In late 2013 I sat in a conference room at Condé Nast in 4 Times Square with a stack of architecture diagrams nobody at the table had asked for. AWS was about three years into selling the enterprise. The deal in front of us was the first full data center evacuation AWS had ever signed, and Condé Nast had already sold the building. The migration had a deadline before it had a plan. I was on the AWS Professional Services team that was supposed to make it real.
The room was split. Two of the IT leaders were quietly in favor. Two were openly skeptical. The skeptics had real concerns: latency, security, the people whose jobs were tied to the infrastructure we were about to retire, the cost predictability of an opex model nobody had run for a decade yet. The believers had a thesis but not a proof. We were the proof. The migration moved forward. Some of it was clean, some of it was ugly, all of it taught us things you cannot learn from a whiteboard. Within two years the entire publishing portfolio was running on AWS, and a year after that "moving to the cloud" stopped being controversial inside Fortune 500 boardrooms.
I have sat in that same room three more times since then, with different logos on the wall and different acronyms in the deck. The fourth one I am sitting in right now is about AI.
I have lived through three of these waves as an operator. The internet at Hearst Corporation through the late 1990s and 2000s, where I started as a network engineer and ended up running technology development inside Hearst Business Media. Cloud at AWS Professional Services and at 2nd Watch through the 2010s. Modern data and analytics at 66degrees and Indicium through the early 2020s. The technology was different each time. The shape of the wave was almost the same.
Each one starts with a few years of skepticism. The smart, conservative buyers wait. They have their reasons, and most of them are reasonable. The data center is paid off, the data team has been trained on the old tools, the existing vendors are renegotiating contracts to make the new option look unnecessary. The skeptics are not wrong, exactly. They are early-mid in a curve that is about to bend.
Then a few brave bets land. With cloud, it was Netflix moving to AWS, then Condé Nast in 2013-2014, then GE writing its memo on going all-in on cloud. With data, it was a handful of mid-market firms discovering that cloud-native warehouses and dbt could replace a million dollars of legacy infrastructure with ten thousand dollars a month in compute. The brave bets are unevenly distributed. They are usually made by an operator who has been burned before by waiting and decides not to be on the wrong side again.
After the brave bets, the stampede. By 2017 every consulting firm in New York had a "cloud practice." By 2022 every analytics firm had been rebuilt around the modern data stack. The market consolidates around two or three platform vendors. The talent market reshuffles. The hardest people to hire are the ones who lived through the prior wave, because they are the only ones who can tell the difference between the real thing and the sales-deck version.
A few other things repeat reliably. The job description of the buyer changes during the wave, not after it. The CIO of 2008 was running a data center; the CIO of 2018 was running a portfolio of cloud platforms; both job titles were "CIO." The vendors who win are not always the ones who had the best technology in year one. They are the ones still standing in year five with a partner ecosystem behind them. A new operator type emerges every wave: cloud architects, data engineers, ML engineers, and now AI engineers and applied AI leaders. Most of those new operators were doing something adjacent five years before.
This fourth wave is following the same script. The skepticism phase is mostly behind us at the leading edge of the market. The brave bets are landing. The stampede is in front of us, not behind. The buyer's job description is changing in real time.
The operators who can tell the real thing from the demo are, again, the rare ones.
I do not want to sell a tidy story. The fourth wave is not a copy-paste of the first three. There are real differences worth being honest about.
The pace is faster. Cloud took roughly a decade to move from controversial to default in the enterprise. The modern data stack took five or six years. AI is going to take three or four. That is not breathlessness. It is reading the velocity of capital, capability, and customer adoption and adjusting the slope. The flip side: the cost of being wrong is also faster. A bad cloud bet in 2014 could be unwound over two years. A bad AI vendor bet in 2026 can be replaced by something better twelve months later, which is a different problem if you have already built workflow on top of it.
The capital intensity is different. Cloud was capital-light at the customer; you stopped buying servers and rented them. The data wave was capital-light at the customer; you stopped buying Teradata and bought a Databricks or Snowflake workspace. AI is, at the foundation-model layer, the most capital-intensive technology cycle most of us will see in our careers. That has knock-on effects. It concentrates power in a smaller number of model providers. It makes the partnership choices on the customer side higher-stakes than they used to be. The companies that picked AWS in 2014 had three or four credible alternatives. The companies picking a foundation model partner in 2026 have a tighter set of real options. (At Indicium I signed Anthropic as a foundational partner in 2024 because the choices in this layer matter more than the choices in the prior two waves did.)
