Pavan Agarwal Is Testing How Much of a Mortgage Decision AI Can Make
Sept. 11 2026, Updated 12:29 p.m. ET
A mortgage approval can now be generated by artificial intelligence in a matter of minutes, according to Pavan Agarwal, who says AngelAi® can make the lending decision while a licensed financial institution remains responsible for originating the loan.
In a recent discussion, Agarwal described a system in which the AI can approve a consumer in a few minutes to a few hours, depending on the situation. He said the decision is final within the system, while the actual lending must still be performed by a licensed entity.
AngelAi therefore goes beyond software that helps a loan officer find information or prepare a recommendation. The system is being developed to make decisions within a regulated lending process.
The idea has been part of Agarwal's work for several years. During a 2023 HousingWire interview, he said AngelAi was being used to calculate income, review documents, and work through large volumes of federal and agency regulations. He also described a process in which the system can involve a human when its confidence level is not high enough to issue a final answer.
The human review mechanism is important because mortgage decisions can carry financial and regulatory consequences. Agarwal explained that AngelAi makes a series of decisions while working toward an answer, with a low confidence score at any stage triggering human involvement. He said the amount of human review had been reduced as the system accumulated more answers and improved its confidence.
AngelAi's current materials describe a similar architecture. The company says its systems are designed around deterministic execution, meaning defined inputs and rules are intended to produce consistent outputs. Its Zero UI materials also describe confidence-based routing for cases that require human review.
As financial institutions experiment with increasingly capable AI systems, the question of what those systems are actually permitted to do becomes more significant.
A general-purpose chatbot can explain a mortgage term or summarize a document. A decision system has to determine what action is permitted, apply the relevant lending rules, and produce an outcome that can be reviewed later.
Agarwal has argued that the mortgage industry has spent too much time treating AI as a productivity assistant. He has said the higher-value application is using AI to perform work such as income calculation, document review, and analysis of extensive lending regulations.
AngelAi's development has taken place inside Sun West Mortgage Company, Inc. (NMLS 3277), where Agarwal serves as CEO. He has previously described Sun West as the testing and research environment for the technology, giving AngelAi access to an operating mortgage business while the system was being developed.
Human involvement also has a defined place in the workflow.
Agarwal told HousingWire that Sun West's AI workflow had humans involved in document verification and in cases where the system's confidence was insufficient for a final decision. He described the goal as reducing human touchpoints as the system becomes more capable, rather than removing the possibility of human involvement altogether.
AngelAi's newer products extend the same principle into individual parts of mortgage production. Angel Income, for example, is designed to calculate qualifying income from documents and return a warranted result. The company's current examples include self-employed borrowers and Schedule C income, with one published scenario showing document reading, analysis, and a warranty step completed in less than 10 minutes.
A June 2026 episode of The Pavan Perspective also focused specifically on Angel Income and its use in automating income calculations, reducing deal fallout, and handling borrowers with self-employment and gig economy income.
AI research in financial services has also moved toward questions of reliability, interpretability, compliance, and the risks created when automated systems operate continuously at scale. Researchers have described financial AI as increasingly capable of functioning as a decision engine rather than merely an analytical aid.
Mortgage lending gives that development a particularly demanding test. A system that can answer questions is useful. A system that can make a regulated financial decision has to meet a much higher standard.
AngelAi is being developed around that second use case.
The August release of Sena also shows AngelAi expanding beyond individual lending decisions. Celligence LLC introduced Angel Agents that can carry out delegated tasks across functions such as research, email, scheduling, and document work, broadening the platform's role from financial decisioning toward wider workflow execution.
The significance of the experiment will ultimately be measured inside the lending process itself, where an automated decision has a direct effect on whether a borrower receives financing and the terms attached to it. As more financial institutions explore similar systems, AngelAi offers a live example of what happens when AI moves from supporting a mortgage decision to producing one.
Never Miss an
Daily updates from the heart of Hollywood, right to your inbox
By entering your email and clicking Sign Up, you’re agreeing to let us send you customized marketing messages about us and our advertising partners. You are also agreeing to our Terms of Service and Privacy Policy.
