AI Disruption Is Now a Credit Question

For most of the AI boom, the debate about winners and losers belonged to equity investors. Lenders could afford to pay less attention: a loan gets repaid as long as the borrower keeps generating cash, and most borrowers kept doing so. In 2026 that separation broke down. When fear that AI would undermine software business models hit public software stocks early in the year, attention quickly turned to the large volume of private loans made to software companies, and to whether lenders had accounted for the risk that some of those businesses could shrink. AI disruption has become a credit question, and lenders now have to answer it.

Why lenders are exposed

A loan is a claim on a company's future cash flow. When lenders financed software companies, many did so on the basis of recurring revenue and high customer retention, sometimes before those companies were consistently profitable. That approach assumes customers will keep renewing at similar or higher prices for the life of the loan. If AI lets customers do the same work with fewer licenses, or lets a new competitor offer a cheaper alternative, renewal rates and pricing can weaken before profits show it. By the time the financial statements reflect the problem, a lender's options may be limited.

The risk is not limited to software. Business process outsourcing, call centers, document processing, basic IT support, and other businesses that sell labor by the hour face the same question from a different angle. If an AI agent can do the work, customers will expect to pay less for it.

Telling resilient borrowers from exposed ones

Not every software or services business is at risk, and broad-brush conclusions are as dangerous as complacency. In our underwriting, we look for the characteristics that make a business resilient. It is the system of record a customer's operations depend on, not a tool layered on top. It holds proprietary data that makes its product better. It operates in a regulated field where compliance, auditability, and trust matter more than price. It is tied to the physical world, delivering care, moving goods, or maintaining equipment, so its value cannot be reduced to text on a screen. And it has a credible plan to use AI itself, so that falling costs become margin rather than lost revenue.

The warning signs are the inverse: per-seat pricing for tasks that agents can perform, little proprietary data, low switching costs, a customer base under its own pressure to cut spending, and a management team that treats AI as a marketing slide rather than an operating priority.

How it changes the structure of a loan

Underwriting disruption risk is not only about choosing borrowers. It also shapes how a loan is built. For businesses with meaningful exposure, we favor lower leverage, shorter duration, and scheduled amortization, so that risk declines over time rather than resting on a refinancing years from now. We want maintenance covenants and regular reporting on leading indicators such as customer retention, renewal pricing, and usage, not only trailing earnings. And we are skeptical of structures that let interest accrue instead of being paid in cash, which can hide deterioration until it is too late.

We also run scenarios that few lenders would have bothered with a few years ago. What happens to this company if its largest customers cut their usage substantially, or if a competitor offers an AI-driven alternative at half the price? If the loan still works in those cases, we are comfortable. If it only works when nothing changes, we are not.

The opportunity on the other side

AI risk is also an opportunity for lenders who can tell the difference. Businesses that use AI well can lower their cost to serve customers, grow without adding staff at the same pace, and generate more cash, which makes them better borrowers. Many essential-industry companies, including revenue cycle firms, specialty manufacturers, and field-service businesses, are early in that improvement. Financing them through it, with structures that reflect both the opportunity and the risk, is work we find compelling.

The Kiron take

The lesson of 2026 is that no sector is safe simply because its historical numbers look stable. Lenders now have to form a view on how technology will change a borrower's business over the life of the loan, the same way equity investors always have. Our team invests in AI and has run businesses in the sectors we lend to, and that combination is how we underwrite: the business first, the structure second, and an honest view of how AI changes both.