The Leverage Gap: What AI Means for How We Pay People

Sales organizations have always linked pay to visible outcomes. Create more value and earn more money. The relationship isn’t perfect, but the connection has generally been clear. Work harder, sell more, get paid more.

AI is starting to pull that relationship apart, but the compensation landscape hasn’t been great about catching up. We think it’s time to start the conversation.

The Leverage Gap

Two employees with similar experience, similar intelligence, and similar roles can now produce dramatically different outcomes depending on how effectively (and willingly) they use technology. The gap between them can become enormous, not because one person tries harder, but because one is more leveraged.

A Harvard and Boston Consulting Group study of 758 consultants put real numbers on this. Given access to the same AI tool, consultants working within the tool’s actual capability zone worked 25 percent faster and produced work rated 32 percent higher in quality. Consultants who used the same tool on tasks outside that zone performed worse than those who used no AI at all. Same tool, same firm, opposite outcomes. 

The same pattern shows up in the numbers at scale. PwC’s 2026 AI Performance Study, based on a survey of over 1,200 executives, found that the top 20 percent of companies capture 74 percent of the economic value AI creates. Most organizations are running similar technology and spending similar budgets. The gap comes from how that leverage gets used, not from who has access to it.

What the Highest Performers Already Knew

The highest performers were rarely the people who worked the longest hours. They knew where to focus and what mattered, and they let everything else fall away. They created more value per hour than everyone around them, long before anyone was using the word leverage to describe it. 

AI only amplifies that principle. The difference between an average performer and an exceptional one may become much larger than what we’ve seen historically, simply because the ceiling on individual output just moved.

This is already showing up in how some companies structure roles and pay. At Lovable, growth leader Elena Verna stepped out of management and back into an individual contributor role. She’s described VP-level candidates wanting that same tradeoff, choosing roles with no direct reports because the pay is tied to what they produce rather than how many people they oversee. It’s an early example of pay tied to what a person builds, and not where they sit on the org chart.

The Fairness Question

Whenever compensation comes up, fairness enters the conversation. It should. Pay isn’t just an economic decision. People compare, evaluate, and draw conclusions about their own value from it.

That gets harder once leverage enters the picture. If one employee produces five times the value of another, most people would say compensation should reflect that. What if they produce ten times the value? What if technology accounts for part of that difference? What if one person embraces new tools while another resists them?
These questions get uncomfortable fast, but we think leaders will need to sit with them.

 

The future may not reward effort the way the past did. It may reward contribution, and those aren’t the same thing.

That distinction hasn’t reached most comp structures yet. A review of roughly 2,500 public company proxy statements in 2026 found that only 58 companies, about 2 percent, have built AI into their formal incentive frameworks. The rest are still measuring performance the way they did before individual leverage started diverging this sharply.

Leverage Doesn’t Always Look Like More Output

It would be simpler if leverage just meant producing more of the same thing, faster. The research says otherwise.

A widely cited study, from Brynjolfsson, Li, and Raymond followed 5,172 customer support agents after AI tools were introduced. Productivity rose 15 percent on average, but the gains weren’t even. Novice and lower-skilled agents improved by roughly 34 percent. The most experienced agents saw only small gains, and in some cases a slight dip in quality. The same tool that closed the gap for newer workers barely moved the needle for the people already at the top. 

Research on customer service teams found a similar pattern. AI didn’t just make agents faster. It automated away the simple tickets and left people handling a higher share of complex, emotionally demanding cases. The hours stayed the same. The difficulty of those hours went up.

Organizations consistently underestimate the overhead that comes with new AI tools, from prompt development to output review to exception handling. For every hour a tool saves, teams often spend more time managing it, at least in the first year.

None of this cancels out the leverage gap. It just complicates what leverage actually measures. Two people using the same tool can end up with very different workloads, different task complexity, and different review burdens, even when their raw output numbers look similar.

Effort vs. Contribution

We don’t believe the future of compensation is simply about paying people more or less. We think it’s about developing a better understanding of contribution itself, and building pay structures that can actually measure it.

What value is someone actually creating? How durable is that value? How much leverage does that person create for the people around them? How much trust do they build, and how much capability do they help others develop? 

These are harder questions than counting output, and they’re also more important, because organizations don’t create value through activity alone. They create it through people who elevate the systems and teams around them.

Hard Work Still Moves the Productivity Needle

Hard work will always matter. But hard work alone is becoming less differentiating, and the data on individual variance backs that up.

The question used to be simple. How hard are you working? The question that’s becoming more important is different. What are you creating?

We don’t have all the answers here. But we think this is a conversation worth having, and we’d like to hear how other compensation practitioners are thinking about it too.