Buy The Incumbent: Sequence Holdings And The $7.7 Billion Baldwin Bet

Introduction

Days after announcing the largest private AI acquisition to date — the $7.7 billion take-private of insurance broker Baldwin, alongside Michael Dell's family office — Sequence Holdings co-founder and CEO Michael Lee joined Sarah Guo on No Priors to explain the firm the deal was built around. The result is less a deal story than an argument: that the biggest technological shift of our lifetime rewards incumbent leaders, and that the only structure which lets you actually rebuild one is ownership.

A $7.7 Billion Vote On Incumbents

Sequence is a perpetual holding company founded twenty months ago, pairing a frontier engineering team with management teams to build market leaders. Brokerage was a focus from the beginning: the firm met with dozens of insurance brokers and judged Baldwin truly one of the best. The cadence is as unusual as the structure. Sequence tries to do one deal a year — that is all. There is no deployment clock, no investor group pressing for more, and Lee hopes to hold the equity for the rest of his life. The Dell family office backed the equity commitment, underwrote the technology platform itself, and will co-manage the acquisition vehicle, Baldwin One, alongside Sequence.

The Thesis, Forming Since 2017

Lee's conviction was assembled at Lone Pine, where one of the first areas he covered after joining in 2017 was artificial intelligence — the era of AlphaGo and the first transformer paper, when the dominant architectures degraded as they scaled. When ChatGPT appeared in late 2022, it became obvious to him that the world had changed forever: finally there was an architecture known to scale with compute. As an investor he concluded that AI's impact on the economy would be uneven — restaurants and golf courses untouched, coding going to startups like Cognition and Anthropic. But the business-school intuition pointed the other way: the incumbent leader had all the advantages — brand, scale, network effects, regulation. Inherit those advantages, add engineering, and could you create a market leader? That question is what Sequence is based on.

Little Machines Versus Reorganizing The Company

What most enterprises do with AI today, Lee argued, is a giant push to put little machines at every person on the human assembly line. Nothing wrong with that — it is better than nothing. But machines run 24/7, scale with electricity, and do things no group of people can, so the right response is a different question: how do you start to reorganize what an organization should be to absorb what the technology makes available? In Lee's framing, the industry is living through the equivalent of the next industrial revolution, and he treats the holding-company model less as ambition than as the only natural response — he does not see a second path for the reorganization to happen on its own.

Three Options, Three Structural Failures

The centerpiece of the episode is a Fortune 500 thought experiment: you are the CEO, with unlimited resources, and your job is to reinvent yourself. You have three options, and each fails structurally.

Do it yourself, and you cannot attract the talent the restructuring needs, because every company on the planet has a famous person — usually an investor, which is exactly the wrong persona when alpha comes from engineering. (Blackstone is his example of the investor-as-famous-person; Palantir, whatever Alex Karp would say about the description, is the engineer-honoring organization that sells that capability to firms that cannot hire it.) Hire the multi-trillion-dollar service industry — Accenture, McKinsey, Palantir — and you meet the incentive problem: service firms optimize for getting funds into your wallet, keeping them in your wallet, and increasing your wallet share — the path to gradualism — and cannot reorganize your people or their incentives. Buy software, and you get a beta available to everyone, sold the only way software can scale: you will always be selling the workflow as it is designed today, tuned to the human assembly line as it exists. The only natural way to combine engineering, aligned incentives, and a long horizon is a holding company.

The BankSouth Pilot

The proof of work came from a cold-start problem. A holding company without money cannot do a deal; without money you cannot make a deal, without a deal you cannot hire engineers, and without engineers nobody gives you money. Rather than buy something arbitrary to prove the firm could buy, Sequence went looking for a client — a dear friend, an Avenir co-founder, whose family owns a bank in Georgia. From August 4 (Lee's wedding anniversary, which he missed) to late November the team worked as a service provider on the bank's key workflows, until the family invited them to become a permanent partner. The investment closed in March, after alignment and approval from the Fed and the OCC.

The difference between serving and owning was night and day, Lee said: the engineering depth you take on when you will be there for the long haul, the enthusiasm of employees who see long-term support, and the ability to lay a technology foundation on a three-, five-, ten-year view.

