Berkshire Hathaway is the gold standard for long-term compounding. Buffett is characterized by his long hold periods across the tokenized “great companies at good prices”. He famously focuses on value companies with enduring long-term growth, rather than the near-term get rich quick opportunities. The strategy behind one of his more critical initial investments, National Indemnity, serves as part of the thesis behind this piece, but more importantly, if Warren Buffett were 35 today, starting fresh in a world shaped by artificial intelligence, I’m not sure he would build the same thing.

We’re living in a new investment era, one where intelligence, not just capital, is the dominant input. That changes the opportunity set for what a holding company can be. We no longer need to simply acquire and hold great businesses. We can acquire good ones — overlooked, analog, operationally inefficient — and make them great by quietly embedding intelligence. This is quite different than the Buffett strategy of finding a diamond in the rough, but rather a story of many companies that, with the right tools, can become their own best-of-breed.

Imagine buying a third-party insurance administrator or a regional HVAC company. These aren’t sexy businesses, but they’re profitable, under-digitized, and operated by teams often overwhelmed by paperwork, scheduling, claims processing, or quoting complexity. These company profiles lack basic technologic integration. Anecdotally, we were speaking to an Insurtech company that built their business on integrating APIs into insurance carriers as it takes 3–6 months to update basic client information with a carrier. Insurance has been around for centuries and constitutes a massive global TAM, yet their technological processes have not changed and kept up with currently available systems. It is within industries like these that this holding company strategy has an opportunity to thrive. Buying companies for the long-term and via change management, transforming them into leaner, more efficient, and more productive enterprises that are directly implementing the changes that the legacy competitors have not.

With just a modest AI deployment — a fine-tuned routing optimizer, a back-office copilot, or an underwriting model — you could drive 20–40% gains in EBITDA without touching the brand, culture, or headcount. The center of gravity in this model isn’t a bloated ops team. It’s a small, specialized internal AI unit, like your own lean Palantir, that builds reusable infrastructure. Think of it as a shared OS for your portfolio: LLM-based assistants for finance, customer support, and legal; ML tools for cash flow forecasting or dynamic pricing; secure model deployment frameworks so every acquisition can opt-in to modernization without having to build anything themselves.

Industries like insurance and services businesses are especially ripe. In the insurance sector, most MGAs still rely on legacy systems and Excel-based workflows for underwriting, claims handling, and policy administration. This creates the optimal environment for AI to drive efficiency and accuracy. LLMs and NLP tools can automate the ingestion of unstructured documents (think claims emails, accident reports, or application forms) turning them into structured data for instant triage. Underwriting models can be trained on historical loss data, combined with third-party sources like property imagery, credit data, or weather forecasts to make faster, more accurate risk assessments. Claims processing can be streamlined with AI that detects fraud, classifies severity, and even drafts denial or approval communications automatically. AI can make these workflows faster and cheaper overnight. You don’t need to change what these companies do. You just help them do it better. And while most industries have already digitized, insurance remains deeply manual, which makes the operational uplift from AI not incremental, but exponential.

The modern holding company, built from scratch today, doesn’t look like a leveraged roll-up or a messy conglomerate. It’s not a venture fund or a tech accelerator. It’s something in between: a vehicle that acquires durable, cash-flowing businesses and systematically improves them using applied AI. Not disruption, but rather optimization. Investing in the recipient of technology rather than the technologies themselves.

Importantly, in industries like this, investing in the product or the service does not mean immediate adoption. These industries have not changed for a reason; they are resistant to change due to their decentralized nature. Therefore, in order to reap the benefits of technological uplift, the best way to do that is to invest in the recipient of the change, rather than the provider.

This model also changes how capital is used. Warren Buffett’s acquisition of National Indemnity gave him access to investment capital via float. Strategic investment gave him “cost-free” investment capital to enhance his investment returns. Importantly, this could not have been done as effectively without his long-term investment focus. When you plan to exit an investment in 5 years, your range of strategic priorities becomes focused on short-term fixes, and your risk tolerance is tempered by your exit strategy. Therefore, in this new age of holding company, the play isn’t to buy and flip; it’s to buy and compound. You raise flexible, long-term capital and use cash-generating acquisitions to fund the next wave. Float and earnings drive the flywheel. AI makes every dollar more productive.