Only 2.6% of companies tie executive compensation to AI performance - and most of those are doing it wrong, according to new research from compensation consultancy Pearl Meyer.
The firm reviewed proxy filings from 2,500 companies and found just 65 disclosed incorporating AI into executive incentive plans. Of those, only 14% used an explicit AI metric. The rest folded AI into broader strategic goals or individual executive assessments.
That gap matters because the way a company ties pay to AI reveals what it actually expects AI to do for the business. The research, based on an analysis of AI talent markets and executive incentive practices, connects three questions boards are grappling with: how AI investments create value, which talent is required to deliver it, and whether executive incentives reinforce that work.
Three ways companies create value with AI
The researchers identified three distinct value-creation models. Frontier AI creators build foundational models, platforms and infrastructure. They rely on scarce researchers and technical leaders whose work can materially affect enterprise value, and may rationally concentrate extraordinary compensation on a small number of critical people.
AI product builders apply AI to products, services and customer experiences. They need people who can translate technical capability into commercially useful solutions, and may emphasize development milestones, customer adoption or commercialization.
AI-enabled operators use AI to improve the enterprise itself through productivity, efficiency, workforce leverage and decision-making. These companies are more likely to focus on deployment, workforce adoption, process improvement and, eventually, measurable productivity or cost outcomes.
This framework gives directors a more useful starting point than asking whether the company is using AI, the authors said.
What early adopters are measuring
Among the 65 companies with AI-related incentives, 57% embedded AI within broader goals such as technology deployment, transformation, efficiency or governance. Another 29% used qualitative or individual assessments covering leadership, adoption or responsible use. Only 14% used defined AI metrics with explicit adoption, utilization or performance objectives.
The more revealing finding emerges when those companies are viewed through the three value-creation models. AI-enabled operators are emerging as the main source of formal AI incentive measures, with disclosures emphasizing transformation, adoption and efficiency. Executives are being held accountable for enterprise-wide change rather than AI-specific revenue.
The research also found that roughly 60% of companies with AI-related incentive objectives have initiatives that could reasonably be connected to identifiable financial or operating outcomes. "They may not measure AI return on investment today, but this suggests that, as AI strategies mature, many incentive measures could evolve from adoption and implementation toward measurable business impact," the authors wrote.
From activity to value
That evolution - from implementation to outcomes - is where incentive design gets difficult. Boards should first consider whether AI's benefits are already captured elsewhere in the plan. If AI improves revenue, margins or earnings, a separate measure may reward the same outcome twice.
A targeted measure may be useful when the organization must build capabilities today to produce financial value later. AI performance measures may need to evolve with the strategy itself: from capability building to adoption to business impact and, ultimately, to measurable financial outcomes.
The challenge is that implementation does not necessarily demonstrate impact. "A company can deploy an AI platform without employees using it meaningfully, or increase utilization without producing better customer outcomes, lower costs or greater productivity," the authors wrote. Measures that appropriately reward early capability-building may become less useful as the strategy matures.
The risk is allowing an incentive to continue rewarding activity after the organization should be demonstrating impact.
Five questions every board should ask
The research suggests compensation committees should understand how their company is actually using AI and whether incentive design supports the strategy before adding AI-specific metrics. Directors can use these questions to guide the discussion:
- How is AI expected to create value for this company?
- Who is accountable for delivering that value?
- Are we rewarding capability-building, adoption or measurable business impact?
- Can AI-related performance be measured credibly?
- Are AI-related results already captured through existing financial, operational or strategic metrics?
AI should not appear in an executive incentive plan simply because it dominates boardroom discussions or is popular in the market. The decision must follow from a clear view of how AI will create value, who is accountable for delivering it and whether current measures already capture the expected results.
Why this matters for executives and strategy leaders
For senior leaders accountable for AI strategy, the research offers a practical framework for one of the hardest conversations in the boardroom: what does AI actually need to deliver, and who gets rewarded for it? The three value-creation models - frontier creator, product builder, enabled operator - provide a vocabulary for aligning AI investment with incentive design. Executives responsible for AI strategy and implementation can use the five questions to pressure-test whether their company's compensation structure rewards real business impact or merely activity. The distinction matters: companies that tie incentives to adoption metrics may find those measures become obsolete precisely as AI starts delivering measurable value.
Your membership also unlocks:
