoice? Whose AI does the product serve? AI efficiency on an activity unit is margin on a shrinking base. Price it that way. **If you’re building one,** move the meter before you scale the product. The step from activity pricing to outcome pricing gets harder with every customer you sign on the old unit. **If you’re starting one,** start on the buyer’s side of the machine: proprietary data their agents need and can’t scrape, proof they can verify, and a price tied to the result. In January I said progress starts by answering three questions honestly. The buyer’s AI isn’t waiting for your answer. It’s giving its own, one shortlist at a time. *The views expressed in Uphoff on Media are entirely my own. They don’t represent the opinions of any company I’ve led, any board I’ve sat on, or any investor who’s had the pleasure of debating strategy with me over the years. If something I write here sounds brilliant, I’ll take full credit. If it turns out to be wrong, I was clearly misquoted by myself.* ***“Uphoff on Media”* is published by Tony Uphoff, Founder and Managing Partner of** *Uphoff Advisor****y,* a strategic advisory practice for founders, CEOs, and investors in B2B information, marketing, and technology. The businesses that drive business.** This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit tonyuphoff.substack.com