A month ago I wrote The Classifieds Problem. Two agency owners on 21 Hats had looked back at my 2025 prediction that the general purpose agency was doomed and said, fairly, that it hadn't happened. I agreed with them about the past and disagreed about the future. My argument was that disruption would come the way it came for newspapers. AI-native companies would take the low-value work at the bottom of the pyramid, the classifieds that pay for everything else, and climb from there. Gradually, then suddenly.
Three things I read in the last few weeks make me think I described only half of it.
The bottom is getting funded
On October 8, Union Square Ventures announced $900 million in new funds, up from $275 million for its last core fund. The thesis is one sentence: "We believe AI will obliterate markets, not automate them."
Fred Wilson's example on a podcast about the fund was simple. Don't hire a lawyer, use an AI lawyer. Same for accountants. These companies won't sell your firm a better tool. They'll sell your client a replacement for your firm.
That matters because of something I got wrong. I assumed owners would adopt AI once they saw what it could do. Most haven't. If disruption depended on owners changing how they work, it could take decades. It doesn't. The new companies go straight to your clients, and your client doesn't care whether you adopted anything.
The top is getting bought
On October 29, Thrive Holdings, the private equity arm of Thrive Capital, agreed to buy ADT for $15.3 billion including debt. OpenAI owns a stake in Thrive and is supplying models and engineers who work inside the companies Thrive buys. The stated plan is to rebuild labor-heavy service businesses with AI.
ADT isn't a professional services firm. What matters is the model: buy a business that already has customers, people and operations, then rebuild how it runs with better AI than its competitors can get.
Then Noah Smith took the idea further. In a piece on Anthropic's coming IPO, he asked why Anthropic couldn't dominate an industry like insurance. Claude is already smart enough. What it lacks is the private information inside real companies: the customer data, the exceptions, the hard-won tricks. So buy a company. Feed the AI everything. Have the employees teach it their jobs. Then compete against rivals who only have access to the public version of the model,
It’s also possible Anthropic retains versions of its larger and better models for periods of time internally before making them public. That would ensure that they always have a smarter general model and the internal data from every company they buy. Repeat in every industry.
He's careful to call it a "huge, teetering tower of assumptions." But he also notes something I hadn't fully considered: "The wild thing is that this business strategy might be perfectly legal."
Christensen's disruptors start at the bottom with a worse product and climb. This attacker skips the climb. It buys the clients, the relationships and the data. Those are the things I said would be left for incumbents when the routine work was gone. Then it puts a model behind them that you can't buy.
So services firms get squeezed from both ends. Funded startups take the classifieds from below. Funded acquirers buy competitors from above and make them better than you.
Why it won't happen next quarter
None of this is immediate. The frontier labs are hesitant to draw the wrath of regulators, voters and politicians, and buying their way into whole industries would do exactly that.
It won't stay that way. OpenAI is using Thrive as a shield and both Anthropic and OpenAI need growth. The IPOs and their competitors will force their hand.
ADT is going to take a while too. Acquirers will need time to figure out how to combine an accounting firm or a security company with frontier AI and actually run it differently. Most of the first attempts will be messy. But they'll get there.
The protection that will probably last longest is licensing. Most states don't let non-lawyers own law firms, and firms that sign audits generally have to be majority-owned by CPAs. That shields the part of the work where someone signs their name and takes responsibility. Agencies, consultants and most of the services world have no such wall.
Two honest options
For most owners, my advice is to prepare what you can. But if you don't have the energy to change your orientation, I think you should sell in the next two to five years, while buyers are still paying for the business you have today.
If you want to win in the new world, you're going to need to radically reorient. That means more than adding tools to the way you work now. Here's the order I'd do it in.
1. Build the foundation now. Attach paid Claude and Codex models to your email, document system, Slack, etc and have it start documenting how the business actually works: how you price, how you decide, which exceptions matter. Log the outcome of every engagement. Get your knowledge out of people's heads and into a form an agent can use. That's exactly what every acquirer in Noah's scenario wants to buy. It's also what you need to compete with them.
2. Learn to build yourself. You can't direct this from a distance or hand it entirely to a vendor. You will need to use vendors to get this done and probably make some internal hires to get expertise, scale, and speed. Despite that, the owners who win will understand what the tools can do because they've built with them.
3. Then figure out how to scale. I see two paths.
Acquire other firms and run your playbook on them. This is what Thrive is doing at a much bigger scale. You buy clients, people and data, and you take on integration risk.
Cut your cost to serve and grow organically as fast as you can. Cleaner, but it only works if your costs drop far enough to win clients on price or speed.
Which one is right depends on your market and on you. What isn't an option is waiting to see.
Gradually, then suddenly
Newspaper ad revenue was still going up in 2005, almost a decade into the internet. Then it cratered. I expect services to look calm for longer than the pessimists think, and then change faster than the optimists think.
The difference this time is that the people coming for your classifieds and the people coming to buy your competitors are both funded. Decide which side you're on while you still get to choose.
Sources