Your employees are already using AI. The question is who keeps the savings.

Your office admin figured out AI three months ago. She didn’t tell you. She didn’t ask permission.

She just started using it. Meeting scheduling that used to eat 45 minutes a day — done in five. Internal memos she used to draft by hand — generated in seconds, lightly edited, sent. Supplier emails that arrive in German? Translated and summarized before she’s finished her coffee. That shared drive nobody could navigate? She had AI reorganize and re-label the whole thing over a long weekend.

From your perspective, she’s gotten really good at her job. Faster, fewer mistakes, never seems overwhelmed anymore. You might even be thinking she deserves a raise.

Here’s what you don’t see: she’s doing about 3 hours of actual work per day. The rest is Netflix on her phone, a side hustle, or polishing her resume — because she knows an AI-savvy office admin is worth a premium right now.

The productivity gain is real. You’re just not the one capturing it.

And she’s not the only one. Your parts counter guy at the auto repair shop is using AI to look up compatible parts and draft quotes — he handles twice the customers but clocks the same hours. Your head chef plugs supplier prices and inventory levels into AI to plan menus — what used to take Sunday afternoon now takes 15 minutes.

This isn’t a few tech-savvy employees experimenting. It’s a pattern. And it has a name.

There’s a window. It’s closing.

Josh Steimle calls it the “AI arbitrage window” — the brief period where early adopters enjoy outsized returns from a new technology before everyone else catches up. He points out that AI coding tools collapsed the cost of building software by as much as 90% in two years. The people who figured that out first captured the difference. But arbitrage windows close. They always do.

We’ve seen this movie before.

When VisiCalc — the first spreadsheet program — shipped in 1979, the accountants who learned it could do in an afternoon what used to take a week. For a few years, they were wizards. Clients paid a premium for speed and accuracy that seemed almost magical. Then everyone learned spreadsheets. Within a decade, knowing Excel wasn’t a competitive advantage — it was a line item on every job listing. The skill went from superpower to minimum requirement.

Or take ATMs. When banks started installing them in the 1980s, everyone assumed tellers were finished. The opposite happened. Teller employment actually rose — from 500,000 to roughly 550,000 over 30 years. ATMs made branches cheaper to operate, so banks opened more of them. But the job changed completely. Tellers stopped counting cash and started selling financial products. The people who adapted thrived. The ones who couldn’t do the new job didn’t.

Michael Porter and Victor Millar wrote about this pattern back in 1985: information technology changes how companies create their products, and the advantages it creates are real but temporary. What’s differentiating today becomes table stakes tomorrow.

AI is following the exact same arc. Faster.

The gains are real. The question is who gets them.

Here’s where it gets uncomfortable for business owners.

Right now, individual employees are capturing most of the AI productivity surplus. They work faster, produce the same output, and pocket the difference as leisure or leverage. That’s the current reality.

But companies are already doing the math. A ResumeBuilder survey of 866 U.S. business leaders found that 54% plan to cut employee compensation for roles where AI boosts productivity. Another 26% plan layoffs specifically to fund AI investment. As their chief career advisor Stacie Haller put it: “Companies are making a clear calculation: AI investment is the priority, and employee compensation is where the budget will come from.”

This isn’t speculation. It’s stated corporate strategy.

Academic research backs this up. A recent paper titled “When AI Captures the Surplus” found that regions with higher AI-patenting intensity see measurable declines in labor’s share of income. AI functions as a capital-biased technology — the returns flow to firms and capital owners, not workers.

This isn’t even new. The Brookings Institution points out that productivity has grown 2.7 times more than worker compensation since 1979. Every major technology wave has widened that gap. AI is accelerating a pattern that’s been running for decades.

The OECD estimates AI could add 0.25 to 0.6 percentage points to annual productivity growth in developed economies. The IMF’s analysis by Brynjolfsson and Unger frames it bluntly: AI could dramatically boost productivity, but history shows those gains don’t automatically flow to workers.

