How solo founders can ride the $5 trillion AI roll-up wave
Wall Street is buying small businesses and running AI agents inside them. Here's how to build your own one-person version of that playbook.
An accounting firm in Bellingham, Washington just ran 7,000 tax returns through an AI agent and cut prep time so dramatically that interns who used to grind through returns now spend their time reviewing them instead. That firm isn't a startup. It's a decades-old regional accounting practice that got bought out, plugged into an AI agent, and turned into a case study for one of the biggest wealth transfers in American history.
Over the next decade, roughly six million small business owners in the United States are going to retire. By 2035, about six million small and medium-size businesses will face ownership transitions as baby boomers retire, with more than one million firms viable candidates for sale, representing up to $5 trillion in enterprise value. Most of these owners have no succession plan. Many of these businesses will simply close rather than sell, taking jobs and decades of institutional knowledge with them.
That's the setup. Here's the part that matters for anyone reading this on a laptop instead of from a corner office at a venture fund: the same AI tools reshaping this wave are available to you right now, for a few hundred dollars a month, not a few hundred million in committed capital.
What is actually happening in the $5 trillion opportunity?
Two of the most prominent venture firms in the world have already built the playbook, and it's public knowledge at this point. General Catalyst, one of the largest venture capital firms in the world with roughly $40 billion under management, has dedicated $1.5 billion to buying accounting firms, call centres, property managers, and IT service providers. They're not investing minority stakes and hoping for the best — they're not investing in them, they're buying them outright, then rebuilding their operations with AI.
Joshua Kushner's Thrive Capital is running a nearly identical strategy through a separate vehicle. Thrive Capital launched a dedicated vehicle, Thrive Holdings, with over $1 billion to pursue the same strategy of buying and rebuilding legacy service firms with AI. The scale moved fast: Thrive Holdings raised $2 billion at a $12 billion valuation from OpenAI, SoftBank, D1 and Altimeter to buy legacy services firms and inject AI into their workflows.
The results so far are the part that gets people's attention. At Larson Gross, one of the accounting firms Thrive acquired, AI built on OpenAI's Codex processed 7,000 returns in a single tax season while accountants saved 31% of their time on average. Zoom out and the pattern holds across other verticals too. General Catalyst's property management platform, Long Lake, reports 18 property management acquisitions and $100 million in EBITDA in under two years. In the UK, a similar play in residential lettings dropped problem resolution times from 50 days to 20 days and doubled EBITDA margins wherever AI was fully deployed.
The mechanics behind the margin jump are worth understanding, because they explain why this isn't just hype. The funds build the agents and software first, buy a trusted firm, run the agents in the background, then move the back office over — so revenue holds while costs fall, and margins can climb from 5-10% to 30-40% if the thesis holds. On the deal structure side, the incentives are built to keep the people with client relationships in the building: General Catalyst companies typically pay 60 to 70% in cash and the founder rolls about 30% into equity, giving the person with the client relationships a reason to stay through the transition.
Why does this create an opening for solo founders?
Here's the thing nobody at a $2 billion fund wants to admit out loud: the technology stack they're using isn't proprietary. It's not some secret internal model built by a hundred PhDs. The tax filing agent at Thrive's accounting arm was built jointly using Codex, OpenAI's coding agent — the exact same tool available to anyone with a ChatGPT Plus subscription.
That's the whole unlock. A solo founder with a laptop, a Codex subscription, and a willingness to learn how small business acquisitions actually work has access to the same core technology that a $2 billion fund is using to rebuild accounting firms. What the fund has that you don't is capital for a $50 million roll-up. What you have that the fund doesn't is the ability to move on a single $400,000 bookkeeping firm or a two-person property management shop without a partner meeting to approve it.
You can go get Codex directly from OpenAI and start there — it's designed to reliably complete tasks end to end, like building features, complex refactors, migrations, and more, powered by OpenAI's frontier coding models. The official quickstart guide walks through setup across desktop, CLI, IDE extension, and cloud environments, and the GitHub repository has the installation commands if you want to run it from the terminal.
