Small Waves

The Transition Is Already Happening, One Office at a Time

Issue 281, September 10, 2026

A few weeks ago at the beach, I fell into a long conversation with a man who runs a real estate firm. A successful outfit, the kind of place that has been a fixture in a small market for years, the name on the signs you drive past on the way in. We got to talking the way two people who run businesses do, and before long we were comparing notes on what each of us had done with AI.

He is up until two in the morning most nights. Not because anything is broken and needs fixing in the moment. He is building. He has been assembling his own systems, databases, and agents in Claude Code, one function at a time. He was as animated, excited, and satisfied describing what he has been doing as anyone I have talked to all year. Document prep. Marketing. The listing workflow. Sales funnels. New services for lead capture. All of the things he used to pay somebody else to handle. He walked me through what each piece does and what it replaced, and he was relishing every bit of it.

Then he said the other half, in the same breath and the same tone, because to him it was one coherent sentence of thought. He no longer needs to employ the people he used to employ. He no longer needs to find himself frustrated by what he got in return; he no longer needs to contend with the rising service rates, nor the constant coordination and follow-up. All he needed was to continue to replenish his tokens. A single need after years of managing contractors and employees to maintain and grow the business.

He’s not the villain of this or any similar story as some might suggest. He saw an opportunity. He built a lot. It all works with much less coordination and overhead. It saves him substantial money and energy every month. And most importantly, he has an outlet with the only expense of his time, to continue to leverage his creativity and bring his ideas into reality. Anyone who has ever run a small business knows exactly what that feels like, me included.

But I have thought about the conversation a lot since, given there is a pattern in his story that is being overlooked. It’s about the handful of people who used to have that income and now don’t. Maybe a few hundred dollars to a few thousand dollars a month that now needs to be replaced in some way. The someone who prepared the contractual documents. The someone who kept the listings current and the marketing going out. The someone who made sure the computers worked.

And nothing about the shift will ever show up anywhere where someone is paying attention. There was no announcement. No restructuring. No press release, no consultant, no formal change or transformation program with a name and a steering committee and a slide that says which quarter it begins. A man stayed up late for a few months, and an office quietly stopped needing five or six people it used to need. Think about that for a few moments.

That is one small wave. It rose, it came down on five or six companies or individuals, and the water went flat again before anyone outside the building had any reason to notice. Which is the whole problem, because a wave that size is easy to miss. It doesn’t draw one’s attention. It likely does not become part of any business or economic survey or report. But it is consequential nonetheless. So, what is the impact of this when there are a few million of these small waves, all rising at once in different cities and towns across a country, none of them big enough to report on or make the news?

Millions of the Same Decision

Let’s start with how many small companies like his there are.

The Small Business Administration publishes these numbers every year, and the February 2026 edition, working from 2022 Census data, counts 36.2 million small businesses in the United States. That is 99.9 percent of all American firms. They employ 62.3 million people, close to half of everyone who works for a private employer. They account for 43.5 percent of the economy. And from 1995 through 2024, small business created 20.7 million new jobs against large business’s 13.2 million, which means roughly six of every ten jobs added over thirty years came out of a small firm.

Now put the beach office back in the frame with those numbers around it. Almost half the people who work for a private employer in this country work somewhere like it. These are not places with a change management office and a pilot cohort and a communications plan. They are places where the owner decides on a Tuesday and it is done by Friday, and where the only approval required is his or her own. Every one of those 36.2 million owners can make the same decision he made, in the same week, without telling anybody, and most of them are hearing about and considering the use of the same tools he leveraged.

One more thing about what I think is the shape of that decision, because it explains why these waves become more numerous and none of them recede. Gartner published research in June 2026 arguing that AI is not reducing workforce costs so much as reshaping them. Jan Bansch and Joe Coyle found that up to thirty percent of AI-displaced roles will be rehired by 2029, often at a higher cost than before. That is a real and useful finding about companies that have hiring cycles, pay bands, and an HR department. But none of it describes the office at the beach or the millions of small businesses that are serving customers in a variety of ways. There are no rehires coming in 2029 or in any future year, because there is no budget line waiting for one and no plan that is in development or crafted just in case. Whatever he saved, he simply saved, and it stays saved. Gone from his operating expenses and from the daily and weekly task list. In a large company, the water tends to wash back. In a small one it doesn’t. It becomes the new normal, in reality and in perception, and the business becomes more efficient.

