Fintech has stopped debating whether to use AI and started competing on how deeply it runs. Eight in ten financial firms now make moderate-to-large investments in AI, according to Broadridge Financial Solutions — and those that lag are not losing a feature; they are losing the business.

The economics have changed. Generative AI spending across enterprises hit $37 billion in 2025, up from $11.5 billion in 2024 — a 3.2x year-over-year jump, according to Menlo Ventures. Within financial services, Broadridge reports that 72% of firms made moderate-to-large investments in generative AI in 2025, compared with just 40% in 2024. That near-doubling in a single year signals something more durable than a technology cycle: it signals an infrastructure build-out. CFOs who still classify AI as an R&D line item are misreading where it sits in the cost structure.

The Infrastructure Argument

For most of the past decade, financial firms treated AI as an enhancement layered on top of existing systems — smarter search, better dashboards, faster reporting. That framing is now operationally obsolete. AI has migrated into the core stack: underwriting engines, fraud detection pipelines, compliance monitoring, and back-office reconciliation. The distinction matters for capital allocation. Infrastructure spending carries different justification logic, different depreciation schedules, and different competitive consequences than feature development.

Broadridge’s 2025 Digital Transformation Study, which surveyed more than 900 financial services executives globally, found that 80% of firms are making moderate-to-large AI investments overall, with asset managers planning to increase AI spending by 28% from current levels.

68% use AI in market and investment research — Broadridge Financial Solutions
Sixty-two percent use it for meeting summarization and note-taking; 58% apply it to analysis and data visualization. These are not pilot programs. They are embedded workflows.

Fraud Detection: The Clearest ROI Signal

No use case illustrates the infrastructure argument more cleanly than fraud detection. The volume and sophistication of financial fraud have outpaced every rule-based detection system. AI models that ingest behavioral data, transaction patterns, and document metadata in real time have become the only credible defense at scale.

The numbers are concrete. Allica Bank, a UK business lender, deployed an AI document-fraud tool that now identifies over £1 million per week in fraudulent loan applications, according to Marqeta’s 2025 fintech research. That is not a marginal efficiency gain — it is a risk-elimination system running continuously without human review queues.

The generative AI in fintech market — valued across fraud detection, underwriting, and personalization — is forecast to grow by $5.56 billion between 2024 and 2029 at a compound annual rate of 36.9%, according to Technavio. CFOs should treat that growth rate as a proxy for how fast the capability gap widens between firms that deploy and firms that delay.

Key Takeaway: AI in financial services has crossed from competitive advantage into competitive necessity. Firms that have not embedded AI into fraud detection, underwriting, and back-office operations by 2026 will face structurally higher loss rates, slower cycle times, and unit economics that cannot match those of AI-native competitors.

Underwriting: Speed as a Strategic Asset

Upstart Holdings (NASDAQ: UPST) offers the most detailed public data on what full AI integration looks like in underwriting. In 2024,

91% of loans fully automated with no human involvement — Upstart Holdings
and 93% of instant approvals converted to funded loans, according to Upstart’s annual results. In Q2 2025, originations surged 154% year over year to $2.8 billion, with conversion rates reaching 23.9%.

The mechanism is straightforward. Upstart’s neural network models evaluate creditworthiness using far more data inputs than traditional FICO-based underwriting, which relies on seven primary variables. The result is faster decisions, lower default rates, and an ability to approve borrowers that legacy models would reject. Traditional lenders relying on manual underwriting workflows cannot replicate that cycle time or approval accuracy without replacing the underlying system architecture.

Traditional lenders competing with AI-native firms face a clear timeline pressure: the longer they delay upgrading underwriting infrastructure, the steeper the margin compression becomes — and the more disruptive the eventual rebuild.

Back-Office Operations: Where the Savings Accumulate

The fraud and underwriting cases attract attention, but back-office automation is where AI delivers the most consistent cost reduction. Gartner’s 2025 finance AI survey found that

59% of finance functions use AI in some capacity — Gartner
with organizations further along in deployment more than twice as likely to report moderate impact and nearly three times as likely to report high impact. The returns compound with depth of integration, not breadth of experimentation.

Morgan Stanley has deployed AI-driven tools that give its financial advisors real-time access to the firm’s research library, compressing hours of document retrieval into seconds. The firm has framed this not as a productivity perk but as a client-service infrastructure requirement — a signal that AI-augmented advisor capability is now a baseline expectation in wealth management, not a differentiator.

AI-enabled fintechs attracted $7.2 billion in global investment in the first half of 2025 alone, nearly matching total AI fintech investment for all of 2024, according to Innovate Finance. Institutional investors are betting that AI-infrastructure-first firms will compound their operational advantages faster than traditional institutions can close the gap.

What Happens Next

The next competitive frontier is agentic AI — systems that do not merely assist human decisions but execute multi-step financial workflows autonomously. Broadridge’s 2026 Digital Transformation Study reports that leading firms are already moving beyond generative AI experimentation toward scaled agentic deployment. CFOs should expect budget requests tied to agentic infrastructure within 12 to 18 months.

The firms to watch are those moving AI from departmental tools into enterprise-wide operating systems, where a single model layer touches origination, servicing, compliance, and reporting simultaneously. That architecture produces network effects within the firm: each additional use case improves model accuracy across all others. Executives who approve AI investment line by line, rather than as coordinated infrastructure, will build the wrong thing at the wrong pace.

The strategic question for 2026 is not the size of the AI budget. It is whether that budget is building infrastructure or buying features.

Sources

  • Broadridge Financial Solutions, 2025 Digital Transformation & Next-Gen Technology Study
  • Broadridge / StockTitan, Gen AI Delivering Now — 2026 Study
  • Menlo Ventures, 2025 State of Generative AI in the Enterprise
  • Upstart Q4 and Full Year 2025 Results
  • Upstart AI Underwriting Edge, deBanked
  • Technavio, Generative AI in Fintech Market 2025–2029
  • Innovate Finance, Fintech Investment Landscape 2025
  • Gartner, Finance AI Adoption Survey 2025
  • Marqeta, AI in Payments and Fintech 2025
  • Banks and Bankers, AI Fintech Trends 2026