No Priors

Michael Lee on why AI can crown the re-founded incumbent

Michael Lee· Co-founder and CEO at Sequence Holdings
·~43 min·English·No Priors
AgentsAI CompanyBusiness Strategy
TL;DR

Michael Lee argues that in the industries where incumbents hold structural advantages, AI's biggest winner will be a re-founded incumbent, not a startup, and that owning the business, celebrating the engineer, and redrawing how people work is what turns that moat into a lasting lead.

01Core Mental Model

Put the Engineer on the Pedestal

Michael Lee argues that every company celebrates one persona and hires the best of it, so if alpha now comes from engineering and AI, you have to rebuild the culture around the engineer.

In a world where you believe that alpha comes from engineering and AI, you need to create a culture whereby the celebrated persona is the engineer.

— Michael Lee, No Priors
Key Insight
The choice of celebrated persona is a recruiting magnet, not decoration: it decides which scarce talent you can aggregate at all, so betting that alpha comes from engineering forces the engineer to the center of the company rather than bolted on the side.
02The Founding Kernel

Buy the Incumbent, Inherit the Moat

He splits the economy into industries AI will not touch, industries a startup will win, and industries where the incumbent holds every advantage, and Sequence goes after that third group by acquiring the incumbent itself.

I felt very strongly that if you could acquire the right incumbent and inherit the advantages of being that incumbent could forge the market leader

— Michael Lee, No Priors
Key Insight
Most AI investing chases the second bucket by funding the startup; Lee's contrarian bet is the third, where brand, scale, network effects, or regulation favor the incumbent so heavily that a startup struggles to break in, so buying that moat lets AI amplify advantages instead of attacking them.
03The Core Argument

The Three Options Fail, So Own It

A CEO who wants to re-found the company is left with three options, doing it alone, hiring a services firm, or buying software, and each fails for a structural reason, which is why Lee's answer is to own the business outright.

Services companies optimize for kind of three things. Getting in your wallet, staying in your wallet, growing the share of your wallet. It is a path towards incrementalism.

— Michael Lee, No Priors
Key Insight
Each alternative fails by design, not effort: talent will not join to be second-class, a vendor is paid to keep you dependent, and packaged software is priced to be generic, so only ownership aligns the time horizon and incentives with a full transformation.
04What Transformation Means

Redraw the Line, Don't Just Speed It Up

Lee's sharper point is that bolting AI onto each existing job only speeds up the assembly line you already have, while the real opportunity is to reorganize the whole line around what machines can now do.

You can't sell a product around a new human assembly line that doesn't exist today

— Michael Lee, No Priors
Key Insight
The trap in 'better than nothing' software is that a vendor can only sell to the workflow that already exists, so it quietly freezes the org in its current shape, and the gain that actually matters requires changing the org chart, which only an owner can do.
05Why a Bank

Regulation Is a Feature, Not a Bug

The regulated nature of a bank, with its well-defined operations, clean data, and clear rules, is exactly what agents need, which is what made banking, a sector private equity has traditionally avoided, the ideal first pilot.

the regulated nature of the bank was a feature, not a bug.

— Michael Lee, No Priors
Key Insight
The reframing is the point: the same rule-boundedness that makes a bank feel rigid is what gives an agent a reliable, legible environment to act in, so regulation is not an obstacle to automation but the precondition for it.
06What Makes a Target

Organizational Physics

Lee looks for organizations that are dense and centralized, because then anything built once at headquarters is amortized across every branch, the opposite of a fragmented rollup that must be rebuilt site by site.

We like organizations that are quite dense and the operations are centralized. So anything you build can be advertised over a large base.

— Michael Lee, No Priors
Key Insight
'Organizational physics' is really a selection filter: Sequence favors businesses where one central build propagates across many branches, unlike a fragmented rollup that has to integrate different systems and standardize procedures and cultures site by site.
07The Platform

Atlas, Built in Four Layers

Because roughly 80% of any business is the same underneath, Sequence built Atlas, a data ontology, an agent builder, an orchestration engine, and an app builder, so the core built at the bank is reusable at every company it buys next.

if you break down the business to its atomic units 80% of it largely homogeneous and 20% of it is vertically specific.

— Michael Lee, No Priors
Key Insight
The stack mirrors how serious agent systems are actually built: you cannot orchestrate agents over a business the models cannot read, so the ontology layer at the bottom, making the company legible in code, is what everything above it depends on.
08The Proof

Do More With Less at Bank South

Six months in, consumer loan underwriting time was down 94% and commercial loans went from 30 days to 11; when Q2 loan volume happened to double, a rise Lee did not credit to Sequence, the bank handled it with the same overall headcount.

what we've done with the tools is that we've allowed the organization to be able to take on significantly more loan volume than was ever possible before with the same amount of headcount.

— Michael Lee, No Priors
Key Insight
The numbers are less about speed than about redeployment: the underwriting team actually shrank as one person retired and another moved to the front office, so the same people now spend their time on exceptions and customers instead of typing numbers, the human-engineering payoff Lee calls the hardest part.