The first AI take-private is a $6.3 billion bet that services can grow like software

Elad Gil opens No Priors with a number that is hard to miss. Long Lake Management, co-founded by Alexander Taubman, has agreed to acquire American Express Global Business Travel, the world’s largest corporate travel platform, for $6.3 billion. Gil thinks it may be the first AI take-private: buy a public company in order to remake it with AI. The firm had already bought around 30 companies on the same premise. The tool underneath is Nexus, a horizontal platform Taubman says they have been investing in since the beginning.

Roughly 80% of the infrastructure is shared across verticals. The rest is deployment: mapping workflows, understanding and cleaning data sources, integrating them so models can actually reach them. Nexus sits between the models — the stack is model-agnostic — and the company’s data, skills, and workflows. That takes applied AI engineering, which Long Lake has built in-house. The first acquisitions took more than a year before AI showed up in business outcomes. Now, Taubman says, they can partner with a company and deploy in days.

Gil asks whether that means instant margin lift in a couple of weeks. Taubman corrects the sequence. What shows up immediately is time savings. The question is what you do with the hours. Long Lake invests them in growth and customer experience rather than cost-cutting, which he admits is a narrative violation. AI, in this view, is positive-sum: more productive people are people you want more of, happier customers mean you grow faster, and you create jobs. HOA businesses that were growing 0 to 5% a year in volume when Long Lake invested are now growing more than 20% organically, which he says makes them the fastest-growing company in that industry. Better products at incrementally lower cost also improve customer-acquisition economics. It is the software go-to-market playbook, dropped into sleepy industries.

If employees are dramatically more productive, Gil asks, can they even go work somewhere else? Retention across the acquisitions is very high. The stated vision is to be the best place to work in every industry they enter: the best people, the best tools, the best customers. Leave for a competitor and you pick the mundane work back up — 25 or 30% of the day. Taubman compares it to giving up email. That is how they become a talent magnet, and how they can pay the most: the people are the most productive, they make more money, and Long Lake is delighted about that. Give those people AI superpowers and customers get happier too. Response times fall, errors drop in board reporting, budgeting, and email, and customer retention rises.

The Silicon Valley alternative would be to find the HOA niche, build software, and sell it as a vendor. Long Lake buys the company instead, for deeper alignment. Own the customer relationships and you own the results; a software company, even a good partner, does not live with the business outcome. Internally the inversion is complete. Engineers treat field employees as the customer. Taubman reaches for the old skunkworks picture: put the engineers in the factory. His engineering team is probably in 20 states at any moment, sitting with people in architecture, HOA, HR services, and specialty tax, then building a tool inside Nexus for the pain point they just watched.

Gil’s summary is the adoption bottleneck everyone else keeps hitting. Change management — processes, organizational design — is the hard part of getting AI used. If you own the company you can change those things. If you only sell software, you mostly cannot. The 25 to 30% of the day that is busywork is the list worth automating first, whether that is HOA budgets and board packets or, in travel, disruption handling when a flight cancels.

Gil has talked to dozens of people trying some version of an AI roll-up. Almost none of them have all three competencies at once: a private-equity-style motion for purchasing companies, an engineering team that can build the AI stack, and real change management. Long Lake was purpose-built as a cross-functional company with technology, change management, and M&A in the DNA from day one, which is how they recruited from their network. Taubman says 100% of the first 20 people were people they already knew well, from shops such as Palantir, Ramp, Robinhood, and Glean.

Rasmus, co-founder and CTO, came through an early investor and board member the group had known for 15-plus years. Gil notes how rare it is for business people to have deep technical networks; technologists hire business people badly early on, and vice versa, so founding teams mismatch. Many of Long Lake’s engineering co-founders had already been CEO or CTO of an applied-AI startup. They had tried selling software into services industries, found it brutally hard, and took the other fork: if you cannot beat them, join them. Taubman wants the firm to be a place where entrepreneurial applied-AI engineers do their best work — "a team of founders."