How would a one-person holding company actually run?
Think of this less like "day trading with AI" and more like buying a small, boring, profitable business and quietly making it more profitable using agents instead of new hires. Here's a rough shape for how that week actually looks in practice.
The folder structure. Every acquired business gets its own directory with a consistent internal layout: client records, financial statements, standard operating procedures, and a running log of what the AI agents are doing inside that business. This isn't glamorous, but it's the difference between an agent that understands context and one that's guessing.
The agent files. Rather than one generic assistant, you build task-specific agent configurations — one for bookkeeping reconciliation, one for client email drafts, one for scheduling, one for report generation. Each has its own instructions file describing exactly what "good" looks like for that specific business, written the same way you'd onboard a new employee.
The human approval rule. This is the non-negotiable part. Nothing reaches a client without a human checking it first. Full autonomy sounds efficient until an agent sends a client the wrong tax figure or promises a maintenance appointment that doesn't exist. The rollups that are actually working keep a human in the loop on anything client-facing, at least for now — even the biggest funds running this playbook admit you still need people reviewing before things go out the door.
The average week. Mornings go to reviewing what agents flagged overnight — reconciliations that didn't match, client emails drafted and waiting for approval, scheduling conflicts the agent caught. Afternoons go to sourcing: talking to business brokers, reviewing financials on potential acquisition targets, building relationships with owners who are quietly thinking about retirement but haven't listed anything yet. One or two days a week go to actually improving the agent setup itself — refining prompts, adding new automations, tightening the SOPs that feed the agents context.
If you want a broader map of which industries are being targeted and by whom, this AI roll-up playbook breaks down the sector-by-sector landscape in more depth than most venture blog posts bother to.
What are the strongest arguments against this?
It would be dishonest to write this up as a guaranteed win, and the skepticism deserves real airtime.
"Roll-ups always fail." This is the most common pushback, and it's not baseless. Traditional private equity roll-ups have a long history of overpaying, over-leveraging, and destroying the culture that made the original businesses valuable in the first place. Critics argue this is just private equity wearing an AI sticker, and there's a real version of that critique worth sitting with.
The results are still early and unstressed. A meaningful caveat applies to nearly every impressive number in this space: many of these figures come from young companies raising money that have yet to face a recession, so they're a strong signal of direction rather than proof. Nobody has run this playbook through an actual downturn yet. Margins that look great in a growth environment can behave very differently under stress.
Integration risk compounds with speed. Buying multiple firms per year sounds impressive until the previous acquisition isn't fully absorbed yet. Moving fast on acquisitions while still stabilizing the last one is exactly the kind of overextension that has killed roll-ups long before AI entered the picture.
The dependency question. For funds relying on a close relationship with a single AI lab for their core technology, there's a real question worth asking: what happens if that relationship changes? For solo founders, the equivalent risk is smaller in dollar terms but just as real — building your entire operation around one vendor's agent tooling means you're exposed if pricing, access, or capability shifts under you.
The most serious version of the argument against solo AI roll-ups is simply that this looks easy from the outside and isn't. Acquiring a real business involves due diligence, financing, legal structuring, and managing actual employees and clients who don't care how good your agent prompts are if the invoice is wrong. AI can compress the operational workload dramatically. It doesn't remove the parts of running a business that were always hard.
Where does someone actually start?
If the acquisition side feels too big a leap for a first move, start smaller. Pick one underperforming, unglamorous process — invoice reconciliation, client onboarding, scheduling — inside a business you already have access to (yours, a friend's, a family member's) and build a single, well-documented agent for it using Codex or a comparable tool. Get comfortable with the human-approval workflow before you ever think about buying anything. For background on the market itself, the McKinsey Institute for Economic Mobility's full report on the ownership transfer is the primary source almost every fund in this space is citing, and it's worth reading in full before deciding this is a market worth entering.
Six million business owners are heading for retirement whether or not anyone builds an agent to help. The only real question is who's paying attention while the deals are still small enough for one person to close.