AI in lending can shut down sneaky liability management schemes

The golden age of liability management transactions — the decade-long run of aggressive structural engineering and contractual gamesmanship — is drawing to a close as of the second half of 2025.

For much of the 2010s and early 2020s, the outcome of a credit negotiation depended less on pricing than on structure. Borrowers with inventive legal advisors and a high tolerance for reputational risk could exploit definitional ambiguity, reclassify assets and subordinate existing lenders, often entirely within the four corners of the contract.

Consider J.Crew’s 2016 masterstroke: The retailer transferred its valuable trademark to an unrestricted subsidiary. In other words, they moved the company’s most valuable asset — the ability to call something “J.Crew” — outside the original credit group and beyond the reach of existing lenders. That intellectual property was then used as collateral for new financing that leapfrogged existing creditors in the capital structure. The maneuver allowed J.Crew to raise fresh capital while leaving original lenders holding significantly devalued claims.

J.Crew didn’t invent this sort of financial engineering; PetSmart had a similarly tricky 2013 maneuver, involving selling a nominal sliver of equity in Chewy (just enough to classify it as a “non-wholly owned” subsidiary) and subsequently shifting its most valuable assets out of the credit facility’s reach. With a 1% sale, PetSmart effectively stripped 100% of Chewy’s value from creditor protection. It was an elegant exploitation of definitional loopholes.

What made such tactics viable in the first place was their asymmetry and novelty. No law firm, lender or sponsor had a complete view of how these structures functioned across the broader market. Every deal was a negotiation with partial information. You could throw in some language based on past financial skullduggery, but how could any company hope to preempt all of the potential creative flavors of liability management engineering? 

The maturation of AI-native infrastructure, capable of extracting and contextualizing structural precedent across thousands of historical transactions, means that the informational playing field is flattening. It is now possible to track, in real time, which clauses are gaining traction; which maneuvers have fallen out of favor; and where pressure points in the documentation are likely to emerge. 

What we’re seeing in the ways these deals’ structures have evolved reflects this new AI-driven transparency and understanding: Anti-PetSmart provisions are near-ubiquitous. “J.Crew blockers,” provisions that prevent companies from moving valuable assets beyond creditors’ reach, surged from 11% of deals in the first quarter of 2023 to 36% in the second quarter of 2025. Pro rata sharing provisions — which ensure equal treatment of lenders in repayment scenarios — now appear in 85% of syndicated loan deals. Lien subordination requirements, designed to prevent silent reprioritization of creditor claims, are present in 61% of new issuances. This reflects a systemic calibration of lender protections, the backlash to the golden age of borrower-led structural arbitrage. The message is clear: The days of asset shuffling without consequence are numbered.

However, as structural loopholes close, borrowers are receiving greater economic flexibility in return. What’s emerging is a more rational, more stable set of trade-offs: transparency in exchange for accommodation. Borrowers who accept tighter documentation are frequently rewarded with improved pricing, accelerated execution and room to maneuver within agreed parameters. This is a fundamental market rebalancing in action. 

Nowhere is this clearer than in the evolution of EBITDA addbacks. These provisions allow borrowers to reflect projected operational improvements as present-day earnings for covenant purposes. Since the first quarter of 2023, EBITDA cost-savings addbacks have remained present in approximately 40% to 45% of deals. But in the second quarter of 2025, the generosity of those addbacks became unmistakable: They were uncapped in 30% of the deals that included them, and exceeded a 20% cap in more than 80% of cases.

This, viewed at scale, is a useful signal. And it is only observable through tools capable of aggregating and analyzing historical deal data across thousands of transactions. The structural horse-trading taking place in real time, this convergence of legal discipline with economic flexibility, would not be visible without AI-native analytics. Armies of associates couldn’t spot this from the ground.

The result is not the end of creativity in structuring. The dopers will always try to outrun the tests. But the days of plausible deniability are over. What was once a game of strategic opacity is becoming a contest of visibility. Distressed borrowers may still attempt novel maneuvers, but recycled playbooks won’t cut it. The market has seen them all.

Until someone invents something truly new, the golden age of liability management is over.