Corporate executives are already planning for this. A Federal Reserve Bank of Atlanta survey found companies reported a 1.8% productivity increase from AI in 2025, with 30% of large firms planning to invest over $1 million in AI this year. They also expect to reduce headcount by 0.8% — small, but directional.

The pattern is clear: companies will recapture the productivity surplus. The only question is how fast.

Two scenarios. You’re in one of them.

Scenario 1: You do nothing.

Your best employees are quietly using AI. They’re more productive, but you don’t know it. You’re paying for output that takes them half the time to produce. That’s money you’re leaving on the table, but it’s not the real problem.

The real problem is your competitor. The accounting firm down the road just automated their invoice processing. The auto shop across town turns around estimates in 3 minutes instead of 30. The restaurant group that just opened their third location did it with the same admin staff they had for one.

They serve more customers with fewer errors and lower costs. They undercut you on price or outperform you on speed. You’re competing with yesterday’s tools against tomorrow’s operations.

Scenario 2: You capture the gains at the business level.

Instead of individual employees using AI as a personal productivity hack, you automate the workflows themselves. Invoice processing that required a full-time person becomes a system. Scheduling runs itself. Estimates get generated in minutes.

The difference is structural. When one employee is 2x productive and keeps the surplus, your business doesn’t change. When your entire operation is 2x productive, you serve more customers, reduce errors, and free your team for work that actually requires human judgment.

This is the same logic behind why your best employees’ knowledge shouldn’t live in their heads. When capability sits in a person, it’s fragile. When it sits in a system, it compounds.

This isn’t about replacing your team.

I know. You’re a 15-person business. You know everyone’s name, their kids’ names, what they order at the holiday dinner. This isn’t a Fortune 500 “workforce restructuring” memo.

But pretending nothing is changing doesn’t protect your team either.

Remember the ATMs. Tellers weren’t eliminated — their job evolved. The ones who adapted became financial advisors and relationship managers. The same thing is happening now. Your office admin’s job shifts from scheduling and data entry to exception handling and coordination. Your dental receptionist shifts from phone tag to patient experience. Your estimator shifts from typing to customer communication.

Some roles will shrink. Not disappear — shrink. A task that took a full-time person might take that same person 15 hours a week. The question is whether you use the other 25 hours for higher-value work or whether you eventually need fewer people.

As TIME argued recently, AI is the first technology accessible through natural language rather than specialized technical skills. That means workers themselves can shape how it’s used — if they’re given the chance. The businesses that handle this well, that retrain, reassign, and grow, will keep their best people. The ones that pretend it’s not happening will lose them to competitors who value their AI skills.

What to do about it

1. Acknowledge the window exists. Your employees are already experimenting with AI. Some are dramatically more productive. You’re not capturing any of that value. That’s OK for now — but it won’t last.

2. Automate at the business level, not the individual level. When AI is a company system rather than one person’s secret tool, the productivity gains flow to the business. Pick one workflow — invoice processing, scheduling, estimates — and automate it properly.

3. Move now, not later. The businesses that adopt AI in 2026 will have a 2-3 year head start on process optimization and operational efficiency. By 2028-2029, AI automation will be table stakes. The advantage goes to whoever moves first.

4. Invest in your team’s transition. The best outcome isn’t replacing your office admin — it’s turning her into an operations coordinator who happens to use AI tools. The businesses that retrain and upskill will keep their best people. The ones that don’t will lose them.

5. Start small. You don’t need a company-wide AI strategy. You need one automated workflow that proves the concept. The numbers typically work out at EUR 3,500-10,000 setup with EUR 300-800/month ongoing. Expand from there.

Figure out where your window is

I do free discovery calls where we look at your specific situation — which workflows are leaking the most value, which ones are the best candidates for business-level automation, and whether the investment makes sense right now.

Sometimes the answer is you’re not ready yet. I’d rather tell you that than sell you something that won’t deliver.