A 15% price cut isn’t supposed to be a victory lap. In most B2B rooms, it’s treated like a concession—something done when demand is soft, competition is brutal, or the product lost its edge.
And yet Jason Lemkin wrote that an internal “AI VP of Marketing” named “10K” recommended exactly that: drop ticket pricing 15%, expect attendance to rise 40%. Then the test ran. Lemkin reported that SaaStr AI Annual 2026 attendance was up 41% year over year, with ticket revenue still up 6%.
That’s the first thread worth holding onto: the outcome wasn’t “more volume, less money.” It was more people and more revenue. Not the usual trade.
The second thread is more uncomfortable. Lemkin also wrote that his team had “10+ years” of spreadsheets, analysis, and lived experience telling them price wasn’t the lever—especially for last-minute buyers who “didn’t seem to care what the price was” as long as it stayed below a mental threshold of “feels like a lot.” That’s a familiar kind of confidence. Data-backed. Repeated. Settled.
Then an AI model looked at the same system and said: you’re wrong.
Why pricing is suddenly a demand gen problem (not just a finance one)
Pricing used to sit behind glass. Finance owned it. Product had opinions. Marketing mostly worked around it: position the offer, explain value, route objections, keep the pipeline moving.
That wall is breaking in 2026 because AI has made pricing a live system. The Research Brief points to AI-driven pricing optimization seeing significant adoption in 2023, particularly in retail and e-commerce, where dynamic pricing is normal and the feedback loop is fast. In that same 2023 context, some cases reported revenue increases of up to 25% from AI-driven dynamic pricing.
Yes, that’s not B2B SaaS. But the implication matters for Maya Chen—the VP of Demand Gen who treats marketing as a revenue function. If pricing can be tuned with the same rigor as creative, targeting, and landing pages, it becomes part of the growth stack. And once it’s part of the growth stack, the demand team can’t pretend it’s “someone else’s lever.”
There’s another reason this matters now: AI isn’t a side project anymore. The Research Brief cites a 2023 estimate of the AI in sales and marketing market at USD 20.0B, with a projected 26.7% CAGR. That kind of growth doesn’t happen because teams are playing with toys. It happens because budgets are moving.
The blind spot “10 years of data” can create
Lemkin’s story (and it is a real, attributed account) lands because it exposes a specific failure mode in mature demand gen: teams get good at measuring what they can see, then confuse visibility with truth.
Here’s the crux, in Lemkin’s telling. The human conclusion came from observed buyers—especially late-stage, late-cycle buyers—who looked relatively price-insensitive. The AI’s counterargument was that the pricing model suppressed an entire layer of potential buyers who never converted, never registered, never appeared in the dataset as “price-sensitive” because they disappeared before the funnel could record them.
That’s not mystical. It’s selection bias wearing a spreadsheet.
And it’s exactly the kind of thing AI is decent at surfacing when it’s allowed to model the full system instead of a single stage. The Research Brief notes a reported correlation (r = 0.67) between AI adoption and decision accuracy in a 2023-oriented expert summary. Correlation isn’t causation, and that number shouldn’t be treated like a universal law. But it does point to a plausible mechanism: when AI helps teams test more hypotheses, faster, with fewer sacred assumptions, decision quality can improve.
Still, the part to copy isn’t “trust the machine.” It’s the discipline of asking what isn’t being measured. Who never makes it into the spreadsheet? Which prospects bounce before attribution can label them? Which segment quietly decides “not now” because the offer feels just expensive enough to postpone?
What an “AI VP of Marketing” actually is—and what it isn’t
Lemkin described 10K as an orchestrator, not a content generator: something intended to coordinate campaigns, offers, channels, and day-to-day actions, updating as new data comes in. He also wrote that it used Claude Opus to analyze years of campaign and conversion data and produce a pricing model that challenged the team’s assumption.
That framing lines up with where the broader market has been heading since the 2023 generative AI surge: rapid experimentation for research, writing, editing, and workflow automation—paired with public warnings about quality failures when humans step away entirely. The Research Brief references that tension explicitly: speed went up, but so did the risk of errors when organizations over-relied on AI without editors and safeguards.
So the cleanest definition for demand leaders is simple. An AI “VP” is not authority. It’s throughput. It can propose, simulate, and schedule. It can also be relentlessly literal, which is useful when teams have grown sentimental about old answers.
But it can’t own the brand risk. It can’t take the board meeting. And it can’t be the final signer on a pricing change that reshapes positioning. In fact, the Research Brief’s expert guidance argues for exactly this: combine human judgment with AI for evidence-based strategies that hold up under uncertainty, and implement ethical and brand safeguards.
That’s the operating model worth stealing: AI as the challenger, humans as the governors.
The practical takeaway: treat price like creative—test it, don’t debate it
Most demand teams have a mature A/B muscle for ads and landing pages, and a surprisingly weak one for pricing and packaging. Not because they don’t care. Because pricing changes feel irreversible, political, and existential.
Lemkin’s account shows a different posture: hold a belief for years, let an AI model challenge it, run the experiment, accept the result. The most interesting part isn’t that the AI was “right.” It’s that the organization allowed a test to settle an argument that had calcified into identity.
Also, it reframes discounting. A price cut can be a blunt instrument. Or it can be a targeted way to widen the top of the funnel—especially if the real problem is the invisible layer of buyers who self-select out before they ever become “leads.”
In 2026, with AI-driven decisioning becoming standard (64% of marketers in a 2023-oriented statistic rated AI as very important to marketing success in the next 12 months), the teams that win won’t be the ones with the strongest opinions. They’ll be the ones with the fastest learning loop—and the clearest guardrails.
Lemkin wrote that the biggest thing they underestimated wasn’t automation, but “honesty”—an AI with no ego, no sunk cost, no attachment to the old model. That’s the circle closing: the decade of “proof” wasn’t useless; it was incomplete. The AI didn’t replace judgment. It forced a better question, then let reality answer it.
