Loan markets have sorted your software vendors into four risk tiers before your next renewal cycle arrives.

By Shashi Bellamkonda · April 27, 2026

The credit analysts got here first. While enterprise procurement teams were still debating whether to build AI clauses into vendor contracts, the leveraged loan market had already sorted the software sector into four distinct risk categories and started pricing each one differently. That repricing is now visible in loan price data that the Wall Street Journal reported on in late April 2026, and it amounts to an unsolicited vendor viability assessment that most CIOs and CTOs have not yet read.

The four categories the loan market is tracking are vertical software, cybersecurity, horizontal software, and software engineering tools. They are not ranked by revenue or market share. They are ranked by perceived survivability in a world where artificial intelligence replaces or substantially automates what each category used to sell. The spread between the best-performing tier and the worst is now roughly eleven cents on the dollar, and it has opened up since January 2026 alone.

Vertical software holds because workflows hold

WSJ loan price data shows vertical software companies outperforming the broader software cohort since January. Names tracked in the loan data include CCC Intelligent Solutions, which processes insurance claims, and Relativity, which handles legal document review. The common characteristic is not industry focus in the abstract. It is regulatory embeddedness and workflow lock-in so deep that an insurer or law firm cannot simply replace the software with a large language model prompt and move on.

This is the pattern I covered in the vertical artificial intelligence post earlier this month: the moat is not the software feature set. It is the friction of leaving. When an industry-specific workflow is embedded in audit trails, compliance reporting, or multi-party data exchange, the switching cost is structural, not contractual. Credit investors appear to have arrived at the same conclusion from a different direction.

RealPage, a property management platform, also appears in the vertical loan data. It serves multifamily rental operators who run leasing workflows, maintenance requests, and rent collection through a single system. That is not a feature any generative artificial intelligence product replaces on a Tuesday afternoon.

Cybersecurity holds for a different reason

Cybersecurity loan prices dropped roughly five cents on the dollar from January through April 2026, making it the second-most-resilient category. Companies tracked in the data include Proofpoint, Sophos, and Gen Digital. The logic here is distinct from vertical software. Cybersecurity does not survive because it is hard to leave. It survives because artificial intelligence is making the threat environment worse at the same rate it is making everything else more efficient.

Attackers are using the same models that defenders are using. Phishing campaigns that once required human authorship are now generated at scale. Synthetic identity fraud is rising. The regulatory response, including updated frameworks from the Cybersecurity and Infrastructure Security Agency and mandatory compliance timelines across financial services and healthcare, is adding institutional demand that does not shrink when budgets tighten.

The credit market has priced this as durable demand, and the debt investors behind these loans carry downside exposure that equity analysts do not, which makes their read worth taking seriously.

Horizontal software faces the harder question

Horizontal software, the category serving broad process functions across industries, dropped roughly 8.8 cents on the dollar in the same period. The loan data tracks companies like Avalara, which automates sales tax compliance, UKG, which handles workforce management, and SUSE, the enterprise Linux vendor. These are not bad businesses. The problem is that horizontal software's value proposition rests on automation and workflow standardization, which is precisely the terrain where large language model agents are making the most aggressive early claims.

A tax compliance workflow that runs on a structured rules engine is easier to prompt-engineer around than a multifamily property management system that integrates with local housing authority data feeds, and that gap in substitutability is widening as agent capabilities improve.

None of this means horizontal software companies are failing. Several have strong cash positions, investment-grade balance sheets, and customer bases measured in the hundreds of thousands. But debt-financed horizontal software companies, the ones owned by private equity with leverage ratios built on revenue assumptions from a pre-generative-AI baseline, face a different calculation. Their debt payments do not adjust when a customer reduces seat count because an agent is handling half the workflow.

Developer tools absorbed the largest markdown

Software engineering tools dropped 16.3 cents on the dollar, the sharpest decline of any category. The data tracks Idera, Perforce, and SmartBear. These are testing platforms, version control tools, and quality assurance infrastructure. The concern is not that software development stops. It is that the developer productivity wave driven by artificial intelligence coding assistants compresses the need for certain categories of tooling that charged per seat or per test run on human-generated code volume.

If a developer using an AI coding assistant produces three times the code output with half the manual testing cycle, the per-seat math on some legacy tooling categories changes, and credit investors appear to believe that compression is already affecting revenue assumptions in leveraged buyout models built in 2021 and 2022.

User review data shows the same fault line

The loan market is pricing switching cost as the primary durability variable. Verified buyer data at SoftwareReviews shows the same pattern from the customer side. Across product scorecards, Plan to Renew scores equal or exceed Likeliness to Recommend scores in roughly 80 percent of cases. Users are more loyal in practice than they are vocal advocates, and the gap between those two numbers is widest in exactly the categories the loan market treats as durable: security, compliance, payroll, and enterprise backup.

The inversion shows up in the categories with lower switching costs. Customer success platforms, tools built specifically to reduce churn for their customers, show some of the weakest renewal intent in the SoftwareReviews dataset. Recommend scores run high. Plan to Renew scores lag by fifteen to twenty points. The product is liked. Leaving it is easy. That combination is precisely what the loan market is discounting in horizontal software.

High renewal intent without high recommendation scores is the stuck-not-satisfied signal. It means users are staying because migration cost exceeds frustration, not because the product is earning their loyalty each cycle. For a CIO evaluating a vendor's long-term viability, that gap is worth reading carefully. A vendor whose customers renew reluctantly is one whose revenue looks stable until a competing platform lowers the migration cost enough to trigger a wave of departures.

The private equity dimension changes the CIO's calculation

The Wall Street Journal analysis and the supporting research from UBS credit analysts point to a specific risk cluster: software companies owned by private equity, carrying high debt loads, whose revenue assumptions were set before generative artificial intelligence became a procurement variable. These are not the same as publicly traded software companies with strong balance sheets. They are businesses where the debt stack is fixed and the revenue line is now uncertain.

For enterprise buyers, this creates a vendor risk category that does not appear on most technology due diligence checklists. A strong product and a fragile capital structure are separate assessments, and the second one determines whether multi-year contract commitments, implementation investments, and data migrations survive a credit event at the vendor level.

Credit markets are not always right. They are, however, priced by people with more downside exposure than most enterprise software analysts. When the loan price spread between vertical and developer tools reaches eleven cents on the dollar in four months, that is a signal, not noise.