Cambridge Judge 2026 AI Financial Services Report: 52% of Firms Now Running Agentic AI

By Hector Herrera | May 10, 2026 | Finance

More than half of financial services firms have moved past the pilot stage and are actively running agentic AI — AI systems that take autonomous actions on behalf of the institution — according to the 2026 Global AI in Financial Services Report from Cambridge Judge Business School's Centre for Alternative Finance (CCAF). Produced in partnership with the Bank for International Settlements (BIS), the International Monetary Fund (IMF), and the World Economic Forum (WEF), this is the most authoritative global baseline on AI adoption in finance published this year.

The Numbers

The headline figures from the Cambridge report are specific:

• 52% of financial services firms are actively deploying agentic AI — not testing it, running it
• 81% of industry respondents say agentic AI will be meaningfully embedded in their operations by 2030
• 81% overall AI adoption rate across the sector (a figure that encompasses AI tools beyond just agentic systems)

The speed of that shift is notable. Agentic AI — systems that can autonomously execute multi-step tasks like routing transactions, flagging compliance issues, or managing customer workflows without a human confirming each step — has moved from a research topic to majority-deployed technology in financial services within roughly 24 months.

What Agentic AI Is Actually Doing in Finance

The report does not just count deployments. It describes where agentic AI is operating:

• Fraud detection and AML (anti-money laundering) — autonomous flagging, escalation, and initial case documentation without human queues
• Credit underwriting — AI agents assembling applicant data, scoring risk, and generating decision rationale
• Customer service escalation — agents routing and partially resolving service requests before human intervention
• Regulatory reporting — automated compilation and formatting of required filings

These are not peripheral functions. They are core operational workflows at banks, insurers, and asset managers. The 52% deployment figure means the majority of firms in the sector have moved agentic AI into production in at least one of these areas.

The Risk the Report Flags

The most significant finding in the Cambridge report is not about opportunity. It is about risk concentration.

Cyber risk from AI-generated code volume is now categorized as a systemic concern — not just a firm-level technology risk. The argument is structural: as financial institutions rely on AI to generate code for internal systems, the sheer volume of AI-produced code outpaces the capacity for traditional manual security review. Code that has not been thoroughly reviewed moves faster into production than previously, widening the attack surface across the entire sector simultaneously.

The BIS and IMF's involvement in the report gives this finding institutional weight. These are not venture-capital-backed optimism documents. When the BIS flags AI-generated code volume as a systemic risk, that is the kind of signal that shows up in central bank supervision frameworks within a few regulatory cycles.

Vendor concentration is the second systemic risk the report identifies. A small number of AI providers — primarily foundation model companies — now underpin a large fraction of the sector's AI deployments. A failure, outage, or policy change at a single provider can cascade across regulated institutions simultaneously. This is similar to the third-party concentration risk that regulators have flagged around cloud providers, but compressing faster.

What Firms Are Actually Prioritizing

The report segments what financial firms are using AI for versus what they want AI to do. The current top applications are operational: fraud detection, compliance monitoring, and customer service automation. The aspiration is toward more autonomous decision-making in credit, portfolio management, and risk.

The gap between aspiration and deployment remains largest in regulated, high-stakes decisions. Credit decisions that involve consumer lending, for example, are still predominantly human-confirmed even where AI is deeply involved in the preparation. Regulators in most jurisdictions have not yet issued clear guidance on what autonomous AI decision authority looks like in consumer finance — and firms are not moving ahead of that guidance.

What to Watch

The 2030 figure — 81% of firms expecting meaningful agentic AI embedding — is a self-reported forecast, not a structural guarantee. Watch for regulatory responses to the vendor concentration and AI-generated code risks flagged in this report; if central banks begin requiring firms to document AI vendor dependencies in their recovery and resolution plans, that will materially change how financial institutions approach model risk governance in the next 12 months.

Sources: Cambridge Judge Business School / CCAF