Waves That Leave No Mark

So if millions of these are happening, why does it feel like nothing is happening?

Part of the answer is that our ways of assessing employment, the economy and more are pointed at the wrong thing. A Federal Reserve note published in April 2026 lined up the available surveys on AI adoption and the answers don’t agree with each other. Count firms, and about eighteen percent had adopted AI by the end of 2025. Count the labor force by where people work, and roughly seventy-eight percent are employed at a firm that has. Ask individuals what they actually do at their desks, and about twelve percent use it daily. Same economy, same points in time, three defensible numbers, and three completely different stories about how far along we are. This is the challenge I talked about last week surrounding measurement. Despite the gaps, despite the omissions, despite the limited or even biased output, the numbers are something that we look to in defining the current conditions. We use that output to make decisions. To inform our strategies. To rank against others.

But notice what none of those surveys ask. Every one of those surveys wants to know whether a firm uses AI, which is a question about tools. A firm that bought licenses and changed nothing counts as an adopter. A firm that quietly went from nine people to four counts exactly the same way. We are measuring who bought a boat, not whose office emptied out.

The second part of the challenge comes from the Stanford Digital Economy Lab, where Erik Brynjolfsson, Bharat Chandar and Ruyu Chen have been working through payroll records from ADP. Their August 2026 update cuts both ways. Workers between twenty-two and twenty-five in the jobs most exposed to AI are now roughly nineteen percent below where they would be if they had kept pace with people the same age in less exposed jobs, and a year earlier that gap was fifteen percent. At the same time, the researchers say plainly that they do not see broad job loss across the economy from AI. What they see is concentrated rather than general, and where AI helps people do their work instead of doing it for them, employment held steady or grew. Anyone telling you the job market is being gutted wholesale, either right now or in a near-term projection, is embracing the media hype and not the evidence.

But between those two findings sits the line that I think matters most here. The change, they write, is happening mainly through hiring fewer young workers rather than through letting people go. There are near-term and longer-term consequences to that finding that surround the quantity of jobs available to those who are seeking to be employed. We are feeling the current, but it seems smaller than it is.

Their sentence explains why these waves don’t leave any mark in the sand. A layoff is an event. It has a date, a number, a memo, and sometimes water cooler gossip about people carrying boxes out to their cars. It’s recorded in the firm’s human resources system. A job that is never posted doesn’t exist. There is no announcement that a firm decided, quietly and reasonably, not to replace the person who left in March. There is no coverage of the coordinator role that used to exist at four hundred small firms and now exists at three hundred and forty. Those seeking employment spend their days searching and searching. For what they do find, they send more and more applications into a market that feels inexplicably harder than the one they walked into three years ago. Think about this for a few moments. If I have more than a handful of stories of this happening to those I know, you surely do as well.

There is something that may be even harder to see. Not everyone the real estate owner stopped paying was an employee. Some of that work was contracted out: the marketing and the document preparation and the fellow who kept the computers running, and in a firm that size a good deal of the work usually is. When an employee is let go, a record exists somewhere. There is a separation, usually an unemployment claim, a number that eventually reaches a statistical agency at the city, state, or federal level. When a contract simply is not renewed, there isn’t anything at all. No separation, no claim, no notice, no filing. The invoice stops arriving and both parties move on.

The Bureau of Labor Statistics does count independent contractors, and as of July 2023, there were 11.9 million of them, 7.4 percent of everyone working, up from 6.9 percent in 2017. But look at how that gets counted. It is an add-on to a household survey, and the Bureau ran it in 2017 and then again in 2023. Twice in six years. The one exercise that might catch what is going on only opens its eyes about once every six years. I have written often about the speed at which everything moves in today’s marketplace. Whether it be AI, consumer spending or other, six years to learn what is happening results in a lifetime of no information.

