Re-Founding Incumbents for the AI Era with Sequence Holdings Co-Founder and CEO Michael Lee

Summary

• Sequence Holdings, founded just 20 months ago, is partnering with the Dell family office on what Sarah described as the largest private AI acquisition to date: the family office’s $7.7B acquisition of Baldwin. Lee’s core bet, formed when ChatGPT appeared and made scalable architectures seem real, is that AI’s impact will be uneven: startups win some verticals such as coding, but incumbent leaders retain brand, scale, network effects, and regulatory advantages that can be paired with frontier engineering to build market leaders.
• Lee’s structural case for why incumbents struggle to transform themselves is the episode’s sharpest framework. DIY faces a talent and culture problem because “every company on the planet has a person of note”—at Blackstone, that person is the investor—but “in a world where you believe alpha comes from engineering and artificial intelligence, you need to create a culture where the person of note is the engineer.” Service providers optimize for wallet share, “the path to gradualism”; buying software means buying beta that “always sells the workflow as it is designed today.”
• The differentiation from PE is deliberate scarcity and duration: “We try to do one deal a year. That’s all.” Sequence is not organized around a fund deployment cadence or an investor group pressing it to invest, versus typical buyout math where “you’re usually trying to figure out how to start preparing packages to sell this thing in 3 years.” Lee says he hopes to hold Sequence equity “for the rest of my life.”
• The BankSouth pilot delivered hard numbers: average consumer-loan underwriting volume down 94% since March, the average loan timeline cut from 30 days to 11 after commercial-loan underwriting was implemented, and Q2 loan volume doubled versus Q1 with a smaller underwriting team and no change in underwriting standards. Counterintuitively, “the regulated nature of the bank was a feature, not a bug”—clearly defined rules and clean data hygiene “work extremely well for agents.”
• The Baldwin/brokerage thesis rests on structural protection from startups: more than $2T in annual premiums, insurers that have “made almost no money from underwriting since the beginning of time” and earn the bulk through investing, 90% retention, and brokers whose clients don’t pay them—making price competition difficult. Sequence screens for “organizational physics”: dense, centralized operations where anything built at headquarters amortizes across every branch.
• The reusable asset is Atlas, a four-layer platform (data ontology, agent builder, “lattice” orchestration, “artifacts” app builder) built on the observation that 80% of a business’s atomic units are homogeneous across industries. But Lee is candid that the binding constraint isn’t code: “We have an engineering problem and a human engineering problem, and the human engineering problem is much more complex.”
• Lee’s closing investing lesson, informed by Goldman, Apollo, and Lone Pine, is that “ideas are cheap, and execution is very difficult”: back exceptional people in big markets. His best specimen is meeting Jensen Huang in 2017—consistency, clarity of thought, and constant reshaping of the business are the traits he says matter across early- and late-stage investing.

Deep dive

1. The founding insight: AI’s impact is uneven, and incumbents hold the winning cards

• Lee’s origin story runs through covering AI at Lone Pine from 2017—AlphaGo, the first transformer paper, and an era of CNNs, GANs, and LSTMs that showed “degrading effects” at scale. When models such as ChatGPT appeared in late 2022, the calculus changed: “the world had changed forever… finally, we had an architecture that we knew could scale infinitely” with compute.
• His resulting map of the economy: some sectors AI will not affect at all, such as restaurants and golf courses; some sectors startups win—coding, where “Conviction and Sequence are going to buy an outsourcing coding business… seems like a terrible idea, but you would give money to Cognition or Anthropic.” But across huge swaths, “the incumbent leader had all the advantages”—brand, scale, network effects, and regulation—and the question became whether inheriting those advantages plus frontier engineering could create market leaders.
• The transformation he’s after is organizational, not incremental: most enterprise AI today is “this giant push to put little machines in every person on the human assembly line,” whereas machines that run 24/7 and scale with electricity demand that organizations be reorganized around what the technology can do.

2. Why Lee thinks incumbents struggle to transform themselves—the three-doors argument

• Door one, DIY, faces a talent and culture problem: “every company on the planet has a person of note,” and at Blackstone that person is the investor. If alpha now comes from engineering, “you need to create a culture where the person of note is the engineer”—which Lee says is essentially why Palantir exists: an organization that honors world-class engineers and wraps that talent for companies that cannot hire it.
• Door two, service providers, fails on incentives: Accenture, McKinsey, and Palantir “optimize their costs for three things: getting funds into your wallet, keeping them in your wallet, and increasing your wallet share. This is the path to gradualism”—and vendors cannot change how people are organized, who works in the organization, or what the incentives are.
• Door three, buying software, fails twice—“if it is available to everyone, then it is available to everyone,” and more subtly, software vendors must optimize a uniform workflow to sell broadly, meaning “you will always be selling the workflow as it is designed today,” the existing human assembly line. Lee argues that ownership provides alignment, an engineer-celebrating culture, and a horizon matched to the transformation.

