Watchlist Management and Loan Workout Workflows

The CRE Finance Council's mid-year conference, held in New York from June 8 to 10, 2026, put a number on something credit teams had already been living with for months: maturity dates have stopped functioning as automatic payoff dates. Stack those years against 2025, and industry estimates put total CRE maturities at more than $1.5 trillion across the three-year stretch. Those are loans sitting on desks right now, waiting to find out whether they get refinanced, restructured, or handed to a special servicer.

What makes 2026 different from the last few years of "extend and pretend" is that lenders have stopped treating extension as the default move, not just the size of the pile. Mike Fratantoni, MBA's SVP and Chief Economist, said as much in the association's February 2026 release, noting that lenders are no longer simply extending loan terms. Office-backed debt makes up roughly $148 billion of the 2026 maturity total, and multifamily maturities are expected to grow substantially from 2025 into 2026. None of this is exotic. It is simply volume, arriving all at once, at a moment when the old playbook of rolling a loan forward and hoping for better conditions next year no longer applies. A single workout team used to handling a dozen distressed loans a quarter is now looking at the same math problem multiplied across hundreds of positions, and that is a different kind of problem than the one most workout infrastructure was built to solve.

Hidden assumption failures revealed by the office workout crisis

Office properties have become the proving ground for a failure that predates any of the macro pressure described above. These reports were not necessarily fraudulent in the legal sense. They were optimistic documents written for a transaction that needed to close, and optimism has a way of aging badly over a seven- or ten-year loan term.

But the more interesting detail is this: equity-side and lender-side property condition reports, prepared on the same asset, often tell different stories. Why does this matter beyond the office sector specifically? Because it reframes what a "bad loan" actually is. It is tempting to treat office distress purely as a story about remote work and falling occupancy, and that story is true as far as it goes. But the PCR gap shows that the problem is also informational: capital assumptions baked in at origination were never tested against the property's actual condition, and nobody circled back to check until the loan was already in trouble. That is not a market problem but a documentation problem, and documentation problems compound quietly for years before they surface as a crisis.

The operational bottleneck: data access, not deal judgment

Ask a workout officer what is hardest about managing a large book of distressed loans right now, and the honest answer is rarely "deciding what to do. Experienced credit teams generally know the menu of options: extend with conditions, restructure the capital stack, push toward a sale, bring in a special servicer. The result is that the same documents get reviewed multiple times, by different people, at different stages of the loan's life, with no shared structured record connecting one review to the next.

That fragmentation is tolerable at small scale. One might argue that more staff solves this. It does not, not durably: workflows that function adequately for ten loans become unworkable at fifty and break outright at two hundred, and the traditional fix of hiring proportionally more analysts does not scale inside a stress cycle that is compressing timelines on every loan at once.

The deeper issue is that the quarterly review cycle is the wrong unit of time for a problem that moves continuously. The data exists somewhere in the system. It is just not visible to the people who need it, at the moment they need it, which is the entire problem this piece is building toward.

Technical default and why catching it early matters more than payment status

Here is a distinction that gets underweighted even by experienced practitioners: a loan can be current on every single payment and still be in default. A technical default occurs when a borrower breaches a covenant, DSCR, LTV, occupancy, or another financial test, regardless of whether the check cleared on time. That breach can trigger a cash sweep that redirects property income away from the borrower, default interest that compounds the borrower's cost structure, or in some loan structures, acceleration of the entire outstanding balance. None of these consequences requires a missed payment. All it takes is a number crossing a line written into the loan agreement, often a line nobody is actively watching.

The practical cost of discovering this late is leverage. Continuous AI monitoring addresses this directly: by reading loan agreements and testing every ongoing covenant as new data arrives, it can flag a breach weeks before it would otherwise register. For most commercial real estate borrowers, covenant tracking still lives in a spreadsheet updated once a quarter, so problems only become visible after the lender happens to notice them sitting in a cell that has already turned red.

Continuous, AI-powered covenant monitoring across the loan lifecycle

If the diagnosis is that information arrives too slowly and from too many disconnected places, the fix has to replace the periodic review cycle with something that runs continuously across every loan position at once. That is a categorical shift in capability, not an incremental improvement on the spreadsheet. At the center of it sits automated covenant monitoring: DSCR, LTV, occupancy, and other financial covenants calculated in real time as new data flows in, with portfolio managers alerted the moment a threshold is approached or breached. When a breach is detected, automated escalation workflows document the response from the moment the signal fires, building an audit trail as a byproduct of the work itself.

The same approach extends to construction loan draw management, where each draw cycle normally requires reconciling a borrower's funding request against approved budgets, completed work, inspector certifications, and lien waiver documentation, a reconciliation process that AI can match automatically. Debt yield, NOI variance, reserves, and sponsor performance get tested continuously against current market signals, so credit teams can see which loans need attention now rather than discovering the answer at the maturity date itself.

Smart Capital Center builds this kind of continuous monitoring across the full debt lifecycle, and its client base gives some indication of where this matters in practice. KeyBank Real Estate Capital has used the platform to cut financial model preparation time by 40%, and institutional clients including JLL and The RMR Group work with it as well. None of this replaces the credit officer's judgment. It changes how much time that judgment has to work with before a situation hardens into something harder to fix.