We have failed at this before, time and time again. When a plant closed, we counted it, wrote about it, and argued about it for decades, because a plant closing is an event. It has a gate, a last shift, a parking lot that empties out. Some reflect on this as the natural cycle of a capitalist economy. Others recognize the span of impact this has across more than the firm.

What happened to small downtowns across those same decades was bigger, and almost nobody noticed it while it was happening, because it showed up as a local hardware store that did not reopen after the owner retired and a local card shop that turned into a vacant window. Nobody held a press conference for that. There was no last shift. The whole thing was made out of things that stopped, and we have never built mechanisms that help us count what stopped and why.

What the Water Looks Like From Shore

Here is the part that I think matters and what has been on my mind the past few weeks. A few million small waves don’t stay small. Individually, they are invisible and they leave no mark in the sand. But collectively they stop being small events at all and become a water level, and the water level is measurable even when none of the waves were.

So look at the water level.

Markets have been strong. The aggregate numbers suggest we are in a really good period that is defying what we thought we previously understood about how the economy works. And yet almost nobody I talk to describes their own current life in that way. So which is it? Are we in a strong economy that people are somehow failing to appreciate or that they don’t see? Or a weak one where the indicators we use to assess the current state are somehow missing it? I think the Federal Reserve Bank of New York informs part of the why.

In a series published in May 2026, Rajashri Chakrabarti, Thu Pham, Beck Pierce and Maxim Pinkovskiy tracked spending by income group and found the growth is coming almost entirely from households earning more than one hundred twenty-five thousand dollars a year. They were the only group whose spending consistently rose. Lower-income households actually spent less for part of that stretch and did not get back to where they started until the middle of 2024. Their companion piece asked why, and found that wealth has split the same way since 2023, the top group up more than twenty-five percent while the middle forty percent gained less than ten, and that lower-income households have been paying higher inflation than everyone else since late 2022. The same year produced a genuinely good number and a genuinely worse life, depending on which house you walked into. This describes what different segments of the working world are experiencing, and it really depends on who you are and what your income level is.

Let’s explore some additional information. In July 2026, one point nine percent of workers quit their job in a month, according to the Bureau of Labor Statistics. In 2021 and 2022, we wrote issue after issue about people leaving jobs in record numbers and what it meant that they felt free to do so. Now they are staying. And not because the jobs got better, but because they feel they have no choice.

Another New York Fed team expanded on what seems to be really happening. In work published in May 2026, Gizem Kosar, Ishva Mehta and Wilbert van der Klaauw looked at households that had skipped meals or run short on food. Among those households, the share who believed they could find a job fell from just under forty-nine percent in the spring of 2020 to a little over forty-one percent in February 2026, and it sat lower still, at just over thirty-seven percent, as recently as last October. And as of February 2026, far more of them expected to be worse off a year from now than better, by a margin of about thirty-two points. That isn’t a mood, as some describe it, and it’s not pessimism as a personality trait. Those are households behind on something that is not optional, who have concluded there is nothing better to move to.

None of those numbers have AI written on it, at least not directly. Thoughtful people will read the same figures and land somewhere different on what is causing them. Everyone has a viewpoint, biases, and an opinion. But this is what it looks like from the shore when a great many small waves come in at once. Not a headline. But a water level that rose.

The Tide Is Already In

There is a version of the AI conversation that treats all of this as deterministic and as a forecast. Something coming, on a schedule, that we will have time to prepare for once the large organizations finish their pilots and publish their lessons learned.

But it isn’t ahead of us. It’s running now at a pace set by thousands of people like the man I met at the beach, each making a sensible decision at two in the morning, none of them coordinating with anybody, none of them announcing anything, and every one of them creating their own small wave and adding to the water level.

Which is exactly why it is so easy to not see. There isn’t anyone to be angry at, no decision to overturn, no announcement to answer. A tide gives you nothing to push against. It is just going to rise, or fall, and most people feel there isn’t anything they can do about it.