Organizational Physics

Why a regulated community bank as the first test? The regulated nature of the bank was a feature, not a mistake: clearly defined operations and excellent data hygiene work extremely well for agents. Just as important is what Sequence calls organizational physics — a taste for dense, centralized organizations where what gets built amortizes over a large base. All of the bank's underwriting is done centrally, so anything created at headquarters amortizes across every branch. For an organization with over $100 million in sales, the business had one central nervous system and one accounting book — which is why significant progress came quickly, and why the mixed-use roll-up alternative, with its systems integrations and engineers scattered across locations, compares poorly.

Ninety-Four Percent, And Thirty Days To Eleven

The measured results are the episode's hardest evidence. A community bank has three parts — deposit-gathering, loan-seeking, and the underwriting core in the middle — and historically loan volumes grew linearly with middle- and back-office headcount. About six months after the March investment, Sequence had a system taking on all of the organization's consumer loans: average consumer-loan underwriting volume fell 94 percent since they started in March. Commercial underwriting, started in recent months, took the average loan from 30 days down to 11. The organization absorbed significantly more volume with the same staff — and in the second quarter, the bank's loan volume doubled compared to the first quarter, with underwriting standards unchanged and a smaller underwriting team. Lee is careful about causation: he will not say Sequence had anything to do with the doubling. What the tools did do is let underwriters stop keying numbers and work the exceptional loans, while loan officers spent more time in the field than writing loan letters — and no one had to puzzle over whether to finance a boat for someone who clearly has the money.

Why Brokerage Resists Startups

The move from banks to brokers runs through Sequence's three-part screen: market scale (targets worth building past $100 million, at most one a year), industries where the incumbent holds the advantages, and the intersection with what AI does well. Brokerage passes all three. Over $2 trillion in premiums flow to insurers each year, with brokers in the intermediary role. The economics are quietly hostile to startups: insurers have made almost no money from underwriting since the beginning of time and earn their returns by investing, which rewards asset accumulation and paying for distribution; the client does not pay the broker — the insurer does; the business is relationship-based, brokers cannot really compete on price, and retention runs at 90 percent. Baldwin offered the fit: a scalable asset, a centralized technology foundation, and a leadership team ambitious about what cutting-edge technology can do.

Atlas And The 80/20

The reusable core is Atlas, the platform built inside the bank and expected to generalize across industries. Its premise is an observation Sequence's engineers brought from Scale AI and Palantir: break a business into atomic units and about 80 percent of them are largely homogeneous, with 20 percent vertically specific. Atlas stacks four layers — a data ontology that defines the organization and its workflows in code so models can understand the business, an agent builder for high-performance agents on that data, the lattice orchestration mechanism that instruments workflows with those agents, and artifacts, the application builder on top. It is also a developer platform for the operating company's own engineers, and the core infrastructure built at the bank is reusable across Baldwin and any future company.

Not A Fund Business

Lee is explicit that he admires the big private equity firms — he and Guo both passed through Apollo — and expects many to make significant progress on AI transformation; they have the incentives, resources, and capabilities. The limits he names are structural. Fund economics reward finding assets, pricing them attractively, structuring right, and creating enough value to clear the LPs' required return; the horizon reflex follows, and managers find themselves starting to plan the packaging to sell in three years. And a firm designed to celebrate the investor struggles to attract world-class engineers who want to be present when decisions are made — a 25-year-old engineer with a say in the investment profile would be very unusual at a typical large-cap buyout shop.

What They Underwrite In People

For management teams, the first signal is how they play the game in the industry they grew up in — Sequence has great engineers and investors, but Lee has never run a bank or a brokerage, so the operators must be the best in their field. The second is groundwork already laid: small teams in the cloud, centralized data infrastructure, and organization-wide OpenAI or Anthropic already being implemented before it was obvious. Baldwin is the example: Trevor Baldwin was early to run Anthropic across the organization, and the company consolidated onto a single instance of Applied Epic — the insurance broker's primary operating system — which takes real effort, investment, and change.

Bet On Exceptional People

The closing stance compresses the career arc — Goldman Sachs for structuring, Apollo for capital structures and credit documents, Lone Pine for studying the world's best companies and seeing that they compound at rates no one had ever thought of. Asked what early-stage asset managers miss, Lee answered from the founder's seat: ideas are cheap and execution is very difficult, so betting on exceptional people is the only thing that matters — people on hard problems in big markets deserve support even when it seems crazy, because the upside makes up for the rest. It held in public markets too, he said: the pattern from meeting Jensen Huang in 2017 — clarity of thought, incredible execution, the smartest committed people, a constantly reshaped business — repeats more reliably than most people would guess.