Your CEO Is Graded on AI. Hours Saved Won't Save Either of You.

Wharton and GBK Collective published The Adaptive Advantage on Tuesday, their fourth annual study of AI at large US companies. 1,416 respondents, leaders and staff. Most of the coverage went to the headline, that 80 percent of leaders see positive returns and only 28 percent call them strongly positive.

The table I'd hand every CPO is further in. Product/Engineering departments use agents about as much as anyone, 62 percent against 68 for the rest. But only 16 percent have AI built into their core processes. That's the lowest of every department tested. Eight percent have transformed their workflows. Nineteen percent report strongly positive ROI, tied with Legal for last.

The team that builds the product is the one where AI shows up least in how the work gets done.

Your CEO has read that, or someone on the board will forward it. This is part 2 of The CPO Transition. Part 1 argued that the CPO stops owning the roadmap and starts owning the operating model. This one is about the person that change has to be explained to first.

The short version

The AI story is the CEO's story, and CEOs are being graded on it: in BCG's survey of CEOs and boards, CEOs said about a third of their performance review rides on AI returns. Most CPOs hand the CEO one of two stories, a roadmap of AI features with dates or a count of hours saved, and the Wharton and GBK data says both are weak. Hours saved is the most common AI measure and net-new revenue the least, yet departments tracking a growth measure were nearly three times as likely to report strongly positive returns. The story a CPO owes the CEO is one sentence per AI bet: the customer outcome it moves, the evidence, the cost per outcome, and what got stopped to fund it. The CEO should hold the CPO to exactly that, plus an honest account of how product's own work has changed. If product doesn't supply the story, the CEO hires a chief AI officer to tell it.

The CEO is the one being graded

BCG's Split Decisions survey, published in May, asked 351 CEOs and 274 board members at companies over $100 million in revenue how AI is being run. CEOs estimated that 35 percent of their performance evaluation now depends on meeting AI ROI goals. Boards put it at 27. Either number is a lot of a CEO's job riding on one topic.

And they own it personally. 47 percent of CEOs said the CEO is running AI implementation. Fewer than 10 percent of either group thought a chief AI officer should lead AI strategy.

So when a CEO asks the CPO "what's our AI story," it isn't a status request. It's the CEO asking for the material they'll be judged on. I think a lot of CPOs miss this and answer it like a planning question.

Here's what happens when product doesn't supply it. Equilar's Jake Dayton reported on October 8 that median chief AI officer pay hit $1.9 million this year, up from $1.6 million, and that 60 percent of current chief AI officers were appointed in 2025 or 2026. A third of them report straight to the CEO. I read that as CEOs buying the AI story from a new seat because they couldn't get it from the existing ones. Some of those hires make sense. But a CEO who adds an AI executive between themselves and product has told you where they think the story isn't coming from.

Two stories that don't hold up

The first is the roadmap story. "We ship agent features in Q1, the copilot in Q2, pricing changes in Q3." It's familiar, it has dates, and it's a feature parade. Boards have learned to discount it, which is why the board deck product section I use has no feature list in it. Part 1 covered why the roadmap stopped steering. It also stopped persuading.

The second is worse because it sounds rigorous: hours saved. Wharton found it's the most common way departments measure AI, used by 40 percent of them. Net-new revenue is the least used major measure, at 27 percent. And the departments that tracked at least one growth measure (revenue, conversion, output, customer satisfaction) reported strongly positive returns 35 percent of the time, against 13 percent for those measuring efficiency alone. The data is self-reported and correlational, and the authors say so. I'd still act on it.

Product has its own version of hours saved. Velocity. PRs merged. Cycle time. Prototypes per week. Those are hours saved in engineering clothes, and they tell the CEO that product got faster without saying whether anything a customer cares about moved. Stefano Puntoni, who co-directs the Wharton research, said it in the release: "What separates the leaders is whether they are changing the business around AI enough to generate meaningful value."

There's also a market reason the efficiency story is getting weaker. Patrick Collison argued last week, as Rich Holmes quoted him in Department of Product, that personal agents "will act as a kind of structural subsidy for product quality." If buyers' agents do the research, the product that's actually better wins more often. That's a quality story. A CEO can tell a board that one.