A couple of weeks ago I put together a sampler called The Conditions Nobody Chose. The thread running through it was that most of what shapes working life shows up as a condition rather than a decision. There isn’t anyone who votes on it, nobody announces it, and by the time it is clear enough to argue about, it’s already the ground everyone is standing on. This is one of those times, with one difference.

The difference is that we are not even looking away. We keep watching the harbor for a big wave. It’s in the media; it’s in countless social posts, blog posts and recaps from conferences. Many are expecting the large swell to come tomorrow, next month or in a few years. Many are feeling anxiety, economic pressure and/or identity threat as they wait. But the tide came in while we were looking.

I keep coming back to something simple. You already know some of these stories. Somebody’s hours that got cut and never came back. Somebody who has been looking since spring and cannot explain why it is so much harder than it was. A vendor who stopped invoicing because there was nothing left to invoice. We hear each one on its own and file it as that person’s bad luck.

That is how a tide stays invisible. It arrives as a few million separate pieces of bad luck, every one of them explainable by itself.

So before the next forecast about what AI is going to do to work in five years, there is a smaller question worth asking. How many of these stories do you already know? And what changes if they are not separate stories at all?

Related Reading

The Identity Renegotiation (Issue 276). What AI asks people to give up before it gives anything back. This issue is what that renegotiation looks like from the outside, in a market where everyone knows everyone.

The Permission Paradox (Issue 269). The workforce had already transformed itself in private while the instruments showed people waiting. Small firms are doing the same thing, at the level of the whole business.

Who Owns That Number? (Issue 280). Last week, on measurement that accumulates and never gets removed. A companion problem, which is what happens when the thing that matters most produces no measure at all.

The Proxy Collapse (Issue 278). On what happens when the evidence we relied on stops working. Employment statistics are a proxy too.

The Conditions Nobody Chose (Issue 279). A late summer sampler on conditions that arrive without anyone choosing them, which is the frame this issue is built on.

Measuring What Matters. The four questions, including the one this issue keeps running into. What decision or action does this measure inform.

The Transformation Readiness Assessment. A baseline on whether your organization is prepared for what is already underway.

On The Human Factor Podcast

The Measurement Ratchet, Season 3 Episode 1. Why nothing you measure ever leaves, and what it takes to retire a number. Out now.

The Long Game, Season 2 Episode 030. Tom Serena on pace and patience as the denominator of change. The pace problem is the whole subject of this issue, seen from the other end.

Research Sources

1. Bansch, Jan, and Coyle, Joe (2026). AI Isn’t Reducing Workforce Costs — It’s Reshaping Them. Gartner, June 1, 2026.

2. United States Small Business Administration, Office of Advocacy (2026). Frequently Asked Questions About Small Business, February 2026, using 2022 Census Bureau data.

3. Brynjolfsson, Erik, Chandar, Bharat, and Chen, Ruyu (2026). Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence, August 2026 update. Stanford Digital Economy Lab, using ADP payroll data through mid-2026.

4. United States Bureau of Labor Statistics (2023). Contingent and Alternative Employment Arrangements, July 2023, Current Population Survey supplement.

5. Chakrabarti, Rajashri, Pham, Thu, Pierce, Beck, and Pinkovskiy, Maxim L. (2026). Tracking the K-Shaped Economy: Who’s Driving Spending? and Explaining the K-Shaped Economy: What’s Behind the Divide? Liberty Street Economics, Federal Reserve Bank of New York, May 1, 2026.

6. United States Bureau of Labor Statistics (2026). Job Openings and Labor Turnover Survey, quits rate, July 2026.

7. Kosar, Gizem, Mehta, Ishva, and van der Klaauw, Wilbert (2026). Food Insecurity and Consumer Pessimism. Liberty Street Economics, Federal Reserve Bank of New York, May 27, 2026.

8. Allen, Jeffrey S. (2026). Monitoring AI Adoption in the U.S. Economy. FEDS Notes, April 3, 2026. Figures drawn from the Business Trends and Outlook Survey, the Real-Time Population Survey, and the Survey of Business Uncertainty.