Evaluating an AI Underwriting Platform: Questions Private Lenders Should Ask Before a Pilot

Updated November 5, 2026: corrected for accuracy.

Before piloting an AI underwriting platform, a private lender should ask what loan and note types the tool was built for, how it reports files it cannot process, how it handles missing documents, how reviewers trace output to source documents, how escalation works, and whether its output covers what the servicer needs at boarding.

Key Takeaways

  • Ask what note structures the platform was built and tested on, and get the answer in writing.
  • A platform should say clearly when it cannot process a file. It should not hand back a score that looks normal.
  • Missing, unreadable and outdated documents should each produce a distinct, visible flag.
  • Every output should point back to the source document and page that support it.
  • Escalation to a person needs a named owner, full context and a recorded decision.
  • Compare the platform’s output with your servicer’s boarding requirements before the pilot starts.

Related Topics

What Loan and Note Types Was the Platform Built For?

Start with scope. Many underwriting tools learn from conventional mortgage files. Private lending includes structures those files rarely contain. Seller-financed notes, wrap-around mortgages, interest-reserve notes, custom balloon structures and multi-investor notes each follow their own patterns.

Ask the vendor for a written list of supported structures. Then ask for a second list of structures the tool does not support. A vague answer such as “it handles most private loans” is not enough. If you lend on fractionated loans, ask directly how the tool treats more than one investor on a single note.

Also ask whether the tool separates business-purpose files from consumer-purpose files. Your own process may treat them differently. The tool should not blur that line.

How Does the Platform Report Files It Cannot Process?

Every tool meets files it was not designed for. What matters is how it tells you. Some tools return a clear error state. Others return a risk rating anyway, even when the input fell outside their design. That second behavior is the dangerous one.

Ask the vendor these questions:

  • Does the tool return an error, a warning or a score when it meets an unsupported structure?
  • Does the error message explain the reason in plain words?
  • Can you pull a report of every file that ended in an error state?
  • Does a file in an error state stop moving forward until someone acts on it?

Ask for a live demonstration with a structure on the unsupported list. Watch what the screen shows. The red flags in AI underwriting post covers other warning signs to watch for during a demo.

What Does the Tool Do When Documents Are Missing?

A missing document is different from an unreadable one. An expired document is different again. A useful platform names each of these states separately.

Ask whether the tool fills gaps with assumptions. Some tools score a file using whatever data is present and say nothing about what is absent. That produces a result that looks complete but is not. Ask the vendor to show you exactly where the screen marks a missing item.

Next, ask how the tool records its checklist. You want a record of which documents it looked for, which it found and which it did not find. That record helps your team follow up before closing. It also helps anyone who reviews the file later.

Can a Reviewer Trace Every Output Back to a Source Document?

Traceability is the core test. If the tool states an income figure, a property value or a lien position, a reviewer should be able to open the source document and find it. The best tools link each output to the document and page it came from.

Ask these questions:

  • Does each data point show its source document and location?
  • Does the tool keep a log of what it read, when it read it and which version of the model it used?
  • Can you export that log for an investor, an auditor or a regulator?
  • Does a human override appear in the log with the reviewer’s reason?

If any of your loans are consumer-purpose, rules on credit decisions, notices and record retention can apply. Which rules apply depends on loan type, state and loan purpose. Consult qualified legal counsel about what your records must show. The guide on how to evaluate AI in underwriting covers broader review criteria.

How Does Escalation to a Person Work?

Every platform needs a clear path from the machine to a human. Ask who receives an escalated file. Ask what that person sees when the file arrives. The reviewer should get the tool’s reasoning, the source documents and the reason for escalation in one place.

Then ask about authority. Can the reviewer override the tool’s output? Does the system record the override and the reason? Can the reviewer send the file back for more documents without losing the history?

