Frontier Finance: Agentic AI Redefines Banking and Insurance

Frontier finance: AI agents move from tool to teammate

AI has moved past demos and pilots. In leading banks and insurers, agents now work beside people to run processes, reduce risk and deliver more personal customer outcomes.

Microsoft leaders Dalia Ophir and Tyler Pichach describe these organisations as frontier firms: places where AI agents handle tasks while humans set direction and provide oversight. The result is faster execution, better decisions and new value at scale.

What this looks like in financial services

In banking and insurance, the shift is an operating model change. Agents and people work side by side, with AI taking on defined jobs and learning from outcomes while teams focus on exceptions, strategy and client care.

The timing makes sense. Technology has matured, regulation is catching up, and customers expect personal service. Smaller institutions can scale with agent workflows, and larger firms can grow without adding headcount. Waiting risks relevance, trust and efficiency.

From pilots to production in banking

Banks are embedding generative AI into core modernisation, not just wrappers around legacy stacks. A major focus is mainframe and application work that used to be slow and risky.

Mainframe and application transformation

Teams now use tools like GitHub Copilot and Azure OpenAI to analyse legacy code, document intent, generate tests and assist in rewriting to C# or Java. Developers report coding up to 55% faster with fewer errors and higher satisfaction.

Speed here lowers risk and improves quality. It also frees talent to work on products and customer value instead of repetitive tasks.

Beyond code: onboarding, documents and lending

Banks are applying generative AI to onboarding, document-heavy workflows and lending platforms. The aim is consistent, auditable processes with controls baked in, not bolt-on automation.

Compliance at speed

Regulatory review that once took weeks can now run in a fraction of the time. Agent suites manage updates, perform control QA and run gap analysis end to end. In one deployment, compliance work ran five times faster with 5,000 unique mappings generated and 300 gaps identified and closed.

Proof inside the enterprise

Microsoft's finance division deployed agents across quote-to-cash, tax and financial close. Results: 75% time saved on reporting and compliance, 60% faster invoice processing and 97% less effort for tax file preparation.

Across the sector, institutions are automating reconciliation and collections, simplifying document workflows and activating agents for treasury and risk. A platform approach with partners lets firms modernise without a rip-and-replace, building a bridge from legacy to future-ready.

Insurance: the value chain reworked

Insurers are integrating agents across sales and distribution, underwriting, risk, claims, engagement and product creation. The goal: reduce low-value tasks so people can focus on clients and decisions.

Generative AI can cut claims payout by 20-30% and reduce loss-adjustment expense. Employees also gain simpler access to knowledge and context, speeding up day-to-day work.

Proactive health engagement with dacadoo

Wearable data shows a drop in activity and disrupted sleep for an insured member over several weeks. dacadoo's Digital Health Engagement Platform, running on Azure with generative AI, flags risk early, nudges the member through the wellness app and recommends a virtual check-in. The insurer's risk model updates in near real time, lowering long-term costs while supporting better outcomes.

Everyday productivity with Copilot

Hiscox piloted Microsoft Copilot with 300 users, expanding to 1,000 within a year. Results: 15% gained an hour back per day, 20% saved 30 minutes and 20% reclaimed 10-15 minutes daily-largely by summarising emails, meetings and documents and reducing manual handoffs.

Claims in hours, not weeks

Agents can compress the claims cycle from weeks to hours. They speed reporting, standardise evaluation and route exceptions to specialists with full audit trails.

Agentic operations: from assist to operate

Financial firms are moving through three phases: copilots that assist employees, agents that join teams as digital colleagues and agents that run entire workflows with human oversight.

In payments, agents handle reconciliation, exception repair, dispute resolution and KYC. They interpret SWIFT messages, diagnose errors, draft responses and take corrective action through legacy systems using computer-using agents.

Data, architecture and trust

AI quality depends on data quality. Harmonise structured and unstructured data; in payments, standards like ISO 20022 are foundational. Open legacy systems with APIs and cloud-native services.

As agents take on more responsibility, policies for authentication, authorisation and auditability become critical. Microsoft's work with Mastercard, PayPal and Visa on tokens for agentic payments shows how security can keep pace with automation.

Mastercard's Agentic Payments Program

Mastercard is introducing agentic tokens and payment passkeys to improve trust, security and control. With Microsoft Azure OpenAI Service and Copilot Studio integrated into its payment solutions, the program supports conversational initiation of payments across millions of online merchants and creates new interfaces to separate trusted agents from bad actors.

Six moves for leaders

• Re-architect core platforms: break monoliths into services and make systems agent-ready.
• Build unified data foundations: create a single, governed estate that supports real-time analytics.
• Establish AI governance: adopt transparency, fairness and accountability frameworks; set culture; pilot Copilot and agent workflows with clear guardrails.
• Upskill your workforce: train teams on AI literacy, prompt engineering and agentic operations. For structured learning by role, see AI courses by job.
• Prototype and scale: start with high-impact, lower-risk use cases like fraud detection and policy intake; measure time saved and ROI; embed compliance checks early.
• Engage your ecosystem: co-innovate with partners to reduce delivery risk and speed adoption.

Partner perspectives

• FIS: an enterprise data and AI environment on Azure gives teams access to data-driven capabilities and agentic AI use cases using Azure Databricks.
• IBM: combining IBM AI with Microsoft AI to modernise core and next-gen systems; solutions include IBM watsonx Orchestrate for workflow automation and Copilot to improve productivity and decisions.
• M-Files: native storage in Microsoft 365 with Copilot integration delivers AI-driven document workflows, automated governance and secure collaboration for banks and insurers.
• Moody's: embedding AI into knowledge-heavy workflows-from credit assessment to portfolio oversight-on Microsoft's cloud and AI stack, with agentic solutions that automate multi-step tasks with transparency.

The bottom line

AI is moving from a layer on top of core systems to the operating model itself. Leaders who act now-modernising core platforms, data and governance-will define the next chapter of financial services.