AI in banking: How banks can maximize the ROI of AI deployments

Just 61% of the banking industry professionals we surveyed for our 2025 Process Optimization Report say they’re realizing the ROI they expect from their AI deployments. Across respondents in all industries, the average was 73%.

Under any circumstances, failing to realize the predicted ROI from your technology investments is disappointing. But when you’re operating in a rapidly evolving industry, where it’s all too easy to lag behind the competition, it’s also a major concern.

The pace of change in the banking industry – and the pace of banking, period – is becoming lightning fast. Today’s customers expect increasingly slick, digital experiences and quick decisions. At the same time, the race is on to develop smarter operations, stay ahead of fraudsters, and successfully navigate a complex regulatory landscape, even as it’s being redrawn.

AI and generative AI should be powerful tools as banks transform to meet these imperatives, and thrive in this new world of accelerated change. So, how can they make sure every AI and gen AI implementation fully delivers?

AI in banking: The breadth of the AI opportunity

Ask someone on the street how their bank uses artificial intelligence, and their mind will likely leap to the applications that most transparently impact their own banking experience. Most probably the AI powered chatbots that help to answer their requests, 24 hours a day.
While AI is playing an increasingly central role in customer service, its applications in banks – institutions that run on a highly complex web of processes, all of which have the potential for optimization – run much deeper.

For the banks that know how to use them effectively, AI and gen AI are already having a profound impact on everything from credit risk assessment and trade document validation, to fraud prevention and regulatory compliance. (How profound? Think double-digit percentage reductions in customer wait times, payment cycle times, and SLA breaches. More on all these later...)

What AI needs to know, to deliver the greatest returns

To deliver the greatest possible returns, AI tools need to understand how your bank works. This means giving AI insight into banking processes in general, as well as into the unique processes at play in your organization. Specifically, AI needs to know:

• Which systems support key processes
• Where these processes handover to each other
• Why exceptions happen
• Your bank’s business rules
• How your bank defines its KPIs

Often, banks don’t fully understand all this themselves – especially the process piece of the puzzle. Their process data is siloed across systems and departments, and while there’s an official way any given process is meant to work, the reality is often very different.

When banks press ahead without addressing these issues, and begin harnessing AI tools to support and enhance processes they don’t fully understand, they put themselves on a fast track to lacklustre ROI.

Helping AI to understand banking

This makes Process Intelligence a foundational technology for banks seeking to capitalize on the front, middle, and back-office opportunities that artificial intelligence presents.

Process Intelligence combines powerful AI of its own with data from banks' source systems, plus standardized knowledge of how financial services processes typically run. In doing so, it allows financial institutions to create a digital twin of their unique processes, from one end of the business to the other, across customer and sales, operations, and risk and compliance.

For many banks, seeing how their processes actually work is a moment of revelation. PostFinance, one of the largest financial institutions in Switzerland, uses the Celonis Process Intelligence Platform to support account opening, credit decisions, and credit processing. “When I saw this picture, I immediately realized: between what we assume our employees are doing, and what they are actually doing, is worlds apart,” says Peter Lacher, PostFinance’s COO.

Using the Celonis platform to view its processes in detail, PostFinance saw that its account opening experience could be greatly improved. “We needed to get away from so many forms,” says Lacher. “Instead, we realized we wanted an app that allowed an account to be opened with a very short interaction. And in the end, that’s what we achieved.”

Crucially, Process Intelligence also helps banks to identify and evaluate potential AI and generative AI deployments, and prioritize those with greatest value to the business. Then, it helps AI to deliver that value, by giving it the knowledge it needs to thrive.

Let's look at just three areas in which banks are using AI and gen AI to great effect.
1. Customer onboarding: Saving time with an AI agent
First impressions matter. A simple, smooth onboarding experience helps you set the tone for a trusted, enduring relationship with your customers.
But inefficient Know-Your-Customer (KYC) often adds time and friction to onboarding journeys. Indeed, 75% of bank staff surveyed for Capgemini’s World Retail Banking Report 2025 said delays in verifying customer identity were causing issues with their card application process. AI tools are ideally positioned to lend a helping hand here. An AI agent can automate identity verification during onboarding to shorten processing times, and ultimately enhance customer engagement and experience at this crucial moment in their journey.

2. Trade finance: Automating operations with AI-driven document validation
Digital solutions are rapidly transforming the processes fundamental to international trade, which for so long, have relied on slow, error-prone, manual work. Here too, AI (including Machine Learning, Optical Character Recognition, and Natural Language Processing) can play a crucial role. For example, presented with a Bill of Lading or Letter or Credit, AI can help to extract and validate its data, shortening processing times and minimizing errors.

3. Fraud prevention: Reducing risk with a fraud investigation co-pilot

From deepfaking customers’ voices and likenesses, to asking generative AI models to write effective scam emails, fraudsters are already finding countless ways to press AI into service for their own, illegal ends.

But AI is also an increasingly invaluable ally for fraud prevention teams. For example, it’s supercharging risk management, by empowering them to analyze vast transaction data sets, identify patterns, and suggest next steps for deeper investigation or escalation.

Process improvement: Optimizing end-to-end operations with Process Intelligence

As noted above, Process Intelligence comes with powerful AI baked in. This AI supports continuous process optimization, helping banks to enhance customer service, improve operational efficiency, strengthen regulatory compliance - and generally make processes work better, in every area of their business. (As well as delivering the data that other AI applications, like the ones we’ve just explored, need to reach their full potential.)

Some of the banking industry’s biggest names are making great use of Process Intelligence to drive enviable outcomes. For example:

• A 34% reduction in customer wait times
• A 30% reduction in cross-border payment cycle times
• An 80% reduction in SLA breaches related to reporting

Credit scoring, loan approval, money laundering prevention – in truth, it’s hard to find banking operations that can’t be improved by applying Process Intelligence, and unlocking new AI applications.