3. A perpetual holding company, not a fund—one deal a year

• The structure follows from the job: matching capital duration to sustained operational commitment, orienting culture toward “creating market leaders, not attracting capital,” and using retained earnings to support existing companies and attract new investment. Against typical PE—which Lee, an Apollo alum, says he admires—the fund structure “incentivizes you to invest” and points toward “packaging to sell this thing in 3 years.” Sequence has no deployment cadence: “We try to do one deal a year. That’s all… if that means we don’t do anything this year, that’s great.”
• His engineer-retention critique of buyout shops is specific: his engineers “want to be present when decisions are made,” and “it would be very unusual to find a 25-year-old engineer who has any say in the investment profile” at a large-cap firm. Application-software developers also tell him their business model is independent of what the model does: better model performance gives Sequence more tools rather than creating existential risk.
• Lee says he hopes to hold Sequence Equity “for the rest of my life,” emphasizing how different that horizon is from a typical fund investment.

4. BankSouth: the cold-start pilot that produced the numbers

• The origin was a bootstrap problem—“without money, you can’t make a deal. If you don’t have a deal, you can’t hire engineers”—solved by calling Avenir co-founder Jamie Reynolds, whose family owns a Georgia bank. The work began on August 4, which Lee remembers was his anniversary and that he missed it. By late November, the family asked Sequence to become a permanent partner; the investment ultimately closed in March after Fed and OCC approval.
• Guo recalled reacting with “a little horror at the complexities and regulatory issues” because banking has not traditionally been a popular private-equity sector. Lee’s inversion: “the regulated nature of the bank was a feature, not a bug”—clearly defined operating rules and excellent data hygiene “work extremely well for agents.” The bank’s centralized operating structure also fit Sequence’s “organizational physics”: underwriting was centralized, so everything built at headquarters could amortize across the branches.
• The results: average consumer-loan underwriting volume down 94% since March; after commercial-loan underwriting was implemented, the average loan timeline fell from 30 days to 11; and Q2 loan volume doubled versus Q1. Lee preserves the caveat—“It’s pure luck. I’m not going to say that Sequence has anything to do with it”—while saying the bank handled the volume with unchanged underwriting standards and a smaller underwriting team. The team was smaller because one person retired and another moved to the front office.
• The operating goal is not simply fewer people: loan officers spend more time in the field than writing loan letters, and underwriters focus on complex exceptions rather than rote processing. The next roadmap is new products, new businesses, better tools for loan officers and relationship managers, and a differentiated customer experience.

5. Baldwin and the brokerage moat—plus what Sequence looks for in management

• The screen is: a market large enough to matter, an incumbent with meaningful advantages, and overlap with what AI does well. Sequence targets companies it believes can become worth more than $100 million, and Lee says “you won’t do this 10 times a year.” Brokerage clears all three: more than $2T in annual premiums, insurers that have “made almost no money from underwriting since the beginning of time” and earn the bulk through investing, 90% retention, and clients who do not pay the broker. Because brokers cannot really compete on price, the industry is difficult for startups.
• Lee’s point about insurers is that investment income encourages asset accumulation, making a high-quality growth business especially valuable and helping explain the brokerage industry’s strength. Baldwin offered a scalable asset, a centralized technology foundation, and a world-class leadership team that is ambitious about what current technology can do.
• On management, Lee is explicit that he has no experience managing a broker or a bank, so Sequence relies on management teams to be the best operators in their industries. It looks for teams that have already shown technological initiative—working in the cloud, centralizing data infrastructure, or implementing OpenAI or Anthropic across the organization before the outcome was obvious.
• The transcript describes Trevor Baldwin as having been early to run “Entropic” at Baldwin, but that wording is ambiguous. The company is working on a single instance of Applied Epic, the insurance broker’s primary operating system. The Dell family office’s Michael Dell and Dan Betar, its head of global private equity, worked with Sequence to underwrite the deal, its technology platform, and its technology-transformation plan.

6. Atlas, human engineering, and the pure-venture coda

• Atlas, built inside the bank, rests on an observation from engineers with backgrounds at Scale AI, Palantir, and elsewhere: break a business into atomic units and “80% of them are largely homogeneous and 20% are vertically specific.” Its four layers are data ontology (“how do we simply define the organization and the movement of this business in code”), agent builder, “lattice” orchestration, and “artifacts” application builder. The core infrastructure is intended to be reusable across Baldwin and future companies.
• The harder problem is people: “we have an engineering problem and a human engineering problem, and the human engineering problem is much more complex.” The bank taught Sequence to be “hypersensitive” to employee anxiety. The playbook is to elevate employees’ work, remove monotonous rote tasks, use the best aspects of being human, and create a sense that the organization is winning.
• In the closing investing exchange, Lee says that if he ever returned to private equity—which he says he will not—he would look for “exceptional people working on complex problems in large markets.” Ideas are cheap and execution is difficult; exceptional people can find the right support and make something work even when the idea initially seems crazy.
• Guo calls that the purest approach to venture capital. Lee agrees and points to meeting Jensen Huang in 2017: clarity of thought, consistent execution, surrounding himself with committed talent, and constantly reshaping the business to compete in a changing market.