Holding companies for the AI era
Founders are told to focus: one problem, one product, one company. That advice still holds for a single team. But AI is changing how much of a company's repeated work a small team can carry, and that makes a different structure practical: a few real businesses, built and operated on one shared foundation.
Why focus made sense
The case for one company at a time was never about ambition. It was about fixed cost.
Every company needs the same groundwork: a legal entity and its compliance, an engineering stack, a brand, testing and release discipline, and a way to reach customers. Some of it can be bought or borrowed, and venture studios have long shared parts of it across the startups they launch. But for most teams, each new company still meant months of setup and a handful of specialists before the first customer.
At that price, a second company was a distraction from the first. Focus was the rational answer to a cost, and that cost is starting to move.
What changed
Shared services are not new. What AI changes is how much of their accumulated knowledge a small team can put to work.
A legal playbook, a design practice and a testing discipline created value long before AI. With AI, the same team can draft from established patterns, adapt proven code to a new product, generate and maintain tests, and handle routine operations at a pace that used to need a much larger staff. The advantage comes from experienced people and repeatable processes, with AI as the multiplier, not from AI alone.
The foundation still has limits. Some work can be carried from one company to the next; some has to be done fresh every time.
| Capability | What can be shared | What stays specific to each business |
|---|---|---|
| Legal and formation | Document patterns, workflows, counsel relationships | Structure, jurisdiction, obligations, risk |
| Engineering | Tooling, infrastructure, architectural patterns | Product requirements and technical trade-offs |
| Brand and design | Design practice, research methods, production tools | Positioning, identity, customer experience |
| Quality assurance | Testing tools, release standards, monitoring | Product-specific risks and acceptance criteria |
| Go-to-market | Research methods, operating tools, lessons learned | Customer relationships, positioning, sales execution |
The aim is not more capital per company but less repeated work per company, while carrying forward what each earlier business has taught us. That is what makes a small holding company of operating businesses practical.
The model
A holding company for the AI era creates companies, owns them and operates them on one foundation. It has three layers.
- The foundation. Legal, engineering, design, quality and go-to-market capabilities, owned by the holding company and improved with every business it runs.
- The companies. Each one has its own product, its own customers, its own brand, its own revenue and costs, and its own accountable results. Shared capabilities support a business; they never substitute for a viable business model.
- The operating discipline. Shared standards for how products are built, tested, launched and measured, so lessons move across the portfolio instead of staying inside one team.
The order of work matters. We test ideas before committing substantial resources to a new company. Where practical, we look for evidence of a real problem, willingness to pay and repeat use before making that commitment. Once we commit, the goal is a durable business, not another launch.
How it differs
The lines between holding companies, venture studios and company builders are not clean, and many share capabilities in similar ways. The real differences are who builds, who stays, and for how long.
| Question | Venture studio | Conglomerate | Our model |
|---|---|---|---|
| Who builds | Typically the studio, then a founder team it recruits | Typically no one: businesses are acquired | The holding company's own team |
| Who runs it after launch | Typically the founder team, often with outside investors | Separate management per business | The same team, on the shared foundation |
| Ownership horizon | Typically spin out and share in the outcome | Typically long term, managed for cash flow | Long term, built and operated |
Three commitments follow from that. We build, rather than only invest. We stay, rather than hand a company off after launch. And we share capabilities where they help, while each business keeps the freedom to choose differently when its customers require it.
Where it breaks
The model has real failure modes, and they are worth naming.
- Attention runs out before money does. Shared infrastructure lowers the cost of a company, not the cost of judgment. Too many companies, and none gets the focus it needs.
- The foundation can become a ceiling. A shared stack that suits five companies may not suit the sixth. Each company needs the freedom to leave the shared path when its customers demand it.
- Brands can blur. Customers buy from a company, not from its parent. Each company needs its own name, its own voice and its own reason to exist.
- Selling does not scale like building. AI helps most with making products. Earning customers is still slow, human work, and it has to be done again for each company.
- Some companies should end. A portfolio needs the discipline to close what does not work and carry its lessons forward, rather than keep it alive out of loyalty.
Shared infrastructure lowers the cost of a company, not the cost of judgment.
The defence against all five is the same: fewer companies than you could run, a foundation you are willing to break, and an honest scorecard for each business.
What it means for partners, talent and investors
For partners, it means one relationship that can reach several businesses. A supplier, a distribution partner or an enterprise customer can start with one company and extend to others on the same standards.
For the people who join, it means building more than one product over a career without starting from zero each time. The work moves between companies; the foundation and the colleagues stay.
For investors, the unit of analysis changes. The question is not only whether one company wins, but whether the shared foundation produces better businesses once its full costs are counted: less repeated work, disciplined capital, and evidence of durable customer demand.
The opportunity is not to start as many companies as possible. It is to build a few businesses well, reuse what genuinely helps, and stay accountable for the rest. AI can reduce the cost of repeated work. It does not remove the responsibility of running a business.