The rise of agentic AI in financial services

Agentic AI is rapidly mastering the drudgery of audits, onboarding, compliance and more

Processing mortgage applications is one of the most manually intensive workflows in financial services. A homeowner may meet a single loan officer, but behind the scenes much larger teams grind through compliance checks, document reviews and underwriting—45 distinct steps in some cases. Such processes are ripe for reinvention by agentic AI.

Mortgage approvals are only one illustration of a broader pattern. The combination of regulatory pressure, vast quantities of data and operational sprawl—characterised by repetitive tasks handled by numerous stakeholders—has made financial services one of the fastest adopters of enterprise AI.

AI agent deployment is happening now

Large organisations report AI agent deployment more than doubled from Q1 to Q4 of 2025.

AI agents are software applications that leverage generative AI and machine learning to autonomously navigate tasks that are complex, multi-step, or involve hand-offs. They have awareness of business contexts and can act proactively or reactively. They are also capable of working in concert to handle complex tasks, a process known as “orchestration”. 

Banks, insurers and other financial institutions are now deploying agentic AI to streamline repetitive, data-intensive workflows that span multiple stages, from audits and fraud detection to Know Your Customer (KYC) evaluations and credit assessments.

I think it is going to create opportunities for much better margins, and opportunities for more investment in the future, be it new lines of business, new products or new geographies.

Chris Harrison, industry executive director, financial services at Oracle

According to Chris Harrison, industry executive director, financial services at Oracle, agentic AI offers financial institutions the chance to rethink how they operate and grow. “I think it is going to create opportunities for much better margins, and opportunities for more investment in the future, be it new lines of business, new products or new geographies,” he says.

For firms struggling with manual reviews and siloed data, assigning AI agents to handle many routine steps while reserving humans for critical decisions and exceptional cases is increasingly compelling. Beyond speeding up processes, it frees teams to focus on higher-value work, from designing more personalised products to exploring new customer segments and lines of business.

Mr Harrison, a former banker, points to lenders’ mortgage processing timelines. To illustrate how a lender might benefit, Oracle modelled an agentic version of the workflow. Dedicated agents were applied to segments of the process, reducing turnaround times from 48 days to 38, with lower projected costs and reduced customer attrition.

Measuring the potential agentic impact

In a simulated test run by Oracle in October 2025, AI agents were applied to many of the manual steps involved in a typical mortgage approval process. The model suggested a potential 21% reduction in approval times and a 13% increase in successfully closed applications.

Simulated test data

Source: Oracle.

What’s driving agentic AI?

Financial services companies are discovering that many of the building blocks for agentic AI are already in place. The massive data assets that underpin their current operations can also power the necessary models and algorithms, while much of their existing database and cloud infrastructure supports rapid deployment.

“Greater adoption is coming from two sources,” explains Karim Haji, global head of financial services, KPMG International and partner, KPMG UK. “One is the platforms that people are using, like Oracle’s. It just becomes easier to use the agents to automate tasks that otherwise you'd have to do manually. Also, what you're seeing is that people are building their own agents.” The Oracle applications suite Mr Haji refers to includes pre-built AI agents with tools that allow teams across the business to build, manage and deploy them on their own, making it easier to deploy agentic capabilities within large, complex organisations. Agentic AI still requires investment in time, people and capital. Early experiments can often draw on existing resources, but scaling models, developing custom features and integrating them across the enterprise quickly adds to the cost. Embedding intelligence directly into core enterprise applications and data can help to generate efficiencies that offset those investments.

Broker analyst investor using pc computer

Getting started

Financial services companies have often been at the forefront of technological innovation, albeit within the boundaries dictated by risk management and regulatory requirements. Banks, insurers, asset managers and others generally start small.  

“They need guardrails in place to make it safe, and then, once they're comfortable with that, they tend to adopt quite quickly,” Mr Haji says. “Particularly with AI, they're very much in the sweet spot because they are essentially data companies.”

Particularly with AI, they're very much in the sweet spot because they are essentially data companies.

Karim Haji, global head of financial services, KPMG International and partner, KPMG UK

Being a data company, however, carries responsibilities and risks. Institutions must properly secure and manage that data. Executives, staff and partners will scrutinise anything with regulatory or customer-facing implications. This is why early agentic AI projects typically limit and tightly control access to customer data. One way to address this is to embed AI directly within the enterprise environment, so sensitive data remains protected and never leaves the business.

Eventually, the focus will shift to strategic needs and associated KPIs. Backtesting helps to confirm whether agentic AI is suitable for deployment across the enterprise, Mr Harrison adds.

Enthusiastic Young Adults Analyzing New Data

A question of culture

Agentic AI is not just a new approach that makes pain points go away. Agents can be transformative, allowing staff to reimagine entire processes and priorities, and to seize opportunities that did not previously exist. This requires institutions to rethink traditional hierarchies and workflows. Leaders need to manage not only human teams, but also the AI agent ecosystems that work alongside them.

“You want to drive behaviours that make processes better, quicker, faster and potentially cheaper,” Mr Haji says. “My personal view is there isn’t enough focus on that aspect: the cultural change, education, and helping people through that journey.”

Mr Harrison points out that the changes can be profound. Routine tasks can be handled more quickly, and complex tasks such as fraud detection can be handled with fewer false positives.

Consider the office-park call centre. Tricky issues escalated to human representatives are often highly specific or involve obscure policy questions. In such cases, says Mr Harrison, instead of putting callers on hold or escalating the issue, AI agents can act as ultra-capable aides that can fetch information and enable faster resolution.

Agentic AI in 2030

80%

automation of today’s tasks

70-90%

reduction in decision latency

60-80%

compression in cycle times

40-60%

reduction in finance costs

60-80%

uplift in forecast accuracy

Commercial models around agentic AI are still being developed, but the opportunity is clear. Organisations that apply creative thinking can capture not only significant efficiency gains, but also far broader transformational benefits by rethinking processes and, in some cases, redesigning them entirely.

”Can you offer much better products and services to your clients because of this technology?” Mr Haji asks. In many cases, the answer is yes. “I think that it is quite an exciting space to be in,” he adds.

Finding the right enterprise AI technology and services providers

Financial services firms that embrace agentic AI are streamlining operations, reducing risk and accelerating growth across every area—from audits and fraud detection to new product launches and the financial close. Oracle and KPMG firms combine deep industry experience with AI-powered Oracle Fusion Cloud Applications, Oracle AI Database, and Oracle Cloud Infrastructure, helping financial institutions:

→ Boost operational efficiency by using pre-built AI agents or developing custom agents

→ Deliver seamless, personalised digital experiences

→ Uncover AI-powered innovation to maximise value from technology investments

Learn more

Thrive in a data and AI-driven future

KPMG firms and Oracle can help you leverage the power of data and AI to unlock value, accelerate innovation and drive sustainable transformation across your organization.

Accelerate your AI adoption

Oracle AI Experience offers a chance to connect with AI experts and get hands-on with groundbreaking GenAI and agentic AI solutions.