60 days
of warning before a loan reaches 30+ days past due
Nightly
rescoring, so each morning starts current
Connector
added inside Claude or ChatGPT, no engineering
1,100+
lenders already service their loans in TMO
The Turing release at a glance
Three new capabilities, one source of data.
A watchlist you would build yourself, a report you would request, a question you would email someone about. Three things that now answer themselves, so your team spends its mornings acting on the answer instead of assembling it.
Portfolio Risk Predictor
A nightly score on every loan, ranked into Watch, High and Critical, with up to three drivers and a 60-day trend behind each one.
See the PredictorAI Data Explorer
Build the exact dataset you need, then ask it questions in plain English and get a verified figure back with the calculation behind it.
See the ExplorerMCP and Enhanced APIs
Add TMO as a connector inside Claude or ChatGPT and ask about loans, properties and liens. Read-only, scoped to your company, fully audited.
See MCP and APIs01 · Portfolio Risk Predictor
See the problem while it is still fixable.
A machine-learning watchlist that scores your whole book every night, ranks it by what is most likely to fall behind, and shows the reasons behind every score. Loans already 30 or more days past due stay off it, because the point is to reach you first.
- 0 to 100 likelihood, not a label.Higher means a greater chance of reaching 30+ days past due inside a 60-day window.
- Signals from the loan itself.Days since last payment, 30+ DPD events, NSF count, late fee count, accruing interest, loan age, and months to maturity.
- Rescored every night, on your book.The model trains on your own loan performance and is reviewed against your portfolio before it is switched on.
02 · AI Data Explorer
Pull the exact data, then ask it questions.
Pick your datasets and fields, filter them, and save the setup so the same pull takes one click next month. Then ask that data a question in plain English and get a verified figure back with the calculation behind it.
Pull every loan that is slipping, with the payment history that explains why.
Give an investor their position without rebuilding it by hand.
One portfolio-wide cut you can hand to the board, saved and rerun each quarter.
See what is coming due before it lands on someone’s desk as a surprise.
03 · MCP and Enhanced APIs
Ask your loan book a question.
Our MCP server connects TMO to the AI assistant your team already uses. Add TMO inside Claude, ChatGPT or your IDE. No API keys to manage.
Example questions
- How is my loan portfolio split between conventional and construction loans?
- Which of my properties are carrying the most loan-to-value risk?
- What’s going on with loan B006584?
- How many active loans are already past their maturity date?
$13.9M across 25 active loans.
Read-only, scoped to the company you signed in to, and audited.
Schedule a demo
See AI Portfolio Tools on your own loans.
Tell us a little about your lending and we’ll walk you through the Portfolio Risk Predictor, the AI Data Explorer and the MCP server on a portfolio that looks like yours.
Scored on your own data
We start with what your portfolio actually contains, because that is what the model learns from.
Connected to your AI stack
We connect the MCP server to Claude or ChatGPT live and ask your kind of questions.
A clear path to turn it on
Leave knowing what AI Portfolio Tools cost, what your instance needs and a realistic timeline.
Schedule a Demo
FAQs
What are AI Portfolio Tools?
AI Portfolio Tools, shipping together as the Turing release, are TMO's first line of AI-native capabilities for loan servicing: a predictive watchlist that flags loans heading toward delinquency, a far more granular data explorer, and an MCP server that lets your team query the portfolio from the AI assistants they already use. All of it runs on the data already in your instance.
How does the Portfolio Risk Predictor work?
A machine-learning model scores each loan from 0 to 100 on how likely it is to reach 30+ days past due within the next 60 days. It is trained on your own loan performance history, and scoring runs nightly in batch, so the dashboard shows a last-scored timestamp rather than live data.
What drives a loan's risk score?
Indicators from the loan's own behavior: days since last payment received, total and recent 30+ DPD events, NSF count, late fee count, accruing unpaid interest, loan age and months remaining to maturity. Up to three top drivers surface per loan, alongside a 60-day trend line.
Which loans get scored?
Conventional and construction loans. Loans already 30+ days past due are intentionally left off the watchlist, because the tool exists to warn you before a loan is late, not to restate what you already know.
Do we need a certain amount of data for the model to work?
Yes. Predictive scoring needs enough loan volume and payment history to train on, so TMO reviews the model against your portfolio before the module is enabled. Your account manager can confirm where your instance stands.
How is the AI Data Explorer different from a report?
You build the dataset yourself. Choose a primary dataset, join the others you need, set a date range on history data, pick your fields, rename and reorder columns, apply filters and download a CSV. No SmartView building and no API work, and sensitive identifiers are masked by default.
Can I ask questions about an export I just ran?
That is coming. Once an export is ready you will be able to open a chat beside it and ask about that data in plain English, with the conversation kept alongside the download so you can return to it later.
What is an MCP server, and is it secure?
MCP is the standard AI assistants use to connect to outside data. TMO's server is a read-only bridge to your own loan data: add the TMO connector inside Claude, ChatGPT or your IDE, sign in with your TMO credentials and pick a company. Sign-in runs through OAuth with short-lived scoped tokens, access is isolated per tenant, queries are rate limited and everything is audited. Nothing in your book can be changed through it.
How do we get AI Portfolio Tools turned on?
Talk to your account manager. The Portfolio Risk Predictor is rolling out to select clients first and depends on a model review of your portfolio, and the MCP server is an add-on for customers with an active API subscription.