The story a CPO owes the CEO

One sentence per AI bet. Four parts.

The customer outcome it moves. Not the feature, the outcome: fewer escalations, faster time to first value, more renewals in a segment.

The evidence that it's moving, from the listening loop and the eval bar, not from a launch announcement.

What it costs per outcome. This part used to be optional. It isn't now. Wharton found 64 percent of managers and above agree that rising token, license, and API costs are a major barrier to using AI more, and 74 percent of VPs and up. Dev Ittycheria, a week into his return as MongoDB's CEO, said on Jason Lemkin's 20VC x SaaStr episode that his conversations with CIOs are about token costs, token budgets, and who gets the data. Your customers' CIOs are asking your CEO the same thing. Gross Margin Is Your Job Now has the mechanics.

What got stopped to fund it. A story with no trade-off isn't a strategy, and the CEO knows it.

Then the part CPOs skip: product's own house. That 16 percent number is about product teams, not the product. A CEO will eventually ask how product's own work has changed, and a list of licensed tools isn't an answer. The answer is the loop from part 1, who decides, on what evidence, against what bar. The handbook version is The AI Product Operating Model, and The Investor and Board Narrative for AI-Era SaaS turns all of this into the standing board update.

What the CEO should hold the CPO to

If you're the CEO, this is the list. It's short on purpose.

1. A growth measure named before each AI bet ships. If the measure is hours saved, velocity, or adoption, send it back.
2. Cost per outcome, reported next to the outcome, every time. Not in a separate finance review.
3. A straight answer on how product's own work has changed in the last two quarters. Who stopped doing what.
4. One bet killed every quarter, with the reason. The kill-the-roadmap muscle is the same one.

And stop asking for three things: dates on AI features, counts of AI features, and tool adoption rates. Each of those rewards the CPO for the wrong story.

One more thing for both sides. BCG found about six in ten CEOs think their boards are rushing AI transformation. Judith Wallenstein, who runs BCG's CEO advisory practice, said CEOs "need to be very intentional about supporting their boards on the same learning journey they've taken." The CPO is the person who should be handing the CEO the material for that. A board that's rushing slows down when it sees outcomes and costs. It speeds up when it sees a feature list, because a list always looks short. If you work for a founder CEO, the dynamics are their own, and I wrote about them in Founders Don't Want a CPO.

Next in the series: the CPO and the CTO, and who owns the substrate, the eval harness, and the deploy guardrails.

One thing to try this week

List every AI bet your org has in flight. Next to each one, write the measure you'd report to the board today. Circle every line where that measure is hours saved, velocity, or adoption. Rewrite each circled line as one sentence: the customer outcome, the evidence, the cost per outcome, and what you stopped to fund it. Then walk your CEO through the sheet before the next board prep, not after.

Sources: Wharton Human-AI Research and GBK Collective (Jeremy Korst, Stefano Puntoni, Prasanna Tambe), "The Adaptive Advantage: Who's Winning with AI and How," October 13, 2026, ai.wharton.upenn.edu, and the release, Business Wire, October 13, 2026, businesswire.com. BCG, "CEOs and Boards Are Aligned on AI in Theory, but Divided in Practice," Split Decisions: The BCG CEOs and Boards Survey, May 2026, bcg.com, with quotes from Judith Wallenstein in the BCG release, PR Newswire, May 4, 2026, prnewswire.com. Jake Dayton, "Where Does the Chief AI Officer Stand in 2026?," Equilar, October 8, 2026, equilar.com. Rich Holmes, "Stripe's CEO says Agents will reward Product Quality," Department of Product, October 11, 2026, departmentofproduct.substack.com. Jason Lemkin, "20VC x SaaStr with MongoDB's Dev Ittycheria," SaaStr, October 8, 2026, saastr.com.

Part of the running argument on Product Leadership: the product org chart was calibrated to a build cost that collapsed, and the CPO's job is now the system that decides what gets built and proves it paid, which is the story the CEO has to tell.

Related answer: What should a CEO hold a CPO to on AI?