The Middle Disappears
Soon the cheapest consultant money can buy and the most expensive one will be the same setup: one person, one laptop. What separates them is invisible from the outside — and it isn't price, niche, or a clever data moat, because those all get copied within a couple of years. It's whether the firm is built AI-native: wired so its judgment sharpens as fast as the work it sells becomes free. That shape is the only moat left. Everything you sell sits on top of it, on borrowed time.
The cheapest competent lawyer, accountant, marketer, or consultant you can hire in 2027 will be one person with a laptop. The most expensive will also be one person with a laptop.
Most of what gets written about AI and professional services in 2026 is half-right. AI commoditizes services, the consensus says, so go vertical, productize aggressively, niche down. Fine. The harder questions — the ones that decide who wins — are which services commoditize, on what timeline, and what is left to defend once the wave has passed. Most operators are answering those poorly, because they are answering the wrong version of the question.
Services are quietly turning into products
For about a decade the smartest writing about service businesses has been about productization. Sahil Lavingia, Jonathan Stark, Brennan Dunn — the minor canon of independent-business thinking all pointed at the same insight. Repeatable, fixed-price offerings beat hourly work because they let people charge for outcomes rather than time.
Those frameworks were clever pricing innovations because humans still did the work. The pricing was the product. The labor was the substance underneath. The arrangement made margins better while leaving the structure of the business intact.
Now AI does most of the work. Once you remove human labor as the substance, "productized service" stops being a clever pricing innovation. It becomes a product. And products, unlike services, compete on price.
This is what the consensus take captures correctly. Horizontal services — the kind any agency, firm, or consultancy can deliver for any vertical with the right LLM stack — are getting commoditized at speed. The work is done by software now. The customer can buy it from anywhere. Margins collapse.
Vertical is necessary. It is not sufficient.
The standard 2026 advice for service operators is to pick a vertical and specialize. Healthcare. Legal. Real estate. Water utilities. Narrow your aperture, become the deep expert, charge more for the depth.
This is correct, and it is also cheap. Anyone can pick a vertical. Hire a couple of subject-matter experts, prompt-engineer an LLM around the right vocabulary, buy a domain name that contains the sector's name, and you have what looks from the outside like a vertical agency. The market is already full of firms that "specialize in legal" or "specialize in healthcare" and are indistinguishable from one another behind the marketing.
Specialization is positioning. Defensibility comes from owning the structural gap underneath a vertical — the layer of information, infrastructure, access, or integration that is hard to assemble in that sector. The services that ride on top of that layer are the same services every other vertical agency sells. What is different is what they sit on top of.
In water utilities, the underlying gap is fragmented compliance data scattered across federal, state, and utility-level systems. In legal, it is structured access to case law, regulatory filings, and contract precedent. In healthcare, it is claims data and the seams between payer and provider systems. The shape repeats in every vertical worth being in. There is something structurally hard to assemble underneath the surface, and the firm that owns it has a real advantage. Everyone else is selling the same generic LLM-powered work in a different costume.
What separates a vertical agency from a vertical-flavored generalist is whether they sit on top of something hard to copy. Most "AI-powered healthcare agency" or "AI-powered legal agency" pitch decks read identically because they don't. They are all selling the same services on the same generic substrate, and none of them are sitting on top of anything proprietary.
The moat is temporary
The structural-gap moat does not last.
The same AI capabilities that let you assemble a fragmented data layer cheaply are lowering the cost for your competitors. APIs standardize. Schemas get bridged. Integrations that took two years cost the next entrant two months. The advantage erodes from underneath, and the rate of erosion accelerates as model capability accelerates.
If I had to bet on a window, it would be somewhere between eighteen and thirty-six months in most verticals before the gap closes enough that a well-funded competitor catches up. I could be off by a year in either direction. But it is a window, not a fortress, and the window shrinks every year.
The mistake most operators make during the window is to spend it widening the moat. Adding more integrations. Acquiring more data sources. Hardening what they already have. None of that buys real time, because the force closing the gap is AI capability itself, and you cannot outrun that by accumulating more of the thing AI is making cheaper.
The window is for building the next thing. Most operators never get there because they spend it defending the current thing.
