The question of who captures the value AI creates is about to land on every leadership team’s desk. The companies that answer it first, and answer it well, will keep their best people.

For the past two years, most corporate conversations about AI have centered on adoption: which tools to license, which workflows to automate, and how to get employees to actually use them. A second conversation is now starting in one-on-ones, performance reviews, and exit interviews, and it is considerably harder. If an employee uses AI to produce 20% or 30% more output, who should get the upside: the company that paid for the tools, the customer through lower prices, or the person who figured out how to use them well?

Most organizations have not decided. Many have not even realized they will need to.

Physicians are the early warning signal

Medicine offers one of the clearest early reads on where this is heading. The Doximity 2026 Physician Compensation Report, which included a June 2026 survey of more than 1,400 U.S. physicians on AI and pay, found that two-thirds of doctors now use AI daily or weekly in their clinical or administrative work.

The compensation findings are what should get executives’ attention. When asked who should primarily benefit financially if AI allows them to complete more clinical work in the same amount of time, 44% of physicians said doctors should, a figure that climbs to 50% among primary care physicians. More than a third (36%) said pay for a service should change if AI substantially reduces the time or effort it requires, while 43% said it should stay the same. And one in five physicians said they have already faced higher productivity expectations because of AI.

Physicians matter here not because medicine is unique, but because their output is counted. Patients seen, procedures performed, and billable units are tracked precisely, so productivity gains are visible to the people generating them. The same dynamic applies to anyone whose work is measured: sales representatives, attorneys and consultants who bill by the hour or the matter, recruiters, claims processors, customer support agents, buyers, schedulers, and planners. When work is measured, gains are visible. When gains are visible, they become negotiable.

The productivity is real, but it is uneven

The best evidence on AI’s workplace impact confirms that the gains are genuine. In one of the most widely cited field studies, economists Erik Brynjolfsson, Danielle Li, and Lindsey Raymond tracked more than 5,000 customer support agents as a generative AI assistant was rolled out. Their NBER working paper found that issues resolved per hour rose about 14% on average, with a 34% improvement for novice and lower-skilled workers and minimal impact on the most experienced agents. The final version published in the Quarterly Journal of Economics put the average gain at 15% and found that top performers saw small speed gains alongside slight declines in quality.

That finding complicates the pay question considerably. The researchers suggest the AI tool worked partly by spreading the habits of the best agents to everyone else. So whose value is it? If you pay the newer employee for the lift, you may be rewarding the system rather than the person. If you pay no one, the veteran whose know-how was effectively encoded into the tool gets nothing for it. Neither answer is obviously fair, which is exactly why leaders need a deliberate policy rather than a series of one-off negotiations.

There is also the enterprise-level reality that individual time savings do not automatically become profit. An hour saved only creates financial value if it is redeployed into more throughput, better quality, faster cycle times, or new revenue. Many organizations are still working out how to make that translation, and they are understandably reluctant to pay out gains they have not yet banked.

History says the default answer is “the company keeps it”

Past waves of workplace technology offer little encouragement to employees hoping for a raise. Word processors, spreadsheets, email, ERP systems, and mobile devices all made workers dramatically more productive, and the gains flowed largely into margins, lower prices, and shareholder returns rather than paychecks.

The Economic Policy Institute has tracked this divergence for decades. Its analysis estimates that net productivity grew 90.2% from 1979 to 2025 while typical worker pay grew 33%, and that the typical worker would earn roughly $16.40 more per hour if pay had kept pace. EPI attributes most of that gap to rising inequality and a shrinking labor share of income. It is worth noting that some economists dispute how the gap is measured, particularly the choice of inflation adjustments, and EPI’s own decomposition assigns part of the divergence to differences in how producer and consumer prices are calculated. But the broad pattern is not in serious doubt: productivity tools have historically raised output faster than they have raised pay for typical workers.

So why might AI be different? Three reasons stand out. First, AI-driven gains are often measurable at the individual level, not just in aggregate. Second, AI fluency is highly portable: the tools are widely available, so an employee who has mastered them can take that capability to a competitor on Monday. Third, and most importantly, the labor market has already started pricing it in.

