The Best Lender Is the One Who Does the Work

In 1981 the economists Joseph Stiglitz and Andrew Weiss explained something that looks irrational from the outside. Why don’t banks just charge a higher interest rate to riskier borrowers and lend to everyone?

Because the lender cannot tell the good projects from the bad ones. Raise the rate, and the careful borrowers with safe projects walk away, while the ones willing to pay anything, often because their projects are the riskiest, stay. So the rate cannot do the sorting. Instead banks ration. They say no, or they demand collateral: something they can take and sell if the project fails.

Collateral, in other words, is a substitute for knowledge. The bank asks for your building because it cannot see whether your idea will work. Which means the projects that get starved of credit are exactly the ones whose only real collateral is the result: new systems, new processes, new work that has not happened yet.

Now put the signature house on the other side of the desk. It is not a stranger guessing at your project. It designed the work, it does the work, and it has already guaranteed that the work will meet a standard or it pays. It knows more about whether the project will succeed than anyone alive, including you.

That changes the whole problem Stiglitz and Weiss described. The signature house does not need collateral as a substitute for knowledge, because it has the knowledge. And the loan does not have to be repaid from your building. It can be repaid from the very outcome the house has promised: the recovered revenue, the saved hours, the cut costs.

This is not a fantasy. There is an industry that already works exactly this way. Energy service companies go into a building, replace the lighting, heating and controls, pay for all of it up front, and guarantee the energy savings. The owner pays nothing at the start. The company is repaid out of the savings, and if the savings fall short of the guarantee, the company covers the gap. The work, the guarantee and the financing are one contract, and the collateral is the result.

Manufactured intelligence is about to make that structure possible for almost everything.

The instinct to hesitate here is right. There are two real dangers, and either one can sink a house.

The first is double exposure. An insurer survives by spreading risk across many unrelated things. A house that guarantees a project and also lends against it has done the opposite: if the work fails, it pays out on the guarantee and loses the loan at the same moment. Two losses, one cause. The structure only works because the house controls the thing that decides whether the project succeeds. The moment it lends against anything it does not control, that protection is gone.

The second is the seller who lends. A firm that finances its own sales is always tempted to sell more than it should, because the loan makes every deal easier to close. We have watched this movie. General Electric built a finance arm to help customers buy its machines. Over time the finance arm grew into something close to half the company, lending far beyond anything GE built, and when the 2008 crisis hit it needed the government’s emergency debt guarantees to stay upright. GE spent the following decade dismantling it.

So yes, the signature house lends. But only a disciplined one survives it, and the discipline fits in one sentence.

Lend only against your own signature.

Finance the work you do and the outcome you guarantee. Never lend against someone else’s work, never lend for its own sake, and never let the lending grow faster than the work. The house that keeps that rule has the best credit book in the world. The house that breaks it becomes GE Capital.

1. “Nothing up front, paid from results” becomes the standard offer for AI work.

   Especially for small businesses, which are exactly the borrowers banks ration hardest. The owner who could never get a loan for software will be able to get the software financed by the people who build it.

2. Signature houses become the biggest financiers of small-business technology.

   Not by competing with banks for ordinary loans, but by financing the one kind of project banks cannot see into: work whose only collateral is the result.

3. Three prices collapse into one.

   Today a project has a price for the work, a separate price for insurance and a separate interest rate. The signature house quotes one number: a monthly amount, tied to the outcome, that bundles the work, the guarantee and the cost of money.

4. The first great failure is a house that lent beyond its signature.

   Some house will grow its lending faster than its work, start financing things it does not control, and discover double exposure the hard way. It will look exactly like GE Capital, and it will teach everyone else the rule.

5. Banks answer by partnering, not competing.

   The bank has the cheap money. The house has the knowledge. The natural deal is the bank supplying the balance sheet while the house supplies the underwriting, and eventually a large bank buys a signature house outright to get what it cannot build: the ability to see inside the project.

Charleston AI already does the smallest version of this. We do the work, the customer receives it, inspects it, uses it, and then pays. Every engagement is a short loan from us to the customer, repaid once the result is proven, and it is safe for one reason only: it is a loan against our own work. We know what we built and whether it does what we said.

That is why I think the rule matters more than the ambition. Pay-after works because the lender and the one who did the work are the same party. Stretch that into financing, and the same rule has to come along, or the whole thing turns into something else.

What happens when a bank buys one of these houses and turns its branches into local signature houses is the subject of the next piece, The Bank That Knocks First.

This extends the set: The Signature Is the Product, Talent Will Gather Where Risk Is Priced, and The Signature House.

Already true

Banks ration credit to projects they cannot see into and lean on collateral instead. Energy service companies already combine work, guarantee and financing and are repaid from results. Manufacturers have financed their own customers for a century. And pay-after service providers already extend short credit on every job.

What has to happen

Signature houses have to exist with real capital behind them, and their record of guaranteed outcomes has to be good enough that lending against it is safer than lending against a building. Once the record is there, the financing follows.

Where I am probably wrong

Regulators may not let a firm do the work, guarantee it and lend against it all at once, because the concentration of risk is exactly what bank regulation exists to prevent. If so, the lending ends up with a partner bank rather than the house itself. And the temptation I described is real: it is possible the discipline simply does not hold, and the sector learns the rule only after a very public collapse.