The shape of the firm is the moat
The reason small AI-native teams are already producing 10x, 50x, even 100x more than their incumbent competitors is not the tools. Everyone has access to the same tools.
The difference is in how the team is wired around the tools.
How work gets assigned. Which workflows are tolerated. What hiring criteria still apply. Whether AI is default or optional. Whether the partners' incomes depend on the old labor model continuing. Whether the people closest to the problem have the authority to redesign the process. Whether failure is allowed cheaply enough that experiments happen. All of these are organizational decisions, not technology decisions, and all of them compound.
Before AI, a great operating model was a 2x advantage over a mediocre one. With AI, the same operating discipline now governs an exponentially more powerful tool stack, and the gap becomes enormous. Same person, same tools, different operating model, wildly different output. The gap widens every month.
I think most discussion of AI and service businesses is solving for the wrong question. The question worth asking is what an agency would look like if it had been designed from scratch to be AI-native, and whether the existing agency is willing to become that. Most are not. Most are not even allowed to be by their own incentive structure.
The new moat is not what you sell. It is how the firm that sells it is shaped.
Christensen, updated
If you have read Clayton Christensen's The Innovator's Dilemma, you already know how this ends. Incumbents fail not because they misunderstand the new technology, but because their organizations are physically incapable of adopting it. The pricing model is wrong. The sales motion is wrong. The hiring pipeline was built for a different kind of work. The cost structure can't survive the transition without a rebuild. Each of those things is individually fixable. All of them together is a rebuild — and rebuilds are what incumbents cannot afford to do without dying mid-rebuild.
Apply this to a 200-person law firm or a 100-person agency in 2026. Pricing, hiring, billing, culture — all of it built around humans doing the work. The partners whose incomes depend on associates billing hours are not going to vote for an operating model that makes the billable-hours layer redundant.
The structure resists the shift before the strategy conversation can even begin. The senior people defend the model that made them senior. The junior people see the writing on the wall but lack the authority to act on it. The leadership group looks at the spreadsheet, looks at the partnership agreement, and decides that adding an "AI initiative" is easier than restructuring. They are correct about which is easier. They are wrong about which keeps the firm alive.
One prediction I am willing to be wrong about on a public timeline. Most agencies, law firms, and consultancies that try to "add AI" without rebuilding the operating model itself will be smaller in twenty-four months than they are today. The timing might be off. Some verticals will move faster than others. But the direction is the part I'd stake the post on — some of the shrinkage will look like consolidation, much of it will be replacement: small AI-native teams quietly eating work that used to require thirty people.
What this means for you
For operators running a service business, the fight that decides whether you exist in three years is not the pricing fight or the positioning fight or the vertical-selection fight. Those are the visible ones, the ones that get airtime in your weekly meetings because they feel solvable in a quarter. The organizational-design fight is the harder one, and the one most operators avoid for that reason. It is also the only one with a payoff that compounds.
For people working inside a service business, the culture is the leading indicator. If leadership is talking about "AI training" and "AI literacy" but not changing how teams are structured, how work is assigned, or how performance is measured, the firm is on the losing side of this. The announcements that matter are not about new tools. They are about new shapes of work.
For everyone else, this pattern is the leading edge of something coming for nearly every expertise-driven business in the next three to five years. Law, accounting, finance, marketing, healthcare, real estate, education — they all run on the same labor-intensive operating model that is about to get rearranged. Professional services are first because their unit economics are the most exposed. Everyone else is on the same curve a little further behind.
The barbell future the opening sentence describes — the cheapest expert and the most expensive expert both being one person with a laptop, with nothing left in the middle — extends well past agencies. It is the future for every business that sells expertise.
The agencies are just the earliest example.
The new shape
The firms that come out of this stronger have a different shape than the firms holding power now. They are built around the assumption that AI does most of the work, that judgment is the rare resource and worth concentrating, that the talent worth keeping wants to be inside a structure that lets them operate at the edge of what is possible rather than the middle of what is profitable.
Scaling judgment is the new game. The kind of organization that can do that is different from the kind that scaled labor — in hiring, in incentives, in how it measures success, in what it concentrates and what it does not.
There is a window to build the new kind of firm. There is also a window to recognize what kind of person you would have to become to work inside one. Both windows are open. Neither one will stay open very long.