Models

Propel adds Plaid cash-flow data to AI credit underwriting

Propel says a Plaid-powered AI model could expand credit access by up to 15%, but third-quarter credit costs rose as the lender prepares for a U.S. rollout.

Propel Adds Plaid Cash-Flow Data to AI Credit Underwriting
Propel Adds Plaid Cash-Flow Data to AI Credit UnderwritingPexels / pixabay
By Rebecca Stone4 min read

Why it matters

  • Propel said on November 18, 2026 that its model could expand credit access by up to 15%.
  • The cash-flow system entered operation in late Q3 2026 and uses six months of bank transactions.
  • Plaid reports connectivity to more than 12,000 financial institutions.
  • Propel and its bank partners have underwritten more than two million loans and lines of credit.
  • Third-quarter revenue reached $201.5 million, up 32%, while originations rose 27% to $260.7 million.

By Rebecca Stone

Propel Holdings said on November 18, 2026 that its enhanced AI cash-flow underwriting model could expand credit access by up to 15%.

Propel said the system targets consumers whose financial circumstances traditional credit-bureau data does not fully capture. The company said early results showed the model could increase approvals while leaving risk thresholds unchanged.

The company did not position cash-flow underwriting as a wholesale replacement for bureau scoring. It said the system supplements bureau information. The comparison matters because the company’s 15% estimate depends on adding approvals without changing those risk limits.

The announcement paired the model with a partnership with Plaid for consumer-permissioned financial data.

The in-house system entered operation in late Q3 2026. It uses bank transactions recorded over six months, produces an assessment in seconds and supplements conventional bureau information.

Propel identified the central constraint: sustaining higher approvals without weakening credit performance. Its third-quarter results show funded originations rising while the provision for loan losses and other liabilities also increased.

What does Plaid add to the model?

Under the partnership, Plaid supplies insights from consumer-permissioned financial data. The inputs cover income, cash movement and borrowers’ financial obligations.

Propel said those inputs are intended to help it and its bank partners identify creditworthy applicants whom conventional underwriting may overlook. The company is using permissioned transaction data to supplement its existing approach to assessing consumers.

Dr. Jonathan Goler, Propel’s co-founder and chief risk officer, said: “By combining open banking data, supported by Plaid, with sophisticated time series deep learning models, we can better understand a consumer’s current financial circumstances and identify patterns that are genuinely predictive of repayment.”

Plaid’s underwriting documentation reports connectivity to more than 12,000 financial institutions. It describes risk signals drawn from cash-flow patterns, categorized income and account-connection activity across that network.

Plaid makes the integration available through an API or a no-code dashboard. Those options give Propel and its bank partners different routes for adding the data to an underwriting workflow.

How large is Propel’s operating platform?\n

Propel reports that it and its bank partners have underwritten more than two million loans and lines of credit. The company says that activity has built a proprietary dataset in a credit segment that can be particularly complex to assess.

That dataset gives Propel a base of prior lending outcomes, while Plaid supplies permissioned information about current income, cash movement and obligations. Propel said its model combines those inputs with conventional bureau data.

Propel’s platform page says its AI can consider up to 5,000 data points per applicant. The same page reports these operating figures:

  • The system considers up to 5,000 data points per applicant.
  • Automatic credit adjudication takes approximately six seconds.
  • The platform processes more than 100,000 unique applications daily.

Propel says the daily application figure covers the trailing 12 months as of June 30, 2026 and excludes QuidMarket. The platform also works alongside Propel’s risk team to incorporate broader macroeconomic trends.

Those figures put the model’s claimed approval gains in the context of more than 100,000 unique applications processed daily. That is the scale on which Propel expects to carry the underwriting technology into the United States in early 2027.

What do the third-quarter numbers show?\n

In third-quarter results published on November 3, 2026, Propel reported revenue of $201.5 million, up 32% year over year. The deployment followed a period of rapid lending growth.

Originations funded by Propel and its bank partners rose 27% to $260.7 million. Revenue therefore grew faster in percentage terms than funded originations during the quarter.

Credit costs provide the counterweight to that growth. The provision for loan losses and other liabilities reached 53% of revenue, compared with 52% a year earlier. Propel attributed the increase mainly to a higher proportion of new-customer originations.

Net charge-offs remained at approximately 12% of the average combined balance of loans and advances. Propel said that rate was consistent with the prior year, even as the provision ratio increased.

These company-wide third-quarter measures do not isolate the new model’s effect. The results therefore show the financial position of Propel’s broader lending operation, not a standalone performance record for the cash-flow underwriting system.

What changes in 2027?

Propel expects to bring the underwriting technology to the United States in early 2027. The company said the capabilities could eventually support more tailored lending offers and additional financial products.

The expansion gives Propel and its bank partners a larger test of the model. The relevant measures are whether approvals remain higher and whether the additional data produces credit performance consistent with the roughly 12% net charge-off rate Propel reported for the third quarter.

Because its quarterly figures do not isolate the model, Propel’s early-2027 U.S. expansion will test whether higher approvals can scale without weakening the credit performance reflected in the company’s broader lending results.

Source: unite.ai

Share this article:

More from Rebecca Stone

Rebecca Stone

Show full bio

Correspondent covering companies, deals and strategy in artificial intelligence and machine learning at AI In Context.

802 articles

Related articles

  1. Plaid Pipes Read-Only Bank Data Into Instinct's Messaging AI Assistant
  2. Four banks make customers liable for third-party AI agent losses
  3. Uptiq launches GovernAI to govern AI agents in banking at no extra cost
  4. Australia's loan AI talks. Humans still make the decisions.
  5. Wall Street Tightens AI Lending Even With Google Guarantees

« Previous articleNext article »