The aggregate numbers on AI are flat. The people doing the actual work tell you a different story, and they're both telling the truth.
McKinsey, State of AI 2025
That number lands hard if you've been writing checks for two years. Boards are starting to ask the obvious question, and most leadership teams don't have a clean answer.
The gap isn't fake. The productivity is real, the spend is real, the financial signal at the top is genuinely missing. Companies are pouring money into AI and the income statement looks the same as it did before. Meanwhile, your best operator just rebuilt a two-week deliverable in an afternoon and went home at four.
Both of those things happened. The interesting question is where the difference went.
The gains are landing on individuals, not on the company
When an analyst saves six hours on a Tuesday, those six hours don't show up on the EBIT line. They show up as a quieter evening, a side project, a longer lunch, or a deliverable that lands at the same time it always did because the deadline never moved. The personal upside is enormous. The company captures none of it.
This isn't malice or laziness. It's the predictable behaviour of a workforce whose deadlines, budgets, headcount plans, and definitions of "a good week" were all set when the work took longer. Nothing in the way the company runs has been reset to match what's now possible, so the slack pools at the individual level and stays there.
The companies pulling ahead are redesigning the work itself
McKinsey's 2025 data shows the AI high performers, the small group actually moving EBIT, are roughly three times more likely than peers to fundamentally redesign workflows around AI. They're not the companies with the biggest model budgets or the flashiest pilots. They're the ones who looked at a process, asked what it should look like now, and rebuilt it from the answer backwards.
That redesign work is unglamorous and it's where the value lives. A few patterns show up repeatedly in the companies doing it well:
- Deadlines get reset before the tools roll out. If a report used to take a week and now takes a day, the new baseline is a day, with the freed capacity routed deliberately rather than absorbed quietly.
- Hiring plans get rewritten, not just paused. The question shifts from "do we still need this hire" to "what shape of team does this work need now," which usually changes both the headcount and the roles.
- Recurring problems get cut at the root. The same fire that flared every quarter for three years stops flaring, because the underlying process was rebuilt, not patched faster.
- Decisions sharpen because the loop tightens. Leaders get the answer in the meeting instead of the week after, and the quality of choices changes accordingly.
Budgets are going up, which makes the gap more expensive
NVIDIA's 2026 State of AI report has 86% of companies planning to increase AI budgets next year. That tells you something useful. The teams writing the checks are not giving up on the technology. They believe the value is real even when their own P&L doesn't show it yet.
The risk is that another budget cycle goes by with the same shape of result. More licences, more pilots, more enthusiastic individual users, and an EBIT line that still hasn't moved because nothing about how the company actually operates was changed to convert that capacity into financial outcomes. The longer this runs, the harder the conversation with the board becomes.
What gets captured is what gets redesigned
The companies who'll look back on this period as a turning point aren't the ones who deployed the most tools. They're the ones who treated the technology as a reason to revisit assumptions about how long things should take, how many people they should require, and what the cadence of decisions should look like. The tools were the prompt for the work, not the work itself.
This is the part most leadership teams haven't started yet. The pilots are running, the licences are bought, the enthusiasm at the individual level is genuine, and the operating model underneath has not been touched. Until it is, the gains will keep landing where they're landing now.
The uncomfortable question
When was the last time you reset a deadline, a budget, or a hiring plan based on what your best operators can now do with AI?
If the answer is "we haven't yet," that's where the missing EBIT is. The productivity already arrived in the building. The question is whether the company is set up to catch it, or whether it's going to keep flowing past the financial statements and into everyone's personal time for another year.
The teams that move on this in the next twelve months will look very different from the ones that don't, and the gap between them will be the story of 2027.