Filing Watch

AI speeds up credit underwriting with data automation

By Natalie Barne
·
Share:
AI speeds up credit underwriting with data automation - ai credit underwriting
Banks’ loan records arrive in spreadsheets or PDFs, while borrowers appear under three different names across systems.

Credit insurance, trade finance, and sovereign risk underwriting have long shared a fundamental challenge: reliance on data that arrives late, in inconsistent formats, and from sources that rarely coordinate. A single bank’s loan records might be delivered as a spreadsheet, while another bank uses a PDF. The same borrower could appear under three different names across systems. Even reinsurers’ essential portfolio models, needed before committing capacity, often require a full week to assemble from disparate inputs.

Gallagher Re’s Q3 2026 Global InsurTech Report, produced with CB Insights, highlights how artificial intelligence is now closing this gap—not by eliminating human judgment, but by automating the data preparation that has historically absorbed so much time. The report examines credit risk protection, trade finance, financial protection, and hedge insurance, sectors long hindered by fragmented financial, economic, and behavioral data.

Global insurtech funding fell 44.6% quarter-over-quarter to $1.36 billion in Q3, but this decline stemmed almost entirely from fewer large-scale investments rather than a broader downturn. Deal volume remained nearly unchanged at 104 transactions, compared with 107 in Q2. AI-focused firms led funding, raising $1.13 billion across 72 deals, representing 83.2% of total insurtech capital.

The transformation is already evident in underwriting practices. Convergence, a Lloyd’s-based credit non-payment insurer established in 2023, built its platform on the premise that credit insurers collect some of the most detailed financial data available—but rarely leverage it effectively. Loan tapes arrive in inconsistent formats, borrowers are mismatched across bank systems, and critical insights often remain buried in scattered emails or spreadsheets.

Convergence’s AI tools now standardize 1,300-row loan tapes that previously required analysts two to four hours to reconcile manually. Portfolio modeling, which once took a week, now completes in 10 to 20 minutes. The platform processes over 500 credit insurance inquiries monthly. The company’s growth surged 140% in 2025, and by mid-2026, it had already surpassed its full-year 2025 volume.

Read Also: Greenlight Re insurtech segment turns profit

AI turns insurers’ data chaos into speed

“Some of the most valuable companies in the world give their product away and monetize what they learn from the people who use it,” said Stephen Pike, co-founder and chief executive of Convergence. “Insurers are trusted with data every day and then let it evaporate into inboxes, spreadsheets and PDFs.”

OAK Global, a Lloyd’s specialist reinsurer operating through Syndicates 2843 and 1440, addresses the same challenge from the reinsurance perspective. Its OCEAN analytics platform integrates bordereaux standardization, obligor matching, loss simulation, reinsurance structuring, and portfolio analytics into a unified workflow. The core issue in credit reinsurance is that apparent portfolio diversification can obscure repeated exposure to the same obligor or economic shock. Multiple cedants may cover bonds from the same company without any single participant seeing the full exposure. OAK’s system identifies these overlaps before capacity commitments are made.

Sovereign risk models predict defaults months early

The sovereign risk market faces a parallel challenge at scale. Traditional ratings and credit spreads serve as lagging indicators, and emerging market sovereigns often lack liquidity or reliable data. Tellimer, an AI-driven data and analytics firm, has developed a sovereign probability-of-default model to address this gap. The model combines logistic regression for structural solvency risk with gradient-boosted machine learning to detect idiosyncratic shocks, such as commodity price swings, data revisions, or abrupt policy changes, that structural models alone would miss.

Backtesting revealed the model correctly identified 26 of the last 28 sovereign defaults, with a 1.8% false-alarm rate, on average more than eight months before the events occurred. For countries including Sri Lanka, Argentina, Ukraine, and Lebanon, the model’s alerts preceded both bond market reactions and ratings agency changes. As of July 2026, the global average sovereign probability of default sat at its highest level in more than two years.

A notable trend across the sector is a shift in investment priorities. Reinsurance firms backed 25 technology investments in Q3, down from 51 in the same quarter of 2025. Through mid-2026, they had made only 59 insurtech investments, compared with 162 for all of 2025. Incumbent carriers, the report argues, are focusing resources on building their own AI capabilities rather than backing external insurtechs, with the broader environment described as “all hands on deck” for internal AI deployment.

Leave a Reply

Your email address will not be published. Required fields are marked *