The ROI gap is a platform count problem
Wharton Human-AI Research, working with GBK Collective, surveyed more than 1,400 enterprise leaders and employees for its fourth annual report on enterprise AI, and the headline finding should reorder a lot of 2027 planning decks. Half of the companies that had standardized on one approved, dedicated AI platform reported measurable ROI. Among companies still experimenting across multiple platforms, only 9 percent said the same. That gap is large enough to decide whether a program survives finance scrutiny or gets quietly cut next budget cycle.
The implication cuts against a popular instinct in procurement, which is to run several AI tools in parallel to hedge against picking the wrong vendor. The survey suggests that hedging strategy is itself the cost. Platform sprawl multiplies integration work, dilutes the data each tool needs to get good, and spreads training and change management thin across teams who end up mastering none of it. Consolidation, not optionality, is what correlates with return in this data set.
A title without a mandate
The report also surfaces a structural problem with how enterprises are organizing around AI. Nearly 11 percent of Fortune 500 and S&P 500 companies have appointed a chief AI officer or an equivalent role, according to an August figure the report cites, and companies including Boeing, Eli Lilly, and McDonald's are among them. An IBM survey from May 2026 of 2,000 CEOs found real momentum behind these hybrid C-suite AI titles as well.
But Jeremy Korst, a Wharton lecturer and CEO of Mindspan Labs who co-authored the report, is blunt about what that title tends to lack. He describes the typical chief AI officer as someone who "has a title, not a success mandate," appointed to signal seriousness about AI without the budget authority or organizational reach to actually redesign how work gets done. A title answers a board's question about who owns AI. It does not answer the harder question of who can change a workflow.
Where the returns actually come from
Korst's central argument is that ROI follows redesign, not deployment. Buying a platform and layering it on top of an unchanged process produces exactly the kind of underwhelming pilot results that have made CIOs cautious heading into 2027 budget conversations. The returns in the Wharton data show up at companies that used AI adoption as a forcing function to rebuild how the work actually happens, rewriting approval chains, staffing ratios, and handoffs rather than asking a model to quietly sit on top of a process nobody revisited. That is a heavier lift than a software rollout, and it is the lift the survey says actually pays.
That distinction matters because it reframes what a chief AI officer, or any executive sponsoring AI at scale, should be measured on. Korst's framing implies the right mandate is closer to a transformation leader with authority over process change than a technology evangelist reporting on adoption metrics. Enterprises that hire for the second job and expect the first job's results will keep landing in the 9 percent column, no matter how capable the platform they eventually choose turns out to be.
What CIOs should take from the platform data
For CIOs weighing whether to consolidate AI tooling before the next budget cycle, the survey offers a specific, defensible number to bring into that conversation: a fivefold difference in reported ROI between single platform and multi-platform approaches. That is a stronger argument for standardization than most internal pitches get to make, and it comes from outside the organization, which tends to carry more weight with finance than an internal recommendation alone. It also gives a CIO cover to say no to a well-intentioned pilot that would add a fourth or fifth tool to an already fragmented stack, because the external data backs the harder, less popular call.
It also gives CIOs a reason to push back on business units that want to bring in a fourth or fifth AI tool because it tested well in a demo. The sharper question is whether adding that tool moves the organization closer to, or further from, the platform discipline this data associates with actually getting paid back, regardless of how well it performed in a demo. Framed that way, a procurement debate stops being a feature checklist exercise and becomes a question of whether the enterprise can still name, a year from now, which platform did the work and which ones quietly got abandoned halfway through rollout.
The chief AI officer question enterprises still have to answer
The growth in chief AI officer appointments, from a low base to nearly 11 percent of the largest public companies in roughly a year, suggests boards have accepted that AI needs an executive owner. What the Wharton report makes clear is that accepting the need for the role and designing the role correctly are two different exercises, and most companies have only done the first one. The IBM survey of 2,000 CEOs pointing to growing interest in hybrid AI titles suggests this pattern is about to repeat at a much larger scale, across industries that have not yet worked out what the job is actually for.
A chief AI officer with genuine authority over process redesign, budget, and cross-functional priority setting looks like a real structural bet. One appointed mainly to field board questions and coordinate vendor relationships looks like the title Korst is warning about. CIOs sitting next to this role, or being asked to take it on themselves, should treat the mandate question as the one worth resolving before the org chart gets finalized. Boeing, Eli Lilly, and McDonald's have already made the appointment; the harder work, and the one the survey suggests most companies still owe themselves, is writing down what that person is actually allowed to change.
