Over the last year, I’ve met hundreds of founders pursuing the AI-enabled rollup thesis. Most activity has concentrated in categories like accounting, IT MSPs, wealth management, insurance brokerages, and HOA management. These are all perfectly reasonable places to build. The challenge is that they’re also the first places everyone looks. Private equity has been consolidating many of these industries for decades, competition for acquisitions is intense, and quality assets increasingly trade at premium multiples.
The markets that interest me most today tend to be the ones sitting slightly off the beaten path. They are still fragmented. They are still relationship-driven. But they also benefit from powerful secular tailwinds that have nothing to do with AI itself. In many cases, demand is growing, labor is constrained, and technology adoption remains surprisingly low.
Below are my requests for startups where I’d be especially excited to meet founders.
Legal Services
Legal is an incredibly exciting market for AI-enabled services and rollups with the obvious driver being AI. A huge percentage of legal work involves reviewing documents, synthesizing information, drafting content, and navigating large bodies of text. Those happen to be the exact activities where large language models perform best.
The less obvious driver is market structure. The U.S. legal services market exceeds $430 billion annually, yet it remains incredibly fragmented. There are roughly 450,000 law firms in the United States and approximately three-quarters have fewer than five attorneys. Legal services firms often have real regulatory moats that prevent competitors from entering easily. Despite its size, consolidation has historically been limited because non-lawyers generally cannot own law firms.
That may be beginning to change. Arizona and Utah have already adopted alternative ownership structures. The law-firm MSO model, where investors own the operational infrastructure surrounding a practice while attorneys retain ownership of the legal entity itself, has also become increasingly common. Lawyers are more open to exploring this structure based on its success in adjacent professional services markets like accounting.
On the AI-enabled services side, companies like Crosby, Soxton, General Legal, and Ivo are building de novo law firms focused largely on low-end transactional work like contracts. This makes sense as a place to play when building a net new firm because barriers to entry are lower and the billables-based business model of incumbent law firms is ripe for disruption. Harvey and Legora are apparently making moves into owning the service layer as well. They likely are running up against TAM constraints being just vertical software providers.
On the rollup side, I am more interested in teams consolidating law firms that are doing higher-end work that is more relationship-based, bespoke, and recurring. These firms tend to be larger and often focused on a particular practice area or geography and are more involved with litigation work. This is where a rollup play makes the most natural sense.
Adjacently, I think there are a lot of non-law firms but legal services-focused plays that are interesting candidates for a rollup. These include court reporting/stenography and trust and probate administration where a lot of the work can be sped up by AI but the existing businesses have regulatory moats and longstanding relationships with customers.
Data Center Services
Everyone wants to own data centers. I’m much more interested in owning the businesses that service them.
The reason is simple. Every investor in the world understands the AI buildout story. Fewer people spend time thinking about what happens after the facility gets built.
US data center power demand is expected to more than double by 2030. Hundreds of billions of dollars will likely be invested in new facilities, power infrastructure, cooling systems, substations, and related assets over the next decade.
Once those facilities are operational, they require constant maintenance and support. Backup generators need servicing. Electrical systems need inspection. Cooling systems need maintenance. Uninterruptible power supply infrastructure needs monitoring and testing. Entire facilities need commissioning (formal testing and verification) before they can even come online and older facilities need recommissioning. These needs are taken care of by a very fragmented base of specialized, regional firms. While much of this work is manual and physical, I imagine lots of it can be improved by AI and technology, whether through fleets of drones doing inspections, thermal anomaly detection, processing data and creating reports.
I wouldn’t be surprised to see a company like Endra, which is an AI-enabled vertical software for MEP (mechanical, electrical, plumbing) engineering, acquire services firms eventually (analogous to Harvey and Legora becoming law firms).
What I find attractive about these businesses is that they often become deeply embedded with customers. If you’re responsible for maintaining mission-critical infrastructure inside a data center, the customer is not shopping the contract every six months. Reliability and trust matter more than price. Separately, there is a real shortage of licensed professionals who understand and can maintain these physical systems, which is analogous and related to the growing demand for HVAC professionals nationally.
