Boards have stopped asking whether companies are experimenting with AI and started asking for receipts. Meanwhile, most senior leaders don’t have much to show. 

Why? Because even when AI tools are working, there’s a critical difference between generating productivity and generating profit. Many organizations are achieving the former, but few have done what’s necessary to convert it into the latter.

This isn’t a technology problem. Many AI use cases are driving real efficiency gains, and the people using them don’t want to go back. But value disappears somewhere between “our teams are more productive” and “our CFO sees it in the numbers.” Understanding why this gap exists and what to do about it is the important AI conversation enterprise leaders need to be having.

The gap between activity and impact

Across large enterprises, the most common AI deployment looks something like this: Microsoft Copilot, ChatGPT, or a similar chat tool has been rolled out broadly, employees have access, and there’s genuine enthusiasm at the individual level. These enterprise AI rollouts are often focused on driving productivity. In fact, in a recent Catalant survey, consultants reported that increasing individual employee productivity and efficiency was the top value driver motivating clients’ AI initiatives. 

Driving productivity or efficiency does not inherently translate to meaningful change in how the business performs financially. 

An early AI initiative from a Fortune 50 financial services company illustrates the problem. They were among the first major enterprises to deploy Copilot at scale, with Microsoft investing consulting dollars alongside them to support the rollout. Leadership set an ambitious target: every function was expected to identify +10% efficiency gains within a year, to be reflected in the following year’s budget. 

Only one area actually hit the target. Ironically, it was the technology organization.

What made technology different wasn’t just that its employees were more sophisticated AI tool users, though that did play a part. The bigger factor was how the tools were deployed. Rather than giving individuals a chat window and hoping for the best, the technology team integrated AI holistically across an entire workflow, from product managers developing specs to developers writing code to QA teams running tests. Each handoff was automated, and the productivity impact wasn’t in any single step; it was in the aggregate, across the full development lifecycle. 

This distinction matters enormously. A chat tool in isolation is useful, but it is not, in most contexts, a P&L mover. When AI gets embedded across a connected workflow with automated handoffs between people and functions, that’s where numbers start to move.

But most large enterprises haven’t gotten there yet. Instead, they remain stuck celebrating localized time savings without a mechanism to capture them in dollars. Enterprise leaders are mistaking individual momentum for corporate margin, and that is where the value disappears.

Why productivity isn’t the same as profit

Efficiency gains don’t automatically flow to the bottom line. They get donated to customers in the form of lower prices that the market eventually comes to expect, or they get absorbed back into the organization without changing what anyone is accountable for producing.

This is an old problem. The same dynamic played out when CRM platforms hit the market in the early 2000s. The promise was simple: give sales reps better tools, reduce their administrative burden, free up their time, and watch deal flow increase. Companies spent millions of dollars on CRM implementations expecting that outcome, and many were disappointed. When sales reps hit their targets, they stopped pushing, and the freed-up capacity was rarely redirected productively. Without changing the incentive structure, activity expectations, and measures of success, efficiency couldn’t turn into revenue. 

The same failure mode is playing out with AI today, just in different functions and at a larger scale.

Consider a professional services firm running M&A due diligence. That work has traditionally been priced for time and materials: a team of five people, for four weeks, billed accordingly. Now, with the right AI tools, a team of two can do the same work in two weeks, a significant productivity gain. If the firm keeps billing for time and materials, they’ve just donated the gain to the client. Efficiency showed up in the work but didn’t impact margins. 

A company in the same space that’s acting strategically is moving to fixed-fee pricing for that work. Yes, the top-line revenue per engagement looks smaller, but the margin is dramatically better. And because they can now deliver the same quality at a lower price point than competitors that are still running on time-and-materials pricing models, they’re taking market share. This is how the productivity gain becomes a structural competitive advantage rather than a gift to the client.

This move to reinvent the business model to capture what AI makes possible is what separates the organizations seeing P&L impact from those that aren’t.

Some of the clearest examples come from companies applying AI to work that would have been cost-prohibitive for humans. 

Look at Salesforce. Their sales team had a pipeline problem: qualified leads were going cold simply because there weren’t enough people to nurture them over time. AI solves this problem by deploying agents to handle follow-up continuously, across every lead, indefinitely. Revenue per employee goes up because AI unlocked capacity for work that simply wasn’t getting done before.

Freight brokerage offers a less obvious example. Flexport, a broker connecting customer shipping needs to available truck capacity has always faced a simple operational constraint — independent truck drivers are on the road, not looking at screens. Reaching them proactively, at scale, is cost-prohibitive through human labor. AI agents can now make those calls, and brokers can find capacity faster while driving greater volume. The economics of the business improve in a way that wasn’t accessible before.

These examples share the same underlying logic. Instead of looking to execute an existing process more cheaply, they sought to enable a capability that was previously impossible. 

What leaders need to do differently

Most enterprises have spent the last few years running a version of the same strategy: put tools in people’s hands, encourage experimentation, and see what emerges. That strategy builds important skills, surfaces new ideas, and creates organizational familiarity with AI. All of those outcomes are valuable. But they won’t move the P&L, and at this point, boards are no longer satisfied with experimentation as the answer.

What’s needed now is a different kind of question: “Which of our profit pools are most vulnerable to AI disruption, and which are we positioned to expand?”

Those are questions leaders need to answer, and answering them requires enough AI fluency to know what’s actually possible.

From there, the path is more focused than most leaders may expect. Identify one or two profit pools where AI can either protect or expand margin. Determine what business model decisions need to accompany the technology investment (pricing, incentives, organizational structure, go-to-market approach). Then drive those initiatives from the top down, with real accountability, rather than hoping bottom-up experimentation eventually connects to financial outcomes.

The combination of both is what works. Broad skill-building across the organization creates organizational muscle. Senior-led, profit-focused initiatives create the business model change that makes the muscle matter. 

The window of opportunity is open

As leaders are struggling to answer questions about what they have to show for AI investments, many boards already have an eye on the next question to come: “How durable is our business model as these tools keep improving?”

When an AI capability release causes a major competitor to lose meaningful market share, it’s not a technology story. That’s a business model story, and every board in that industry is paying attention. 

If you feel behind, ignore the headlines. Most companies are behind, even if they consider themselves AI forward, and the window of opportunity is still open. Many of the highly publicized layoffs attributed to AI, when examined closely, are actually about cost structure and capital efficiency, not about AI being deployed at scale. The evidence that any large organization has truly operationalized AI across its enterprise in a way that shows up reliably in financial performance is thin, and the game of AI-at-scale is just beginning.

That means there’s still time to get on offense. It’s time to make the business model decisions many organizations have been deferring. Productivity without profit is just an interesting experiment. The leaders who close the gap are the ones who will have something real to show.