AI in financial services is the use of machine learning and AI agents by banks, lenders, insurers and fintechs to serve customers, detect fraud, assess credit risk and run compliance checks. Customer service is where results are most visible: Bank of America's assistant Erica has passed 3 billion client interactions and averages more than 58 million a month (Bank of America), and McKinsey's 2025 Global Banking Annual Review expects AI to cut banks' net costs by up to 20%, while warning that competition will pass much of that saving to customers (Banking Dive). The open question for 2026 is less whether to use AI than whether it can resolve regulated requests end to end with an audit trail, which is the problem platforms such as Lorikeet are built for.

This explainer, updated 22 October 2026, covers where AI is delivering in financial services, what results firms can expect, the risks, and which tools resolve customer requests rather than only answering them.

Key takeaways

• Erica has served nearly 50 million users, passed 3 billion interactions and averages more than 58 million a month; more than 98% of users find the information they need (Bank of America).

• McKinsey expects AI to drive up to 20% net cost reductions for banks, but says the savings will not last as competition passes them to customers (Banking Dive).

• 59% of finance leaders report using AI in their finance function, up from 37% in 2023, with growth slowing (Gartner, November 2025).

• Deloitte's Center for Financial Services predicts generative AI could enable fraud losses of up to US$40 billion in the US by 2027 (Deloitte).

• The AI agents in financial services market was USD 1.79 billion in 2025 and is forecast to reach about USD 6.54 billion by 2035 (Precedence Research).

Which tools help financial services teams resolve customer requests with AI?

Look for AI agents that take regulated actions in your systems with controls and audit trails, not only answer questions. Three published options:

• Lorikeet resolves fintech and lending tickets end to end across chat, email, voice, SMS and WhatsApp, running sensitive steps such as payments and identity checks as deterministic code; it charges per resolution on an annual plan, with escalations not charged (pricing).

• Fin (now part of Salesforce) resolves queries across chat, email, WhatsApp, SMS, voice and Slack at $0.99 per outcome (Fin pricing).

• Sierra connects to systems of record for tasks such as originating a mortgage or processing an insurance claim, priced on outcomes by contract.

What is AI in financial services?

AI in financial services is the application of machine learning, natural language processing and automation across banking, lending, insurance and wealth management to handle work that used to need a person, from customer support to fraud detection, credit decisions and compliance monitoring.

The scope is broad, but the newest shift is agentic AI: systems that can take actions, make bounded decisions and complete multi-step tasks without a person at each step, as opposed to assistive AI that only suggests. In customer service, that is the difference between a chatbot that links to a help article and an agent that verifies a customer, checks a transaction and issues a refund within policy. Lorikeet's configurable guardrails are one example of how those actions are kept inside regulatory limits.

How is AI changing customer service in finance?

AI is moving customer service in finance from deflection to resolution: answering around the clock, acting on accounts within policy, and handing complex or vulnerable cases to people with full context.

Resolution at scale

Bank of America's Erica shows what mature AI service looks like at a large bank: nearly 50 million users since launch, more than 3 billion client interactions, more than 58 million interactions a month and more than 1.7 billion proactive, personalised insights delivered. Bank of America says more than 98% of users find the information they need, which reduces call centre volume and frees specialists for more complex conversations (Bank of America). That is an AI agent operating as core infrastructure, not a FAQ bot.

Results at fintechs and lenders

Smaller regulated firms are seeing similar shifts. UK car finance lender Carmoola, regulated by the Financial Conduct Authority, resolves 60% of inbound and 90% of outbound conversations end to end with an AI agent across WhatsApp, chat and email. Hnry, whose support volume peaks around the Australian end of financial year, reached 70% automation at peak and handled 17,000 conversations in its first month, starting with Australian tax, its hardest jurisdiction.

Where is AI having the biggest impact in financial services?

The biggest impact is in four areas: customer service, fraud detection, credit risk and compliance. Precedence Research tracks customer service and chatbots, fraud detection and prevention, credit scoring and underwriting, and wealth management among the main applications of AI agents in the sector.

