Better.com has launched a conversational credit decision engine that lets mortgage and home equity lenders run underwriting through ChatGPT using the company’s Tinman AI platform, the company announced Thursday.
The integration, built with OpenAI using a Model Context Protocol (MCP) connector, allows loan officers and bank teams to access Better’s Tinman engine directly in ChatGPT Enterprise. Users can connect their own guidelines, pricing and CRM data and receive decision-ready underwriting results for mortgage and home equity loans inside the ChatGPT interface.
Better said loan officers using the Tinman AI app in ChatGPT can take consumer application data and documents and fully underwrite loans to the specific guidelines of more than 45 institutional investors. These include Fannie Mae, Freddie Mac and the Federal Housing Administration, as well as major banks such as JPMorgan Chase, Truist, Citizens Bank, U.S. Bank, Huntington Bank and Fifth Third.
The process can be completed in as little as 47 seconds, with a median time of 2 minutes and 24 seconds. That compares with an industry average of about 21 days for a traditional mortgage underwriting process, according to the company.
“Loan officer teams and banks can simply log into their ChatGPT Enterprise account, download the Tinman AI credit decision engine app, connect their guidelines, pricing, and CRM to process, underwrite, and fulfill loans nearly instantly; passing thousands of dollars in savings to consumers,” Leah Price, general manager of Tinman AI Platform, said in a statement.
Through the MCP connector, ChatGPT acts as the front end while Tinman serves as the decisioning layer. Better said its platform maintains a real-time snapshot of each loan file — including facts, documents, actions and outstanding conditions — and uses an “underwriting orchestrator” to route work to specialized agents that read documents, apply guidelines and return a decision-ready status with a granular decision tree.
The company positioned this as an alternative to legacy loan origination system (LOS) stacks.
Tinman is trained on more than a decade of Better’s internal mortgage data, according to a company press release. Better said the platform has mapped roles, tasks, rules and decisions across over $110 billion in funded loans, more than 12 million recorded customer calls and more than 5 billion pages of credit, income, collateral and asset documentation.
It also includes the criteria and pricing parameters of investors representing more than 80% of the U.S. mortgage market.
“With OpenAI, Better is not only advancing mortgage intelligence for the industry but also demonstrating how AI can transform how financial institutions operate from the inside out,” said Giancarlo Lionetti, chief commercial officer at OpenAI.
The launch extends Better’s existing use of OpenAI tools. The company has deployed ChatGPT across its roughly 1,400 employees and uses OpenAI’s multimodal models to power Betsy, its AI loan agent that supports both consumer and enterprise customers.
For originators, the key change is where and how underwriting decisions get made. Instead of working inside a traditional LOS or proprietary portal, loan officers with ChatGPT Enterprise access can query Tinman from a conversational interface — asking if a borrower qualifies, what conditions remain or how a file would price across multiple investors — and receive answers grounded in their organization’s own overlays, product matrices and pricing models.
“The most important part of my day is delivering mortgage purchase commitment letters to homebuyers that need a mortgage to buy a home,” Better loan officer Tony Song said. “With this launch, I’m already benefitting from the efficiencies, and I will be able to serve 10x more customers daily compared to what was possible with the traditional mortgage underwriting process.”
Why this matters for housing professionals
The move underscores how quickly generative AI is moving from front-end lead capture to core credit decisioning in mortgage. By embedding investor guidelines and pricing logic into a conversational interface, lenders and brokers could shorten cycle times, reduce manual touches and potentially compress the cost to originate — a key focus in a volume- and margin-constrained market.
At the same time, the strategy highlights a broader shift: Instead of building standalone lender tools, some vendors are meeting loan officers inside horizontal AI platforms such as ChatGPT Enterprise. For lending executives, that raises new questions around data governance, model explainability, investor acceptance and regulatory expectations as AI-driven decision support moves closer to underwriting.
Better said it plans to continue building additional tools on OpenAI’s stack aimed at reducing both the time and cost to originate loans, with a stated goal of making mortgages more affordable for consumers.
This article was produced with the assistance of HW Automation.