The market is paying for AI skill, not AI usage

PwC’s 2026 Global AI Jobs Barometer, which analyzed more than a billion job postings across 27 countries and territories, found that jobs requiring AI skills now carry an average wage premium of 62%, up from 57% a year earlier. The premium varies widely by sector, from 118% in some areas such as consumer markets to 16% in government and public sector work.

PwC also distinguishes between two kinds of AI-exposed roles. In what it calls “professionalized” roles, such as radiologists and recruiters, AI amplifies expert judgment; in “democratized” roles, AI simplifies the work so non-experts can do it. According to coverage of the report, professionalized roles are seeing twice the job growth and 42% faster salary growth, and the companies most exposed to AI are growing wages faster than the least exposed (24% versus 17%).

The physician data points the same direction. In the Doximity survey, 67% of doctors said colleagues who stay current with AI tools will have a meaningful earnings advantage over those who don’t, and at least 39% said AI proficiency already factors into hiring and promotion decisions in their specialty.

The distinction matters for how leaders think about reward. Markets have always paid a premium for scarce, relevant capability. They have rarely paid people simply for using a tool everyone else also has. The employee who redesigns a workflow, applies judgment to AI output, catches its errors, and teaches others to do the same is creating value that is hard to replicate. The employee who simply turns on a copilot is not. Companies that fail to recognize the first group will not keep them for long. Doximity found that three-quarters of physicians are open to relocating for the right opportunity, a reminder that measurable productivity is also a very portable bargaining chip.

The hidden cost of saying nothing

Many leadership teams are tempted to stay quiet, raise targets as productivity improves, and hope the question never becomes formal. That approach carries real risk.

When people see that higher output only leads to higher quotas, they adapt. They finish in five hours and look busy for eight. They stop sharing the prompts, templates, and workflows that made them faster. Call it shadow productivity: gains that exist but never reach the organization. This is precisely the wrong outcome, because the research suggests much of AI’s value comes from spreading best practices across a team, not from a few individuals hoarding them.

There is also the burnout risk of ratcheting expectations on an already stretched workforce. Doximity reports that 82% of physicians describe themselves as overworked and that 44% say they are somewhat or very likely to leave clinical practice within two years. AI did not cause that strain, but it illustrates what happens when productivity demands keep rising without a corresponding sense of fairness.

Interestingly, the same data shows reward does not have to mean cash alone. About three in four physicians said they would accept lower pay in exchange for more autonomy or better work-life balance. For many employees, a share of AI’s gains delivered as time, flexibility, or growth opportunity may be as valuable as a bonus.

A practical playbook for leaders

Measure before you reward. Establish a clear baseline for output, quality, and cycle time before and after AI deployment. Without it, you cannot separate gains driven by the tool from gains driven by process changes or individual ingenuity, and every pay conversation becomes a matter of opinion.

Decide your philosophy explicitly. There is no single right answer, but there are several defensible ones. A company may retain the gains and reinvest them in growth, job security, or tools, and simply explain why. It may share a portion through team-based gainsharing, an approach with a long track record in manufacturing through plans such as Scanlon, Rucker, and Improshare that tie payouts to measured improvements. It may build skill-based pay and career paths that reward demonstrated AI capability. Or it may offer a “time dividend” in the form of flexibility, learning time, or reduced hours. What matters is choosing deliberately and applying the choice consistently.

Reward contribution to the system, not just personal output. The employees who build reusable workflows, document best practices, and coach colleagues multiply AI’s value across the organization. Recognize them explicitly, or they will stop doing it.

Don’t silently ratchet targets. If performance expectations are going to rise, say so, explain the rationale, and pair the change with something employees value.

Benchmark externally. Your competitors are figuring out their answer too. Know what the market is paying for AI-fluent talent in your industry and roles before your best people learn it from a recruiter.

Communicate before you are asked. Employees will eventually expect a logical explanation of how AI-driven value is shared. Leaders who articulate a clear position early will be seen as fair. Those who wait will be seen as defensive.