A Risky Match?
As regional bankers lean into private debt matchmaking — what could possibly go wrong?
Banking has held our attention for a long time — from advising on the Chase Manhattan/Chemical Bank merger during the consequential consolidation moments of the 90’s, to sitting alongside Goldman Sachs in 2008 as the financial plumbing froze, to helping Midwest regional banks find their footing when the ground shifted. Same discovery every time: the risk everyone thought had moved was still in the room. It had just changed its name.
Your banker has a new job.
Did anyone tell you?
There's a banker you know well. Experienced, sound, been in the market for fifteen, twenty years. They know which companies are growing, which founders keep their word, which deals have a meaningful story behind that deck. You'd take their call on a Sunday.
Over the last decade, their job changed. Not loudly — there was no announcement. They used to lend from their own balance sheet, carrying the risk through the full life of the loan. Now they make introductions.
THE NEW MID-MARKET BANKER
They source the deal — a $40M, $60M corporate loan — using relationships no out-of-state fund could replicate. They know the borrower's third-quarter number before the CFO files it. That local intelligence is genuinely irreplaceable.
However now, they hand that loan off to a “strategic partner”. A private credit fund provides the actual capital. The bank keeps the origination fee. The loan leaves the bank's books, and regulatory obligations are deftly dodged.
This isn't cynicism — it's creative adaptation. Rising deposit costs on one side, tightening capital requirements on the other. But creative adaptations create new exposures. And this one creates an exposure almost no one is naming out loud.
When the person who knows you best is paid to make the introduction rather than live with the outcome — and the fund that took the loan depends on a credit line from the very bank that just handed it off — where, exactly, did the risk go?
That credit line — the private credit fund's operational lifeline, logged on the bank's books as low-risk — is the hidden cord. In calm conditions it barely registers. It's only visible when something pulls on it hard.
Then AI walked in
and changed the calculus.
Private credit funds moved into lending to software companies for a perfectly rational reason: these companies generate contracted, recurring revenue on multi-year terms. You can model it. Unlike a manufacturer with inventory risk or a retailer with foot traffic, a SaaS vendor's revenue felt almost mechanical in its predictability— fee by fee, seat by seat, month by month.
That popular SaaS model had two assumptions baked in so deeply that no one bothered to write them down: namely (1) that investing the time and money to create the software was a durable strategic “moat”, and (2) the humans using the software would still be employed.
The SaaS playbook is being rewritten as AI-driven coding slashes the barriers to entry, turning what used to be sophisticated features into cheap, easily replicated commodities. For incumbents, survival is no longer about out-building the competition, but about ruthlessly defending the only moats that still matter: proprietary data integration, structural inertia, and enterprise-grade trust.
Generative AI is eliminating seats — not eventually, but in current renewal cycles. Most business software is priced per person. When a company automates its customer support team or its finance analysts via AI, it doesn't call the software vendor to celebrate. It calls to renegotiate.
Not all software is equal in this brave new world. Tools priced on headcount doing repeatable work — Asana, Monday.com, Zendesk — are directly in the line of AI-fueled fire. Compliance and audit infrastructure — Workiva, Veeva, Archer — is better insulated, because AI can eliminate the people but not the compliance/legal obligations they were fulfilling. A credit portfolio holding loans against both looks identical from the outside. Same revenue figures. Same loan terms. Completely different story underneath.
AI won't destroy enterprise software. It may change the competitive environment, quite rapidly. AI almost certainly will destroy the pricing model — and the loan terms written against that model will find out the hard way.
How it moves.
Including the part that looks like good news.
THE SEQUENCE
(I) Renewals soften faster than anyone modeled. Software companies hit renewal season with customers carrying 20–30% smaller teams than the prior year. Revenue doesn't collapse — it contracts, account by account, deal by deal. Loan terms start tripping. Fund managers flag it internally. They've seen worse. They draw on the bank credit lines designed for this moment.
It works. Cash stabilizes. A carefully measured note is sent to investors.
(2) The collateral behind those lines changes character. The credit lines are secured against future funding pledged by pension funds and endowments — institutions now navigating their own pressures, deferring commitments, quietly reweighting. The security the bank thought it held turns out to be contingent on conditions no one stress-tested. What was classified as safe is no longer. It just hasn't been reclassified yet.
By the time this exposure becomes visible on a balance sheet, the window to move has already closed.
In 2008, the gap between "the system is functioning" and "the system has seized" was measured in weeks. Software renews on quarterly cycles. How much runway does that actually leave?
A CONSIDERATION
From the outside, the whole arrangement looks elegant. Local banker who knows the market. Private fund with patient capital. Institutional investors providing the backing. Each party doing precisely what the system intended. The loan gets made. The fee gets earned. The risk, officially, has moved.
The dominoes are standing, in the shadows. Their alignment is elegant. But the risk everyone thought was offloaded is still in the room—it just changed its name, again.
And trust hangs in the balance.