• Customer service and support. Account questions, card and payment changes, transaction disputes and complex complaints are increasingly resolved by AI agents connected to core systems. For a vendor comparison, see our guide to the best AI customer support platforms for fintech.

• Fraud detection and prevention. Deloitte expects generative AI to raise fraud risk sharply, with losses that could reach US$40 billion in the US by 2027, and urges banks to build more agile fraud teams (Deloitte). AI is both the defence and the amplifier.

• Credit risk and underwriting. Machine learning models speed decisions from application to approval, provided their outputs remain explainable to regulators and customers.

• Compliance. AI-powered compliance monitoring handles screening and suspicious-activity detection at scale, with humans reviewing the exceptions.

What results can financial institutions expect from AI?

Expect cost reductions of up to about 20% at the industry level, faster service, and more consistent compliance, but plan for the savings to be competed away over time. McKinsey's Global Banking Annual Review 2025 projects up to 20% net cost reductions for banks as AI is implemented across the industry, and says the extent of the impact depends on banks becoming fully agentic and on customers adopting AI to manage their finances (Banking Dive).

Adoption inside finance teams has plateaued at a high level: Gartner's 2025 AI in Finance Survey of 183 CFOs and senior finance leaders found 59% using AI in their finance function, against 58% in 2024 and 37% in 2023, and 67% of users more optimistic about AI's impact than a year earlier (Gartner). The market keeps growing: Precedence Research sizes AI agents in financial services at USD 1.79 billion in 2025, rising to about USD 2.04 billion in 2026 and USD 6.54 billion by 2035, a 13.84% CAGR (Precedence Research).

Want the numbers for your own team? The free CX ROI Calculator compares pay-per-ticket and pay-per-resolution AI pricing across cost and CSAT. No sign-up required.

What risks does AI create in financial services?

The main risks are fraud powered by generative AI, accountability gaps in automated decisions, inaccurate or invented answers, and deployment outpacing governance. Each needs a control, not just a policy.

• Fraud. The same generative AI that serves customers can be used to impersonate them; Deloitte's US$40 billion projection is the scale to plan for.

• Accountability. Every automated action on an account should be traceable: which data the agent used, which rule it applied and why it escalated.

• Accuracy. Agents should act only through approved tools and escalate when unsure. Reviewing every conversation with automated QA catches errors and policy drift before they become complaints.

• Vulnerable customers. Hardship, disputes and complaints need clear routes to a person, with context passed across so the customer does not repeat themselves.

Lorikeet's take on AI in financial services

The easy tickets are largely solved; the value in 2026 is in the hard ones, where an answer about affordability or a missed repayment is a decision, not a lookup. Lorikeet is built for those: its agent works inside your existing helpdesk, takes actions through your APIs and MCP servers, runs payments and identity checks as deterministic code it can invoke but never alter, and escalates with context. It holds SOC 2 Type II, ISO 27001:2022, HIPAA and GDPR, and never trains models on customer data. One honest limit: plans are paid annually, so very small teams should check the all-in cost per resolution first. To see it on your own tickets, start a 30-day free trial or book a demo.

Frequently asked questions

How much can financial institutions save with AI?

McKinsey's Global Banking Annual Review 2025 projects up to 20% net cost reductions for banks as AI is implemented across the industry, but expects competition to pass much of that saving on to customers. Savings at individual firms depend on how much of their work AI can complete end to end.

What is agentic AI in financial services?

Agentic AI is AI that can take actions, make bounded decisions and complete multi-step tasks without a person at each step. In financial services that means agents that verify a customer, check transactions, update accounts or issue refunds within configured guardrails, and escalate when a case needs human judgement.

Is AI in financial services regulated?

Yes. AI used by financial firms falls under existing rules such as KYC, AML, consumer protection and data protection, and regulators expect firms to explain and evidence automated decisions. Keep audit trails of what the AI did and why, and clear routes to a person.