The adoption paradox: why most firms have AI but few have changed their workout outcomes

A fair skeptic might point out that most CRE firms already have AI. Deloitte's 2026 Commercial Real Estate Outlook found that CRE leaders see real promise in AI but face significant implementation challenges, and that turning that promise into results requires reliable data and genuine application readiness. The gap between firms that have adopted AI and firms that have seen it change an outcome is wide. The share of executives reporting a transformative AI impact fell sharply from 2025 to 2026 because most firms pointed it at the wrong workflows. Covenant monitoring, financial spreading, and workout data extraction are where the leverage sits, and running AI on report summaries while leaving those three processes manual captures only a sliver of what is available.

Two failure modes explain most of this gap, and neither is a reason to retreat from AI monitoring, even though both have burned early adopters. The fix is not walking away from the monitoring, it is improving the inputs: layering in asset-level leasing performance, sponsor behavior history, and near real-time data feeds with validation checks built in, which cuts false positives while keeping the early-warning advantage intact.

Second, and arguably more consequential: data quality acts as a multiplier, in both directions. If firms feed clean, standardized property and financial data into an AI platform, they see dramatically more value from it than firms feeding it inconsistent, siloed records pulled from six different systems that have never been reconciled. The platform does not fix a firm's underlying data mess. It exposes it, often in front of a loan committee, which is an uncomfortable but useful place to find out.

None of this argues for black-box automation, and regulated lenders would not accept it if it did. Loan committees, audit teams, and regulators all expect a documented decision trail, so the workflows that hold up are the auditable ones: AI narrows the decision set and documents its reasoning, and a credit officer makes the final call. That is the actual difference between AI that adds risk to an institution and AI that reduces it. The other is transparent, traceable, and reviewed by a person whose name is on the decision.

The regulatory floor that spreadsheet-based monitoring can no longer meet

Spreadsheet-based covenant tracking has a second problem beyond the scaling limits described earlier: it is becoming harder to defend to an examiner. OCC Bulletin 2026-13 on Model Risk Management lays out a principles-based, risk-based framework for governing, validating, and monitoring models on an ongoing basis. The bulletin does not demand that every output be traceable to its inputs and reviewable on demand, and it explicitly excludes simple arithmetic calculations, including covenant math, from its own definition of a "model." That technicality does not let spreadsheet monitoring off the hook. It arguably makes the standard harder to meet, because there is no bright regulatory line protecting a manual process that cannot produce the systematic, documented, scalable program examiners now expect to see.

A chief credit officer walking into an OCC or FDIC examination needs to show a monitoring program that is organized, documented, and built to handle volume, not a patchwork of spreadsheets maintained by whoever happened to own that loan last quarter. Examiners increasingly look for automated testing workflows, centralized exception tracking, and board-ready reporting that shows covenant health across an entire CRE portfolio in one consolidated view. If you handle sensitive loan and property data at current volumes, SOC 2 Type II certification, AES-256 encryption, and US-based data storage are not differentiators anymore. They are baseline requirements.

What makes this section connect back to the earlier operational argument is that the exact audit trail an examiner wants to see, who was notified of a breach, when, and what action followed, is the same institutional record that makes a workout decision defensible months later in a negotiation or a courtroom. Manual monitoring fails on two fronts at once: it cannot keep pace with volume, and it cannot produce the paper trail regulators now expect. Those two failures reinforce each other. Fixing one without the other leaves an institution still exposed.

Workflow infrastructure for high-volume watchlist and workout

Putting the pieces together shows what separates institutions that move quickly from early warning to resolution from those still working case by case. It starts well before a loan shows any sign of trouble: refinance risk conversations need to begin 12 to 18 months before maturity, because borrower outreach, capital stack analysis, sponsor communication, and lender negotiation all take real time, and waiting until 90 days out strips leverage from everyone at the table, lender included. From there, continuous monitoring feeds the watchlist automatically, so loans move onto it based on live covenant signals rather than a date on a calendar, and portfolio managers can see trouble developing before any formal trigger is reached.

When a workout does start, it should draw on the same document intelligence that underwrote the loan in the first place, appraisal updates, modification terms, current financials, and PCR data available in structured form rather than requiring someone to re-extract it all from the original closing binder. Data captured at origination has to stay connected to the asset for the life of the loan, rather than getting locked away in an origination system while the asset management team works from an entirely separate set of documents years later.

Generic AI tools dropped into this infrastructure problem introduce their own risk, because they were not trained on CRE document structures, CRE-specific metrics, or the covenant frameworks that govern commercial loan agreements. A platform built specifically for this work, trained on rent rolls, T-12s, leases, and covenant schedules, and grounded in a firm's own underwriting IP and templates, produces outputs that can actually survive a loan committee meeting or an examiner's questions. Smart Capital Center was built around this exact operating model: continuous AI monitoring across the full debt lifecycle, configurable covenant thresholds per loan, automated escalation with full audit trails, and portfolio-level dashboards giving credit teams one consolidated view of the entire CRE book, developed by people who have worked through more than $500 billion in CRE transactions. With $875 billion maturing in 2026 alone and special servicers carrying some of the heaviest workout volumes in years, the institutions handling it best are not the ones with the sharpest individual judgment calls. Their infrastructure is what gives that judgment somewhere to actually work.