What is AI performing note investing analysis? AI performing note investing analysis is the use of artificial intelligence to price and diligence a seasoned, currently paying commercial mortgage that is being sold on the secondary market. It covers three jobs: converting a contract rate into a bid price at your required yield, reconciling the loan documents and payment history, and re-underwriting the collateral standing behind the payments. The buyer here is not originating a loan and not stress testing a book already owned. The buyer is deciding what to pay for somebody else's paper. For the broader context, see our complete guide to AI CRE finance and capital markets.
Key Takeaways
- A performing note is priced off yield, not balance. The gap between the contract rate and your required return sets the discount, and AI can solve for it in seconds.
- Seasoning shapes the math. A five year old loan on a 30 year amortization schedule still carries most of its original balance, so the balloon drives most of the price.
- The assignment chain, the allonge, and the servicing file are where these deals die. AI reconciles the document set faster than a paralegal and flags gaps before you bid.
- Payment history is not underwriting. Re-price the collateral at today's NOI and cap rate, because a 1.42x DSCR built on 2021 rents can be much thinner in 2026.
- Prepayment protection is part of the price. A note that can be paid off at par next quarter cannot be bought at a discount and held for yield.
What You Are Buying When You Buy Seasoned CRE Paper
You are buying a contractual cash flow stream plus a lien position, not a property. The package is the promissory note, the mortgage or deed of trust securing it, the unbroken assignment chain from the originator to the current holder, any allonge endorsing the note, the loan agreement and guaranty, and the servicing file that records what the borrower has actually done.
Supply exists because maturities keep arriving. The Mortgage Bankers Association reported in February 2026 that 17%, or $875 billion, of the $5.0 trillion in outstanding commercial mortgages is scheduled to mature during 2026. Depositories and credit companies managing CRE concentration do not always wait for maturity; they sell performing whole loans at a discount and redeploy the proceeds. That is the paper reaching the secondary market.
Three adjacent things are often confused with this, and each has a different workflow. Buying a CMBS bond is not buying a whole loan; securitized positions are monitored through servicer data in the CREFC Investor Reporting Package, which we cover in our guide to AI CMBS loan surveillance and special servicing watchlists. Committing capital to a debt fund means underwriting a manager rather than a loan, which we address in our guide to AI diligence on CRE debt funds for LPs. And stress testing loans you already hold is a portfolio exercise, covered in our guide to AI loan portfolio stress testing. Buying a single seasoned note is a pricing decision, and it starts with arithmetic.
The Price-to-Yield Math AI Should Run First
Before any document review, know the number. A note's price is the present value of its remaining scheduled payments plus its balloon, discounted at the yield you require. If the seller's ask implies a yield below your hurdle, no amount of clean documentation fixes that.
Consider a representative loan. A $2,000,000 mortgage originated in 2021 at a 6.25% fixed rate, amortizing on a 30 year schedule with a 10 year term, now five years seasoned:
- Monthly payment: approximately $12,314
- Unpaid principal balance today: approximately $1,866,800, or about 93% of the original balance, because early payments are mostly interest
- Balloon due at maturity in five years: approximately $1,684,700
- Price to an 8.0% yield: the present value of 60 remaining payments plus that balloon is approximately $1,738,000
- Implied price: about 93 cents on the dollar of UPB, a discount of roughly $129,000
A spreadsheet has always done this. What AI changes is throughput and direction: it extracts the inputs from the note and servicing statement instead of you keying them, re-solves across a tape of 40 loans in one pass, and answers the reverse question just as easily. Given the seller's asking price, what yield am I actually being offered? That last question is the one sellers prefer you not compute precisely.
One caution. A nominal yield is not a return. It assumes every payment arrives and the balloon pays in full, and it ignores servicing cost, legal cost, and illiquidity. Treat the output as a starting bid.
The Document Set AI Should Reconcile Before You Bid
Note diligence is a reconciliation problem, and reconciliation is what AI does well. Load the full file, have the model build one structured table, and instruct it to flag every place two documents disagree rather than resolving the conflict on its own.
- Promissory note and allonges: confirm the endorsement chain runs unbroken from the originator to your seller. A missing allonge is the most common defect in this asset class and can stall enforcement.