The M&A side comes from the other world. Manny, a co-founder and M&A lead, left GTCR after working on what Taubman says was the most profitable deal in that firm’s history the month before, and was named employee of the month the month he resigned. Others came from Blackstone, TPG, and H.I.G. Those firms are extraordinary at what they do, Taubman says; they are not AI-native. For the subset of deal professionals who believe the thesis, there are not many places that look like Long Lake. The thesis itself is a gap: labs and hundreds of billions are making models better every day, trillions more will follow, and someone has to make that work in the real economy.

That combination produced the $6.3 billion attempt to take private American Express Global Business Travel, a company Gil notes is on the order of 110 years old. There is only so much Taubman can say while the deal is in progress. He and his investors have been customers. Gil’s joke is that it started when the only things you could book were railways and boats, and that it predates air travel. Taubman confirms it: 111 years old, founded in 1915 by American Express to get travelers-check customers out of Europe during World War I. Carlson Wagonlit, which Amex GBT acquired late last year, goes back further. Wagonlit, he says, is French for sleeping cars on a train, and that business dates to 1876. These are franchises that have already navigated a century of technology change. He thinks they will be extraordinary for another.

Long Lake picks industries with a prepared mind: a whiteboard of 15 to 20 high-value sectors, with travel always on it. The reasons are specific. Corporate travel is mission-critical, the cost of failure is high, and most trips are revenue-generating — miss a recording with your favorite podcaster, Gil offers, and you have lost a whole travel day. A hundred years of customer trust is the asset. The company has already, in public disclosures, charted an AI transformation; Long Lake’s plan is to double down on that rather than rip it up. In the press release Taubman has already said the quiet part: Nexus should give team members superpowers, faster response times, faster disruption resolution. "Imagine your travel counselor with AI superpowers." That is the future he wants for Amex GBT’s customers.

Traditional private equity, in Gil’s sketch, buys for a few years, loads the company with debt, cuts costs, and flips it. Long Lake is closer to the Berkshire mold: buy what you can hold. The services TAM Taubman cites is more than $20 trillion, and the ambition is to be market leader in every segment they enter. The companies he names as inspiration are Danaher and TransDigm: a differentiated operating model, better growth, better customer satisfaction, better employee retention, then years of consolidation. Doing this in a year or two is a fantasy. Better tools attract better people, who can be paid more, who deliver better customer outcomes, who let you grow faster — a two-to-five-year cycle. Build the best company in the industry and then sell it? "That just doesn't make sense to me." He wants to own it forever and compound the advantage for decades.

That posture is also how they win auctions, sometimes as the only buyer a seller wants to talk to. Permanent capital is already rare; permanent capital plus an AI platform you can deploy on day one is something these industries have never seen. Real enterprise AI use cases, he guesses, are about 1% penetrated. Ninety-nine percent of American businesses are small and do not have big-company resources; even the big ones are struggling to get maximum impact from models. The pitch is a cross-functional team that lives in your office for two years — he names engineers, Varun, Taras, Pratik, Jason — plus rollover equity so original founders keep a piece of the upside. In the first four verticals, founder rollover has been significant. As the model gets proven, he expects cost of capital to fall, which is how Danaher and TransDigm stopped losing deals on price: operate better, pay more, own a faster-growing platform.

The growth obsession has a bookkeeping reason that founders in labor-intensive services already feel in their bones. Most of those companies are growth-oriented; the people built them over 20 or 40 years by knocking on doors. Growing 20% now means hiring 20% more people, finding them, training them, managing them, and then keeping about 20 cents of every incremental dollar because the rest goes to labor. That is a very high marginal tax on the activity you claim to want. Make the existing team 30 to 40% more efficient and the mindset flips: the company starts growing with software-like incremental margins, which funds more growth. Taubman says you could call any of his CEOs, some of whom have run their businesses for decades, and they would tell you they are having the best time of their careers. Team members are paid more and happier; customers are happier. He grew up in a family that ran a grandfather’s company for 75 years and recently exited. Long Lake was designed as the buyer he would have wanted: a long-term steward of the employees and the customers.