AI rollups won't work
VCs discover PE investors' favorite playbook and it still won't work
VCs have become besotted with the promise of layering AI on top of PE-style rollups. The idea is intoxicatingly simple. AI agents automate complex workflows. Many industries, including plumbing, HVAC, accounting, healthcare back offices, etc., are fragmented across thousands of small operators. So why not buy them up, plug in agentic AI workflows, and watch margins expand?
The pitch is seductive:
AI is improving faster than distribution.
If you own the end business (say, a chain of dental clinics), you can deploy AI inside it, capture efficiency gains directly, and avoid the messy task of selling software licenses to reluctant customers.
VCs have billions in dry powder needing a home, and rollups provide deal volume big enoguh to justify writing chunky checks.
On the surface, it looks like arbitrage: buy boring companies, sprinkle AI on top, compound forever.
But this isn’t a new idea. It’s an old playbook dressed in new clothes.
What a PE rollup actually is
Private equity has been doing rollups for decades. A rollup is exactly what it sounds like: acquire lots of small companies in the same sector, consolidate them into a larger platform, systematize back office functions, and capture economies of scale.
For example:
In healthcare, PE funds have bought local dermatology or dental practices and merged them into chains.
In home services, they’ve bought HVAC or plumbing companies across a region to create multi-state brands.
Constellation Software buys dozens of tiny vertical SaaS businesses every year.
The math looks compelling:
Buy 20 firms at 5x EBITDA
Consolidate back-office functions (HR, finance, IT) and push purchasing into bulk contracts.
Standardize workflows, increase margins.
Sell the whole package at 10x EBITDA.
The spread between 5x and 10x, multiplied across dozens of deals, looks like alchemy.
Why PE rollups usually fail
The problem is that most rollups don’t live up to the Excel spreadsheet. The failure points are well known.
Overpaying at entry. PE firms get caught in bidding wars, pay growth multiples, and assume synergies will justify it. When the synergies don’t appear, the debt stack groans.
Integration complexity. Every company has its own ERP system, HR practices, compliance quirks, and customer contracts. Harmonizing them takes far longer than expected. The synergies evaporate while integration costs mount.
Cultural friction. Founders often are the culture. Once they cash out, employees lose motivation and customers defect. Centralizing workflows in a platform company alientates the local staff who actually deliver the service1.
Misaligned incentives. Corp dev teams are rewarded for closing deals, not for integration success. Local managers resist centralization because it strips autonomy. The integration debt piles up.2
Debt & leverage. Rollups are typically funded with debt. If revenue wobbles or integration slows, leverage becomes unmanageable and the equity is wiped out.
Lack of true synergy. Some industries are simply too localized or labor-intensive to centralize effectively. You can’t create meaningful economies of scale if every customer requires custom, on-the-ground execution.
The result of all of this? A long trail of failed dental rollups, botched home-service aggregators, and service chains that never achieved their promised margin expansion.
Why AI is not a magic elixir
Now enter AI. The new argument goes sure, rollups failed before, but this time is different because we have AI agents to automate workflows. Let’s examine why that’s an illusion.
Messy data still rules. Agents are only as good as the underlying data they run on top of. Rollups inherit dozens of incompatible systems and half-broken databases. AI doesn’t erase this; it makes integration even more complex, because now you must standardize both the old workflows and the new agent layer.
Liability doesn’t go away. Even if AI achieves 95% accuracy, regulators and customers demand 99.999% reliability in healthcare, finance, or safety-sensitive fields. That means humans remain in the loop, and the full labor arbitrage never materializes. Worse, liability insurance and compliance audits eat into whatever savings AI provides.
Commoditization kills the moat. If your competitive advantage is we deploy AI agents inside rollups, so can everyone else. Unless you control a choke point—payments, licenses, regulatory custody—you’ll be trapped in a race to the bottom as AI lowers industry-wide prices.
Change management is human, not digital. People cling to old workflows because they encode tacit knowledge, risk controls, and trust relationships with customers. AI can help, but it doesn’t dissolve organizational politics or human resistance to change. Rollups still choke on adoption debt.
Short-term capital vs long-term integration. Venture capital wants a liquidity event in 7-10 years. Successful rollups require decades. AI doesn’t shorten the half-life of integration. It adds another layer of complexity that takes time to stabilize. The mismatch between venture timelines and rollup reality remains fatal.
The core contradiction
The heart of the AI rollup thesis is that technology will erase execution risk. But exeuction risk is the thing itself in rollups. Acquisitions are easy; integration is hard. If you believe AI eliminates integration debt, you are essentially claiming that this time gravity runs upward.
AI can help at the margins. It may speed up invoice processing, call routing, or compliance documentation. But those are incremental efficiencies. They don’t solve the reasons rollups fail: overpaying, leverage, cultural decay, lack of discipline.
The mirage of arbitrage
Venture investors love AI rollups because they promise the best of both worlds: the excitement of cutting-edge tech and the steady cash flows of boring industries. But it’s a mirage. The same structural reasons most rollups fail still apply, and AI does not dissolve them.
The lesson is not never touch rollups. The lesson is that successful rollups are boring. They rely on discipline: conservative multiples, permanent capital, operating playbooks, and ruthless capital allocation. They are the opposite of the VC growth at all costs mindset.
So when you hear a VC brag about AI-enabled consolidation in dental clinics, HVAC services, or insurance claims, remember: this isn’t a revolution. It’s the same old rollup math, now with a chatbot strapped on. And it will likely end the same way: another tombstone in the PE graveyard of failed rollups.
If you enjoy this newsletter, consider sharing it with a colleague.
I’m always happy to receive comments, questions, and pushback. If you want to connect with me directly, you can:
follow me on Twitter,
connect with me on LinkedIn, or
send an email to dave [at] davefriedman dot co. (Not .com!)
Warren Buffett’s Berkshire Hathaway is not a rollup but it has acquired dozens of companies under its conglomerate umbrella. Crucially it retained the founders in virtually all of these deals, and the founders often stayed for decades post-acquisition. And each company operated as an autonomous unit with its own back-office processes remaining intact. It is worth pondering why Buffett’s strategy worked, repeatedly, while many PE rollups fail.
You can see the same failure point with Sandy Weil’s Citigroup. His deputies were incentivized to acquire dozens of banks under a financial supermarket umbrella, but integration of the disparate banks remained haphazard. Even today, decades after Weil retired, his successors are still dealing with the consequences of his poorly-integrated acquisitions. He and his deputies, meanwhile, earned tens to hundreds of millions of dollars from Citigroup while its shareholders have suffered subpar returns across decades.
Well done.
Excellent article.