Deal File No. 008: What if the AI company bought the customer?

An insurance broker is an unusual thing to buy if your big idea is software.

Unless the customers are the hard part.

Sequence and Michael Dell’s family office have agreed to take Baldwin private in a $7.7 billion transaction. Announcement

A software startup usually builds a product, then tries to persuade an established business to use it. Sequence is approaching the problem from the other direction: invest in the established business and rebuild how it works.

What if the better AI business isn’t another tool sold to an insurance broker, but the insurance broker itself?

That is a much more interesting bet than “AI will save time.”

It also has a catch. Saving time and making money are different accomplishments. Warren Buffett learned a version of that lesson in a textile mill.

We’ll get to the mill.

Sequence Holdings and DFO Management are buying a majority interest, with eligible Baldwin colleagues retaining a significant minority stake. The offer is $32.50 a share. Enterprise value includes approximately $4.6 billion of equity purchase price and $3.1 billion of net debt assumed or refinanced. Closing is expected in Q1 2027, subject to approvals. Baldwin

Baldwin advises clients on risk and distributes insurance. Sequence is a permanent holding company combining investment with embedded engineering.

The obvious AI investment is a company with a clever product.

The less obvious one could be a company with established customers, experienced professionals and a backlog of expensive work.

If capable models become widely available, access to the model becomes less distinctive. The scarce part shifts toward knowing what to do with it, having permission to change the process, and having customers who will pay for the result.

Consider the ingredients. Code can be copied quickly. A client's willingness to trust your advice takes longer.

The opportunity is to combine them.

Sequence calls its approach “refounding.” Its founders describe engineers working alongside operators, using its Atlas platform to redesign established service businesses. Their explanation

An owner can participate in the improvement of the whole business rather than charge for one piece of software. It also gets the whole implementation problem.

At the roughly 20 times trailing adjusted EBITDA multiple described in Baldwin’s announcement, that is an expensive place to discover your product is really a consulting project. Valuation basis

Sequence’s early BankSouth account gives us something more useful than an AI mission statement.

It reports reducing loan origination from nine stages to three, cutting application-to-closing time by approximately 52%, and supporting loan volume that doubled year over year with the same team. These are Sequence-reported results, not an independent estimate of AI’s contribution. September update

Picture the application passing through nine stations. You could give everyone a faster computer. Or you could ask why the application needs nine stops.

Some stops prevent expensive mistakes. Others may exist because two systems cannot talk, two teams need the same information, or nobody has revisited the process.

The prize is not removing six boxes from a presentation. It is removing unnecessary work without losing necessary judgment.

Now picture an illustrative insurance renewal. A client sends a file. Someone copies the information. Someone checks the copy. Someone chases a missing detail. An adviser finally gets something usable.

The client did not hire a relay team. They want the right cover.

AI could make the relay faster. A redesigned workflow could eliminate repeated entry and give the adviser a specific exception to resolve.

Better software helps people do the work. Better design asks which work should still exist.

A conceptual workflow, not a measured before-and-after at Baldwin. Fewer redundant handoffs, not fewer safeguards.

Baldwin announced a firm-wide Claude rollout in May, following earlier deployments. This is not an analog business waiting for someone to show it a chatbot. Anthropic announcement

That is a reason to take the partnership seriously, not dismiss it. An ambitious management team can be a better starting point than a turnaround disguised as a technology thesis.

But it changes the test. Sequence needs to help Baldwin move beyond what its existing team was already on course to achieve.

And ownership alone is not the answer. BankSouth’s partnership involved a minority investment and embedded engineers, not a takeover. BankSouth announcement

The useful combination is authority, expertise and shared incentives. A controlling stake is one way to assemble it. It is not a substitute for it.

In his 1985 shareholder letter, Buffett explained the trap in Berkshire’s textile business.

New equipment could reduce costs. Individual investment proposals looked attractive. But competitors could make similar improvements. Once enough did, the savings fed into lower selling prices rather than durable profits.

His description was blunt: “their reduced costs became the baseline for reduced prices industrywide.” Berkshire’s 1985 letter

The machinery could improve while the investment remained poor.

Insurance advice is not commodity cloth. Relationships, judgment and service can differentiate one broker from another. The parallel is narrower: a productivity gain is not necessarily an advantage if competitors can reproduce it.

For Baldwin, the question is not simply whether AI lowers the cost of serving an account.

What allows Baldwin to keep some of that benefit?

It could go to customers through better service or lower advisory fees where those apply. Producers could negotiate for a share. Technology and oversight could absorb it. Shareholders get what remains.

The spreadsheet that counts every saved hour as profit has skipped the negotiations.

There are three ways I would look for the operating thesis to pay off:

The third is the most interesting.

Suppose a smaller account needs too much expert attention to justify the revenue. Reducing the preparation and administration could change that calculation. The company might sell useful advice to someone it previously could not afford to serve.

That is a growth thesis, not just a headcount thesis.

But spare capacity needs somewhere to go. An adviser who gets two hours back still needs a qualified prospect, a suitable insurance solution and a client willing to act.

And the difficult cases do not disappear because the easy ones become automatic. If exceptions pile up, you get a faster acknowledgment and a longer wait for a useful answer.

A dashboard can celebrate both. A client knows the difference.

There is another prize inside the holdco model: reuse.

If Sequence solves a recurring problem once, can parts of the solution work across its other businesses? Shared tools for handling documents, permissions and exception review could reduce the effort required for the next implementation.

But banking and insurance are not interchangeable. The customer promises, decisions and controls still need domain expertise.

The test is what travels.

If every acquisition requires a completely bespoke rebuild, growth demands another team of engineers each time. If useful components and operating knowledge carry over, later investments can benefit from work already paid for.

The first implementation demonstrates capability. Reuse tests whether that capability can compound.

That is where I would look for the distinction between an AI holding company and a collection of businesses with expensive technology projects.

Three questions, in this order:

My bet is that the winners will combine something new with something difficult to replace: better ways of doing the work with customer relationships worth protecting.

Baldwin is a compelling place to test that combination. The company already has an AI program. Sequence has a theory of how ownership and engineering can accelerate it.

Now the theory has a purchase price.

Buffett’s mills are the reminder: better technology can be real, useful and still a disappointing investment.

The job is not just to make the work cheaper. It is to build a business that gets to keep the difference.

Know someone who would have a view on this deal? Forward them this issue.

Received this from a friend? Subscribe to The Multiple.