Wharton Study: AI Is Everywhere, Strong ROI Is Not

Almost every large organisation now uses AI, but few can show that it pays off strongly, and the gap between leaders and their staff is widening. That is the picture in the fourth annual Wharton Human-AI Research and GBK Collective study, "The Adaptive Advantage: Who's Winning with AI and How," published on 13 October 2026. Its central finding is that returns track management choices, such as what gets measured and who owns AI, more than they track mere adoption.

What the study covers

According to the press release as carried by FinancialContent, the survey sample is 1,416 US enterprise leaders and entry- and mid-level employees; the release does not state fieldwork dates. The Knowledge at Wharton report page describes it as a repeated cross-sectional study, with Wharton directing the research agenda and GBK Collective designing the instrument and sample. It organises findings around three markers of what it calls an "adaptive organization": proving value, transforming workflows, and aligning the organisation. This year it also compares leaders with entry- and mid-level employees for the first time.

Because this is a survey of self-reported outcomes from a particular sample, the figures describe perceptions of ROI, not audited financial returns. The Wharton page itself did not display the sample size or fieldwork dates, so the 1,416 figure comes from the release.

Returns: broad but shallow

The release reports that 80% of enterprise leaders see positive AI returns, but only 28% describe them as strongly positive. The study's most quoted comparison concerns measurement: 35% of departments that track at least one growth measure report strongly positive ROI, against 13% of those tracking efficiency alone, making the first group nearly three times as likely to report strong returns. Hours saved is the most common ROI metric, used by 40% of departments, while net-new revenue is the least-used major measure at 27%. Meanwhile, 52% of leaders say AI's main job over the next 12 to 24 months is productivity or cost reduction, and 22% point to growth.

That pattern is worth interpreting carefully. The data show association, not causation: departments that measure growth may already be more advanced or better resourced, and the survey cannot say whether the choice of metric produces better returns or reflects them. Still, the finding suggests that organisations judging AI only by time saved may be undercounting, or under-pursuing, its upside.

Structure and ownership

The study finds a link between organisational structure and reported returns on AI investment. Among respondents in management positions and above, 58% at organisations with a newly appointed Chief AI Officer report strongly positive returns on investment (ROI), compared with 24% where AI responsibilities were added to an existing executive's role and 8% where no such role exists or is planned. A similar pattern emerges in platform adoption: 50% of organisations that have standardised on a single, approved AI platform report strong ROI, compared with just 9% of those still experimenting with AI.

The Wharton page adds that 67% of leaders whose departments meet all three markers of transformational success report strongly positive ROI, versus 19% of those that do not. According to the release, only 7% of organisations succeed across all three markers, and only about one in five meet all three workflow-transformation measures. Departments with strong returns were also more likely to have changed management structure because of AI (64% versus 47%), according to Wharton, though they are no more likely than others to report staff reductions or hiring freezes.

Agents: used widely, trusted narrowly

The agent numbers show caution. The release says 67% of leaders use AI agents in some form, but only 28% say AI is fully embedded in core processes, and 17% let agents act without a person approving each step. The Wharton page reports that, in departments where agents may take irreversible financial or legal actions, 29% have transformed their workflows, against 13% that give agents more limited responsibilities. It also reports that departments monitoring agents automatically have transformed workflows at 24%, versus 9% that do not. Brian Smith of GBK Collective is quoted in the release as saying most companies sit in a "supervised middle," a description that fits those figures.

Leaders and employees see different companies

The new leader-versus-employee comparison is among the report's more consequential contributions. Per the release, 73% of leaders say AI has significantly changed how their departments work, against 58% of entry- and mid-level employees. Adequate AI training is reported by 78% of leaders and 56% of employees; confidence in daily AI use by 79% and 62%; and a clear long-term AI vision from leadership by 79% and 64%. Wharton's page calls this a "substantial gap" in understanding and following a cohesive AI strategy.

Employees are not negative about the tools themselves. The release says 73% of entry- and mid-level employees report a positive effect on their role and 84% of those who use AI say it saves them time. Daily use among managers and above rose about 16 points to 62%. At the same time, sentiment cooled: "excited" fell 8 points year over year, "cautious" rose 7 points, and "skeptical" rose 8 points, which the release describes as the first year usage and sentiment moved in opposite directions.

Workforce signals

On jobs, 37% of leaders report AI-related staff reductions and 33% report slowed or frozen hiring, per the release. Some 24% expect fewer entry-level hires over the next two to five years, versus 14% who expect fewer executive roles. The same release reports that 82% say AI helps newer employees contribute faster, and that concern about skill atrophy is held by 48% of VP-level and above, 46% of entry- and mid-level employees, and 32% of managers and directors.

How to read it

For market-watchers, the useful takeaway is that adoption is no longer the main variable. Wharton's own framing is that adoption is no longer the question, and the differences that matter lie in measurement, ownership, standardisation and training. Two caveats apply. This is a survey of one national sample with self-reported ROI, so it should be read alongside spending and usage data from other sources. And the study is published by a research centre together with a market-research partner whose own commentary appears in the release; the quoted researchers' views are interpretation, not independent audit. What to watch next is whether the leader-employee gaps in training and strategy narrow in the next wave, and whether the growth-measuring departments hold their lead as more firms add revenue metrics.