Compliance Brief
TL;DR
You can legally call an applicant's number with an AI voice agent to encourage them to finish their application, provided two things are true. First, the consumer affirmatively opted in to AI voice contact on the application itself (a checkbox they actively select, not a pre-check). Second, the operational stack supports real-time revocation, ongoing reassigned-number checks, and clean recordkeeping. The 2008 FCC ruling alone is no longer a safe harbor post-Chevron. Independent contractual consent at the application stage is the strongest protection.
Scope
Lenders using AI voice agents to follow up with consumers who started a credit application but didn't finish it, with the goal of getting them back into the funnel.
Section 01
When a consumer starts a credit application (giving their name, phone number, and other basic information) and then drops off, can a lender use an AI voice agent to call that number and encourage them to complete the application?
Yes, with the right consent architecture in place. The framing of the call matters, though, because it determines which consent standard applies.
Section 02
TCPA distinguishes between two types of outreach, each carrying a different consent burden.
01 · Lower bar
Calls directly related to an existing transaction or relationship. For example, a call informing an applicant that their application is incomplete and providing guidance on how to complete it. These require only prior express consent, which, under the FCC's 2008 Declaratory Ruling, is established simply by the consumer providing their phone number in the application.
02 · Higher bar
Calls whose primary purpose is to encourage purchase or use of a product or service. These require prior express written consent, a higher standard that must be clearly documented (typically through an affirmative opt-in checkbox or e-signature at the point of collection).
App-funnel follow-up calls often sit at the boundary. A call saying âyou started an application, here's how to finish itâ leans informational. A call saying âyou should really open this account, here's why it's great for youâ is telemarketing. Most real-world calls blend both, so the safest approach is to treat them as telemarketing and secure written consent upfront.
Section 03
Under the FCC's 2008 Declaratory Ruling, a consumer who voluntarily provides their phone number on a credit application has given prior express consent to the lender contacting that number using an autodialer or an AI voice agent for calls related to that application.
This is sufficient for informational follow-up. It is not sufficient for telemarketing under current best practice, or the heightened scrutiny courts are now applying.
Section 04
The application should include explicit consent language when the consumer provides their phone number. Strong language covers:
The phrase ânot a condition of obtaining creditâ matters. Tying consent to credit eligibility creates separate regulatory exposure.
Section 05
Even with valid consent in place, the following controls apply.
Consent is tied to the specific number the consumer provided on the application. It does not extend to skip-traced or third-party-sourced numbers. If a number was not provided directly by the applicant, do not call it with an AI voice agent.
If a consumer says âstop calling meâ at any point (during the AI call, through a text reply, or by any other channel), that withdrawal must be logged immediately and honored across all systems. Batch processing creates gaps, and gaps create liability. Revocation must propagate in real time.
For every AI voice campaign that touches app-funnel leads, maintain the application or consent record, the source of the phone number, the consent language in effect at the time, and the dialing logs. The ability to produce this quickly is the difference between a nuisance dispute and a six-figure settlement.
Phone numbers get reassigned. Even a number provided in good faith by a genuine applicant may now belong to someone else. If you are calling past applicants, you should check against the FCC's RND every 59 days if no right-party contact has been confirmed. Numbers that cannot be verified as still belonging to the applicant should not be called.
Section 06
Courts are no longer required to defer to the FCC's 2008 consent ruling. This doesn't invalidate the ruling, but it does mean lenders can no longer treat it as a safe harbor. The best protection in this environment is consent language in the application itself: an independent contractual basis that doesn't depend on regulatory interpretation.
Lenders who explicitly capture AI voice consent at the application stage, using clear opt-in language with proper assignment provisions, are well-positioned even if courts diverge on FCC deference.
Section 07 · The bottom line
AI voice and text agents are a powerful way of engaging with 100% of applicants in the funnel to help drive conversion and increase revenue in a scalable and cost-effective way.
However, it is essential to do this in a TCPA compliant way. The framework exists today: have a clear contractual basis for consent, and make sure you have strong recordkeeping for revocation.
Veritus is an AI communications platform for financial services, with AI voice and text agents that are designed for compliant conversations across origination, servicing, and collections.
This memo is for informational purposes and does not constitute legal advice. Consult qualified TCPA counsel before implementing AI voice programs.