Your deal files and the rules your team teaches the analyst stay on your account and are used for your firm's reviews only. They are not pooled with another client, not put into anyone else's work, and not used to train anything of ours. Below is what happens to each thing you send, and the qualifications we state rather than leave for your counsel to find.
What you send, and where it goes
Starting does not take a security review of new software, because there is no software to install. You forward a deal by email, as it reached you: the offering memo, the rent roll, the T-12, the broker's email. The review comes back in the same thread. Nobody on your team logs in anywhere.
What you forward is stored on your account. A room, its runs and its files belong to that account, and nothing is pooled with another client. The analyst reads your files to answer the deal you sent, and the reviews after it use what it learned from your corrections, on your account only.
Your corrections are your firm's rules
The part of an AI analyst that is most yours is not the files. It is what your team teaches it: "our debt yield floor is 9%", "we never underwrite retail with a single anchor", "use the broker's rent roll, not the OM summary". Each correction is written down as a rule for your firm.
Those rules are stored under your account, applied to your reviews, and read by nobody else's. A rule your acquisitions team gave the analyst does not show up in another firm's review, and it is not put into shared models without your consent. If you leave, your rules go with the rest of your data: returned or deleted. How your analyst learns shows the loop that turns a correction into a rule.
We do not train on your files
There is no training pipeline in our own code. Your documents are not used to train anything of ours, and section 12.8 of our customer agreement says so: we do not train AI models on your content or customer data, and our subprocessors do not either and do not retain or log it for human review.
The honest qualification: model providers
Runs execute through outside model providers, and six hold credentials in production. Routing decides which one handles a given step, so we will not answer a residency or sub-processor question by naming one of them; that would be convenient and false. Under our agreement they do not train on your data and do not retain or log it for human review. What each does under its own API terms is a question about that provider, and we answer it in writing, provider by provider, rather than with a blanket guarantee covering companies we do not control. Data and security lists them.
What we do not offer yet
Accounts are individual. There is no shared firm workspace, no seat management and no administrator view; work is shared by sending it. We do not offer a client-facing per-source audit export today. Runs, their messages and their files are recorded per account, so if you need to know who ran what and when, we can tell you, but we will not demonstrate an audit trail that only works on some products.
When you leave
On your request, or within 30 days after the service ends, your data is returned or deleted, and every sub-processor is directed to do the same. The Data Processing Addendum has the terms. If your firm needs a signed data processing agreement, a named sub-processor list or a residency commitment, ask: a person here decides, and you get either the document or a straight answer that it does not exist yet.
During probation
The same rules hold during a 7-day probation on your live deals. The analyst reads what you forward and reviews it, and that is all it does: it signs nothing, writes to no one outside your firm without your ok, and spends nothing. If you decide on day 7 not to continue, your files and rules are deleted on request.
If you would rather see the analyst on a deal of yours first, book a demo.
Updated 22 October 2026. Written by the Titleman team with AI assistance.
FAQ
Is my deal data safe with an AI underwriting tool, and does the vendor train on it?
With Titleman, your files are kept on your account. There is no training pipeline in our code, and our agreement says we do not train AI models on your files. Runs go through outside model providers; they do not train on your data or retain it for human review under our agreement, and we will give you the specifics of each in writing.
What does it take to start with an AI underwriting service, and what happens to my files?
With Titleman, you forward a deal by email, as it reached you, and the review comes back in the same thread. Nobody installs anything or logs in. The files you forward are stored on your account, read for that review and the reviews after it, and returned or deleted when you ask or within 30 days after the service ends.
Do the rules our team teaches the analyst go to other clients?
No. A correction such as "our debt yield floor is 9%" is written down as a rule for your firm only. It is applied to your reviews and is not put into anyone else's work or into shared models without your consent.
Can we find out who ran what on our files?
Yes. Runs, their messages, their states and their files are recorded per account and are retrievable, so if you need to know who ran what and when, we can tell you. A client-facing per-source audit export is not offered today.
Do we get a firm workspace with admin controls?
Not today. Accounts are individual: there is no shared firm workspace, no seat management and no administrator view. Work is shared by sending it. We would rather say so than describe something that does not exist.