Agent autonomy is the genuinely new thing. The first three waves automated workflows that humans designed. The fourth wave introduces software that can take actions, hold context, and chain steps without a human in the loop on every move. That changes the risk model. It changes the audit and governance model. And it changes what "useful" looks like, because a tool that is 80 percent accurate at one step is fine, while a four-step agent at 80 percent per step lands at 41 percent end-to-end.
The math is unforgiving in a way the first three waves were not.
And the buyer is changing while the technology is changing. CIOs and CDOs are being asked to own AI by boards that have not yet decided whether AI belongs in the technology org or somewhere else. The job title that ends up running this for a mid-market company in 2028 may not be Chief AI Officer. It may be the COO, or a head of operations who has gotten serious about AI, or a hybrid role we do not have a clean name for yet.
Most of the AI advice aimed at mid-market CEOs is either a 12-step framework or a vendor pitch dressed up as analysis. I am not going to write either one. (I wrote separately about the five patterns I keep seeing fail in mid-market AI strategies. Most of what follows is the operator answer to those failure modes.) If you are running a $50M to $500M company and you are trying to figure out what to do about AI between now and the end of summer, here is what has a real chance of being right.
First, get an honest map of where AI is already happening inside your business. Not a policy memo. A map. Which teams are running which tools, on which data, with whose credit card, producing what output. Most CEOs I talk to are surprised by how far this has already gone without a central decision. The map is not the goal. It is the starting point for every other AI decision you make this year. If you cannot answer "where is AI happening in our company today" in one page, the next twelve months will be harder than they need to be.
Second, pick one or two places where AI has a real chance of moving a number that matters to you. Revenue, cost, retention, time to close, gross margin, working capital, something your CFO would recognize. Then say no to ninety percent of what is technically possible so you can do those one or two things well. The companies that get the fourth wave AI moment right will not be the ones with the most pilots. They will be the ones who finished a small number of high-leverage projects and put them in production.
Third, decide how senior AI judgment is going to get into the room. The full-time hire is not the right shape for most mid-market companies in 2026. The market for a real Chief AI Officer is too expensive, the role is too undefined, and the wrong hire wastes a year you cannot afford to waste. The fractional and embedded operator models exist for this exact moment. Whether you bring in a fractional AI executive, an embedded sprint team, or a board-level advisor with operating chops, you need someone in the room who has done this before, at the scale just above yours, who does not have a vendor's finger on the scale. That is the call that costs you the least and protects you the most.
If any of that is useful, the AI Opportunity Sprint is the two-week version of the map and one-or-two-bets exercise above, and the fractional AI executive page describes the embedded operator option.
The conference room in 4 Times Square in 2013 is the one I keep coming back to. The skeptics were not stupid. They were doing what cautious operators do at the start of a wave: protecting the business from a story they did not yet have evidence for. Some of them were proven right in the short run. None of them were proven right in the long run.
The cost of waiting one more year on cloud was small. The cost of waiting three was the difference between leading their category and chasing it.
I started Strong Tide AI for the version of that room that is happening right now in companies between $20M and $500M in revenue. I have been the technical lead in that room, the AWS partner across the table, the cloud consultancy operator telling the CEO which of the bets in the deck were real, and the CRO at a data and AI consultancy realigning around Databricks and Anthropic. The mid-market version of that conversation is the one I want to be in.
If any of this hits close, the fastest way to find out whether a conversation is worth having is a twenty-minute call. No deck. No pitch. We talk about where you are, where you are trying to go, and whether one of the engagements I run is the right shape for the next ninety days. If it is, we spec it. If it is not, I tell you. Either way you get an honest read from someone who has done this for twenty-five years.
The fourth wave is here. The shape is familiar.
Twenty-minute scoping call. No slide deck, no pitch. We talk about where you are and whether a Sprint or a Fractional engagement fits.
Book a scoping call