The AI Roll-Up: Buying the Business, Rebuilding the Work

The Next AI Company Might Already Exist

The next interesting AI business might already have customers.

It might have a familiar name, a team that knows the clients, and years of work behind it. It might prepare tax returns, manage properties, or answer calls. From the outside, it could look like the sort of company technology investors used to walk past on their way to something more exciting.

Now some of those investors are stopping at the door.

Their idea is to buy established service businesses and rebuild how the work gets done. Keep the customer relationship. Bring engineers into the operation. Put AI inside the daily service, where it can help produce the thing the customer was paying for all along.

That is the emerging AI roll-up: a business built by acquiring other businesses, with technology intended to improve what the combined operation can deliver.

The pitch can be reduced to a provocative line: the customer wouldn't buy the software, so the investors bought the customer.

The reality is more substantial than the line. Buying a firm does not buy its clients' loyalty forever. But it can give a technology team something that another software demonstration cannot: the authority to change the work, alongside the people who understand it.

For anyone who has looked at a perfectly useful business and thought there must be a better way to run the machinery behind it, this is an idea worth getting excited about.

Sell the Result, Rebuild the Work

Consider the difference between selling an accountant a tool and helping run the accounting firm.

A tool has to fit into the business as it stands. Someone must evaluate it, find the time to introduce it, teach the team, and connect it to the rest of the work. Even a promising product can become one more thing the owner has to manage.

Ownership changes the starting point. The buyer can fund an engineering team, choose a process to redesign, and work through the awkward details of introducing it. The service business becomes the place where the technology is built and used.

The customer can keep buying the result: a completed return, a resolved support request, a property looked after. They do not have to become an AI implementation project themselves.

There is an appealing reversal here. For years, businesses have been asked to adapt their working day to the software they purchase. This model puts the software builders inside the working day and asks them to make it better.

That does not make buying companies easy, or software sales obsolete. It gives builders another route to the problem.

General Catalyst has made that route part of its investment thesis. It frames selected U.S. service industries as a market producing more than $6 trillion a year, compared with a much smaller software market. Its broader Creation fund has grown to $1.5 billion, supporting both new company formation and the acquisition-led transformation of established businesses.

The appeal is not simply that the market is large. It is that customers already spend money to get this work done.

An AI roll-up does not have to create the need for accounting or customer support. Its opportunity is to deliver those services with a different balance of people, technology, time, and cost.

The Firm Behind the Familiar Name

Thrive Holdings offers a concrete view of what that can look like.

Its accounting platform, Current, was previously known as Crete Professionals Alliance. It brings together dozens of firms, with local businesses retaining their branding, leadership, and equity while gaining access to shared resources and technology.

That is an important part of the proposition. A local firm's value includes the relationships its people have built. Replacing the sign above the door is not necessarily the most useful change an owner can make.

The more interesting changes happen behind the service.

OpenAI took an ownership stake in Thrive Holdings and announced plans to embed research, product, and engineering teams in its companies. Current describes engineers and researchers working with accountants to build tools around their actual work.

One early example is Tax AI. OpenAI reports that a pilot processed 7,000 returns and saved practitioners roughly a third of their tax-preparation time, with people reviewing the output.

Those are results reported by the organizations involved, but they make the ambition tangible. A busy professional gets some time back. The next question is what the business does with it.

More client conversations? More work completed without extending the day? A service that used to be too costly to provide?

Current's stated aim is to make room for deeper advice and more client contact. Whether that happens consistently will matter as much as the speed of preparation.

A faster process is useful. A faster process that leaves the customer better served is the beginning of a stronger business.

Where the Excitement Has to Meet the Books

The financial ambition is considerable. General Catalyst describes businesses targeting margins of 30 to 40 percent as part of the strategy.

Targets are doing important work in that sentence.

They are not a promise that a newly acquired firm will produce that return, or that the same economics will hold across accounting, call centers, and property management. Each business has different costs, customers, and demands on the people responsible for the service.

The distinction between saving time and making money is especially important.

If a task takes less time, the business has gained capacity. Turning that capacity into profit takes another step. It may need more customers, a better service offering, a different way of pricing the work, or lower delivery costs.

And the new system has its own bills. Engineers must build and maintain it. Models cost money to use. People must review uncertain answers and correct mistakes. Integrating another firm can add work before it removes any.

General Catalyst itself describes a period in which technology and integration costs rise before the returns arrive. That is a more useful picture than the idea of acquiring a company, switching on AI, and immediately discovering software-like profits.

The strongest version of this model is a patient operating project. Its buyers say they want to expand what the business can do, rather than merely shrink what it spends. That intention still has to show up in the service, the staff's working lives, and the accounts.

People are central to that test. An experienced employee may know why a straightforward request is actually an exception. If the system makes that person more effective, the business gains something. If it removes the person before it can handle the exception, the apparent saving can become someone else's problem.

Then there is the question an exciting early result cannot settle: what happens when business gets harder?

When clients spend less, can the company still support its technology costs? When acquisitions slow, does the existing operation keep improving? When a customer leaves, was the service really better, or had the business simply been growing fast enough to hide the loss?

An established accounting firm may have a long history. That history does not automatically prove the resilience of the new AI ownership model placed around it.

The current evidence is enough to take the idea seriously. It is not enough to declare the margin question settled through every kind of economy.

A Bigger Future for an Ordinary Business

The reason to watch AI roll-ups is not that every local firm needs a technology investor at the table.

It is that the model changes where we look for the next business opportunity.

A company does not have to invent a new customer habit to become more valuable. It can begin with work people already need and improve the way that work reaches them. The innovation may be a better service business rather than another product someone has to learn to use.

That is an encouraging thought for owners who know an industry from the inside. Their understanding of the customer, the exceptions, and the work that gets in the way is part of what makes a useful system possible.

The technology team needs that knowledge. The customer still needs someone responsible for the result.

For a smaller operator, the useful question comes before any acquisition: which part of the service could become meaningfully better if the repetitive work stopped consuming so much attention?

The answer might reveal room for more personal service, a new offering, or a business that can grow without asking its owner to absorb every extra task.

That is the hopeful version of the AI roll-up story. Familiar firms keep the relationships that made them worth buying and gain the ability to do more for the people they serve.

It will take more than an acquisition announcement to make that future real. Customers have to stay because the work is good. Employees need systems that help them handle it. The financial gains must survive the full cost of delivering the service.

But the possibility is larger than a cheaper back office.

Somewhere inside an ordinary business, there may be a much better business waiting for the work to be rebuilt.

Rich Oneday

Emerging Models explores new ways to build a business: what makes them promising, how they work, and what still has to be proved.