The dominant story about AI right now is greenfield. Founders build new companies on top of new models, capture new markets, and watch the incumbents lose. The story is partly true. It is also incomplete. The other half of the story is the one almost no one is telling: a meaningful share of the most valuable AI-era companies are already alive. They were built ten, twenty, sometimes thirty years ago, and the AI rewrite reaches them through acquisition rather than launch.
What incumbents already have that startups have to build
A mature business that has compounded for two decades has a set of assets that are very expensive to recreate from scratch. The greenfield narrative undervalues these — partly because they are invisible from the outside.
- Customer relationships. Trust accumulated over a decade is not replaceable by a year of growth marketing. It cannot be bought; it can only be inherited.
- Distribution. Search authority, brand recall, channel partnerships, and word of mouth that took fifteen years to build are unavailable to a startup pitching the same market.
- Data. A long-running product has accumulated structured behavioral data that an AI-native rebuild can train and personalize against immediately. The new entrant has to wait for it.
- Operating discipline. A team that has shipped through real cycles has reflexes a new team will spend a decade developing — pricing, support, renewal, hiring filters, the small operating details that decide whether scale is healthy.
- Cash flow. A profitable mature business is its own funding source for the rebuild. No outside capital, no dilution, no external pressure on the rebuild timeline.
What incumbents typically lack is exactly what AI is now cheap to provide: a modern intelligence layer in the product, a modern automation layer in the operations, and a modern architecture underneath both. The asymmetry runs in the opposite direction of the dominant narrative.
Why the startup playbook will not catch up to it
A founder building greenfield can ship features faster than an incumbent in the first year. By year four, the incumbent that has been re-platformed is shipping at the same pace while still owning the customer base, the data, and the distribution. The startup is pouring growth capital into a channel the incumbent owns for free. That dynamic does not appear in pitch decks. It shows up in retention curves at year five.
The structural advantage compounds: the rebuilt incumbent gets stronger as the AI wave continues, because every additional year of the wave widens the moat that started with two decades of customer relationships.
What we look for in an acquisition target
- Durable customer love. A recurring-revenue base that has held through several macro and product cycles. Not a hot quarter — a long curve.
- Real margins. The business already operates with software economics. We are not interested in the version that needs a turnaround before it can be modernized.
- A surface that AI rewrites. The product has obvious places where intelligence collapses cost or expands capability — search, classification, automation, personalization, content generation — that the previous owners had no reason to build.
- Founder fatigue or transition. Not in the negative sense — in the structural sense. The people who built the company over two decades are often ready to hand it to operators who can run the next twenty years. We want to be those operators.
- Mispriced by the market. Public software multiples and venture multiples both miss the value of mature private software being re-platformed onto AI. The dislocation is the entry.
Why this is not a turnaround thesis
Turnarounds are about fixing companies that do not work. Acquire-and-rebuild is about extending companies that already work into the next era. The first is operationally intense and high-risk. The second is operationally intense and high-leverage — because the underlying asset is already proven.
Two-decade software companies that have been running on legacy stacks are not failures. They are unfinished. The finishing is what we do.
The proof point we already operate
Business in a Box is one of these companies. Twenty-five years of operation, 250,000+ paying customers, real margins, and an underlying product surface that AI can rewrite end to end. Every assumption in this thesis is one we are testing against a company we run. The rebuild is happening now.
The rest of the portfolio will look like that. Some of it we will build. Some of it we will acquire. The acquired companies will, on average, be twenty years older than the built ones — and just as central to the next twenty years as the ones we start from scratch.