Finally, ask what triggers escalation. Some triggers come from the vendor. Others come from your own policy. You should know which is which, and you should be able to see every trigger. The post on how to choose AI for private mortgage underwriting covers how to weigh vendor control against lender control.

Will the Output Cover What Your Servicer Needs at Boarding?

Underwriting ends at closing. Servicing begins at boarding. A gap between the two shows up as delays and follow-up requests after the loan closes.

Before the pilot, ask your servicer for its boarding requirements. Then compare that list with the fields and documents the platform produces. Look for items the platform does not capture at all. Common examples include the full payment schedule on a custom-structured note, insurance details, the collateral description as it appears in the recorded security instrument and the investor split on a multi-lender note.

The list of documents a servicer collects at loan boarding is a good starting point for that comparison. If you are still choosing a servicer, the questions to ask a private mortgage servicer include how it handles files that arrive with gaps.

Frequently Asked Questions

Should a pilot include files the platform is likely to struggle with?

Yes. Clean files show what the tool does well. Unusual structures and incomplete files show how it fails. You need both views before you decide.

What is an error state?

An error state is the tool’s signal that it could not process a file. A clear error state stops the file and tells a person why. It is safer than a score built on input the tool was not designed to read.

Why does traceability matter for a private lender?

Investors, auditors and regulators can ask how you reached a decision. Traceability lets a reviewer show the source document behind each data point. Record requirements vary, so consult qualified legal counsel about your own obligations.

Who should own escalated files?

A named underwriter or reviewer with the authority to override the tool. That person should see the tool’s reasoning and the source documents together.

Should the servicer be involved before a pilot?

It helps. The servicer can share its boarding requirements so you can check whether the platform captures them. That check is easier before rollout than after.

Where can I learn about the costs of servicing?

Servicing costs depend on the portfolio. You can request an online price quote.

Sources

  • Underwriting, disclosure, and record-keeping requirements vary by loan type, state, and loan purpose (business or consumer). Consult qualified legal counsel about the rules that apply to your loans.

Next Steps

If you are weighing an AI underwriting platform and want to talk about what a servicer needs from your files at boarding, we can help. Contact Note Servicing Center to start a conversation about your servicing needs.

This content is provided for general informational purposes only and does not constitute legal, financial, or compliance advice. Always consult a qualified attorney or advisor regarding your specific situation.

Share This Story, Choose Your Platform!

Disclaimer

The information provided in this article is for general educational and informational purposes only and does not constitute legal, financial, investment, tax, or professional advice. Note Servicing Center, Inc. is a licensed loan servicer and does not provide legal counsel, investment recommendations, or financial planning services. Reading this content does not create an attorney-client, fiduciary, or advisory relationship of any kind. Nothing in this article constitutes an offer to sell, a solicitation of an offer to buy, or a recommendation regarding any security, promissory note, mortgage note, fractional interest, or other investment product. Any references to notes, yields, returns, or investment structures are illustrative and educational only. Past performance is not indicative of future results, and all investments involve risk, including the potential loss of principal. Note investing, real estate transactions, and lending activities are subject to federal, state, and local laws that vary by jurisdiction and change over time. Before making any decision based on the information in this article, you should consult with a qualified attorney, licensed financial advisor, certified public accountant, or other appropriate professional who can evaluate your specific circumstances. Some articles on this site include hypothetical stories, examples, and scenarios created to illustrate concepts and demonstrate the types of situations Note Servicing Center, Inc. handles. Any names, companies, properties, and circumstances in these examples are fictitious or have been anonymized to protect confidentiality, and any resemblance to actual persons or entities is coincidental. These examples do not describe specific clients and do not guarantee any particular outcome. Some content may be created with the assistance of generative AI tools and may contain errors or omissions. While we make reasonable efforts to ensure the accuracy of the information presented, Note Servicing Center, Inc. makes no warranties or representations regarding the completeness, accuracy, or current applicability of any content. We disclaim all liability for actions taken or not taken in reliance on this article.