The broader data center services ecosystem already exceeds $50 billion globally and is growing rapidly alongside AI infrastructure spending. My guess is that most investors still underestimate how large the services layer around AI infrastructure ultimately becomes and the opportunity to be a consolidator in the space.
Right-of-Way Acquisition, Land Surveying, and Infrastructure Permitting
This is probably the least glamorous category on the list, which is exactly why I like it. Everyone talks about building data centers. Nobody talks about the businesses required before construction can even begin.
The United States is trying to build more transmission lines, power generation, renewable energy infrastructure, broadband networks, and transportation infrastructure. Every one of those projects requires surveying, permitting, environmental review, easement negotiation, and right-of-way acquisition.
Land surveying alone represents a roughly $12–15 billion market and consists of approximately 35,000 firms nationwide, the vast majority of which are small local operators. The adjacent permitting and right-of-way acquisition ecosystem is similarly fragmented. I bundle these together since the workflows are related and a rollup consolidating multiple players across these segments can create interesting cross-sell/upsell opportunities. All these subsectors benefit from meaningful regulatory barriers to entry. Survey work must be signed off by licensed professionals who are licensed on a per-state basis. Permitting often requires deep local jurisdictional expertise, and right-of-way projects involve navigating complex legal, engineering, and governmental processes.
AI can automate much of the labor-intensive work surrounding them including document review, title and deed abstraction, easement analysis, parcel research, permit application preparation, and municipal code interpretation. In many firms today, highly trained professionals still spend a substantial portion of their time on administrative and research-heavy workflows rather than judgment-intensive work. AI creates an opportunity to dramatically increase throughput per licensed employee, allowing firms to scale revenue without scaling headcount at the same rate.
Employee Benefits and Benefits Administration
Benefits brokerages may be one of the most consistently underrated businesses in America. The industry isn’t particularly flashy, but the underlying economics are fantastic. Revenue is recurring, customer retention is high, and relationships are deeply embedded.
Benefits brokers advise employers on health insurance, dental, vision, life, disability, and other employee benefit plans, while benefits administrators handle enrollment, eligibility management, compliance, carrier communications, employee support, and ongoing plan administration.
Much of the industry’s work remains highly manual and document-driven, creating significant opportunities for AI to automate quoting and carrier comparisons, plan design, enrollment processing, compliance monitoring, employee communications, support inquiries, and renewal preparation.
What makes the category particularly interesting today is that complexity continues to increase. Healthcare costs continue rising and regulatory requirements continue expanding. Employers are being asked to navigate increasingly complicated decisions around plan design, compliance, enrollment, and employee communication and are thus increasingly outsourcing this work. There are also meaningful moats as benefits brokers must be licensed on a state level. The broader employee benefits ecosystem exceeds $70 billion annually, and despite decades of consolidation there are still an estimated 30,000–40,000 independent brokerages operating throughout the country.
There are certainly lots of tech-enabled players biting around the ecosystem, including payroll companies like Rippling, next-gen PEOs like Deel, and modern benefits platforms like Nava. Nonetheless, there remains a long-tail of brokers and administrators ready to be consolidated and modernized, particularly those serving the SMB and main street markets.
Closing Thoughts
These markets fascinate me because they collectively represent hundreds of billions of dollars in annual spending, hundreds of thousands of potential acquisition targets, and decades of underlying growth tailwinds. These are also industries where labor remains one of the largest constraints on growth and profitability and there is a paucity of licensed professionals able to do the work. But most importantly, these industries impact the backbone of our economy and improving them will push our country forward in meaningful ways.
The most exciting AI rollups won’t simply buy businesses and cut costs. They’ll use AI to fundamentally change how those businesses operate, improving the gross margin profile, unlocking meaningful growth, and delivering superior experiences for customers. My guess is that many of the most important AI companies built over the next decade will not look like traditional software companies at all.
If you’re building one of these, I’d love to hear from you!