Carrington Labs Unveils AI, Analytics-Driven Cash Flow Underwriting Capabilities

By Roy Urrico

Earlier this month, Carrington Labs, a provider of credit risk analytics and cash flow underwriting models, launched Cashflow Score to meet the growing demand for complementary underwriting capabilities. Built on open banking transaction data and delivered in real-time, the score helps lenders approve more borrowers without increasing risk.

Jamie Twiss, CEO of the Sydney, Australia-based Carrington Labs, sat down with Finopotamus to discuss how credit unions and banks can leverage supplemental data to reshape financial services.

Jamie Twiss, CEO of Carrington Labs.

“I am a data scientist by training and that is why I got into this part of the industry. I just love the application of contemporary data and analytics techniques to the challenges of lending overall and credit risk management in particular,” said Twiss.

Helping Credit Unions and Banks with Lending Decisions

Commercially live since March 2024, Carrington Labs is accelerating its North American expansion of its artificial intelligence (AI)-powered software as a service (SaaS) platform. Their solutions provide insights (credit risk analytics and cash flow underwriting) to help lenders including credit unions, banks and fintechs make more accurate lending decisions.

Carrington Labs delivers a suite of credit risk solutions that span the entire loan lifecycle — from cash flow underwriting and product-specific credit risk models, to post-origination financial-health and early-warning monitoring. It is compliant with U.S. regulations, offering a scalable solution for institutions seeking to upgrade their risk models on a standalone basis or alongside a traditional credit score.

“Carrington Labs is an analytics company that builds tailored cashflow risk models for lenders…banks, credit unions and non-bank lenders of various kinds. One of the core abilities is we have a heavily automated model creation process,” remarked Twiss. “We can come to a smaller or a mid-sized lender and take their data and use that to build a machine learning based model. We can put that together quite quickly in an economic setup that makes sense for smaller lenders that would not have that capability themselves.”

What Does Cashflow Score Do?

Financial institutions needed a cost-effective entree into this new world of cash flow underwriting, explained Twiss. So, Carrington Labs introduced its off-the-shelf solution, Cash Flow Score, which leverages detailed transaction data analysis to provide a more comprehensive view of an applicant’s creditworthiness and their ability to repay. This enables credit unions, banks and non-bank lenders to more quickly and accurately assess those underserved by traditional scoring methods, including thin-file and no-file applicants, Twiss further explained. Using a score based on cash flow patterns and financial behaviors, lenders can expand credit access with confidence while maintaining risk discipline.

Twiss told Finopotamus: “You obviously would still pull their credit score, the credit file. What we can do is by taking transaction level data for an individual customer (consumer or small business), we can put together a different point of view on that customer, looking at everything from their income and expenses, but also a lot of behavioral characteristics around how they relate to money.”

Carrington Labs’ platform also analyzes transaction data across a wide range of behavioral indicators, including gambling outflows, cash-withdrawal patterns, wage advance usage, non-sufficient funds, buy-now-pay-later ratios, and account balance volatility, among many others, said Twiss. These systematically categorized inputs generate detailed financial health metrics, which power the final Cashflow Score.

Cashflow Score also enables “second chance” lending scenarios for lenders, where they can view customers who do not meet standard criteria through a modern credit risk approach, enabling more inclusive lending decisions. Beyond origination, lenders can use the score to monitor customers' general creditworthiness and financial health over time.

How Does AI Help to Achieve Carrington Labs’ Goals

Twiss, a data scientist, provided his input on how AI is used today: “The definition of artificial intelligence has changed quite a bit over time. Until the release of ChatGPT in November of 2022, when people talked about artificial intelligence, they usually meant a range of different techniques, which included machine learning, things based on what are called neural networks, the technology underlying generative AI and large language models and often some other techniques as well.”

Twiss continued, “More recently, a lot of people use AI only to mean generative artificial intelligence (that large language model kind of ChatGPT type of capability). My personal view, I do not really mind how people want to define it. We use a number of different techniques under that umbrella.”

Twiss conceded that “generative AI, large language models, ChatGPT stuff, is an incredibly powerful tool, but it is not always reliable. You can ask the same question 10 times and get 10 slightly different or very different answers. It will make stuff up. It has its place in lending. We do not use it for that actual decision-making point,” he continued. “What we do is we use a whole bunch of different techniques, including quite a bit of AI to set up the model itself. But the model itself is a much more just like statistical, predictable deterministic model. When we take Mrs. Smith's data and run it through, we expect to get essentially the same answer every single time.”

Integrating Cashflow Score

Carrington Labs, Twiss noted, can integrate with third party decisioning, or loan origination and loan management systems. “It is very easy to turn us on because you are already wired into them. It is not quite flicking a switch, but it is not far off it. If we do not have any systems in common, then we do use an API (application programming interface).”

Lenders who integrate Carrington Labs’ Cashflow Score into their existing lending workflows can anticipate a significant uplift in approvals, according to Twiss. Early applications of the Cashflow Score have shown up to 30% more accuracy in scoring high-risk customers and 2.5 times more accuracy in scoring low-risk, high-value customers. Lenders can choose to use Cashflow Score as a standalone signal or layer it with existing rules and models. The deterministic scoring system ensures consistent results, with the same inputs always returning the same outcome.

“Traditional credit scores rely heavily on static and often outdated data, missing the nuance in how people actually manage their finances,” said Twiss. “Our Cashflow Score is built differently. It draws directly from bank transaction data, regularly refreshed to capture the full context of income, spending, and saving behaviors over time. This dynamic view of financial health allows lenders to make more accurate and fair decisions, especially for borrowers who may have limited credit history but demonstrate strong financial discipline.”

Key features of the Cashflow Score include:

• Real-time results: Delivers fast, bureau-style scores ranging from 1 to 100 for immediate decision-making.

• Clear risk signals: Indicates top financial behaviors that are key drivers of an applicant’s score.

• Compliance-ready: Equal Credit Opportunity Act (ECOA)-compliant approach that supports adverse action reporting requirements.

• Easy implementation: Can be deployed quickly with minimal effort via API or batch files, and is also available through partner decisioning and lending platforms.

While Carrington Lab’s product and analytics teams are based in Australia, the sales and service team is based in Columbus, Ohio.

Carrington Labs works with a wide range of lenders, from financial institutions to fintechs. It has not publicly announced a credit-union yet, but recently confirmed an engagement with CCBank (Provo, UT; approximately $871 million assets). Carrington Labs also announced another U.S. customer, Doc2Doc Lending, and recent fintech partnerships with LendAPI, Taktile, and Oscilar.