- Recorded mortgage or deed of trust and every recorded assignment: gaps in the recorded chain create title problems that surface at the worst possible moment.
- Loan agreement, guaranty, and any modification or forbearance: AI should surface modifications that reset the rate, extended maturity, or waived a covenant. These frequently do not appear in a seller's one page summary.
- Payment history and servicing comments: count the actual late payments. A loan described as performing is not the same as a loan that has never been late, and a loan with four 30 day lates in the last two years prices differently from one with none.
- Title policy, endorsements, and current tax and insurance status: escrow shortfalls and lapsed coverage transfer with the loan.
- Estoppels, SNDAs, and the collateral rent roll: these tell you whether the income supporting the payments is contractually durable.
What AI does not do is give a legal opinion. It can tell you the assignment recorded in 2023 names an entity appearing nowhere else in the file, which is exactly the flag you want; counsel confirms enforceability. Teams building this into a repeatable bid process can work with The AI Consulting Network to define the extraction schema and exception rules.
Re-Underwriting the Collateral Behind the Payment History
A note that has paid on time for five years tells you about the last five years. You are buying the next five. Re-underwrite the property as if you were making the loan today.
Continue the example above. Suppose the collateral currently produces $210,000 of NOI. Annual debt service is $147,772, which puts DSCR at 1.42x. Valued at an 8.0% cap rate, the property is worth roughly $2,625,000, so the loan's $1,866,800 UPB sits at about 71% LTV. Your $1,738,000 purchase basis sits at roughly 66% of value. That basis cushion, not the 6.25% coupon, is what protects the position if the borrower stops paying.
AI's contribution here is rebuilding the numbers rather than accepting them. Have it reconstruct NOI from the borrower's T12 operating statements instead of trusting the seller's stated figure, test the rent roll's rollover schedule against the note's remaining term, and check whether the balloon is refinanceable at today's rates. A balloon that needs a 75% LTV refinance on income that has declined since origination is a workout waiting to happen, however clean the payment record looks. The AI Consulting Network helps note buyers turn that re-underwrite into a standard pre-bid checklist.
Prepayment, Extension, and What Breaks the Stated Yield
Buying at a discount only produces the yield you modeled if the borrower cannot escape at par. Prepayment terms are therefore part of the price, not a footnote to it.
- Open prepayment: a borrower free to refinance next quarter hands your discount straight back as a windfall and leaves you with cash to redeploy. Price open paper close to par.
- Yield maintenance and defeasance: these protect your economics but complicate the borrower's exit, which can raise the chance of a maturity default you then have to work out.
- Step-down penalties: model the actual schedule by year. A penalty that steps down from 5% to 1% over five years offers almost no protection in its final year.
- Extension options: a borrower's one year extension at the original coupon lengthens your hold at a below market rate. Model the extension case, not just the contract maturity.
- Default and discounted payoff: run a scenario where the balloon does not pay and you extend or settle. If the deal only works in the contract case, the discount is too thin.
AI will not tell you whether a borrower intends to refinance. It will read every clause that changes the cash flow, list them, and force you to price each one. That is the realistic division of labor.
Frequently Asked Questions
Q: What is the difference between buying a performing note and lending directly?
A: Direct lending lets you set the terms. Buying seasoned paper means accepting terms someone else set, adjusted through a discount or premium. Price is your only real lever, which is why the price-to-yield calculation comes before diligence rather than after it.
Q: How do I decide what yield to require?
A: Start from what a comparable new loan on the same asset would price at today, then add spread for illiquidity, servicing burden, and any documentation defects. A note with a clean assignment chain and a 1.40x DSCR justifies less spread than one carrying a prior forbearance.
Q: Can ChatGPT, Claude, or Gemini price a note accurately?
A: They run present value math reliably when you supply the inputs and check the output, and they are strong at pulling those inputs out of notes and servicing statements. Verify every balance against the servicing statement, because a confidently stated wrong UPB is worse than no answer at all.
Q: Is a performing note safer than owning the property?
A: It sits ahead of equity in the capital stack, which limits downside, but it also caps your upside at the contract rate plus your discount. The real risk concentrates in the balloon and in whether you can enforce the lien efficiently if payments stop.