3 ways AI is overhauling financial services in 2025
By Slavena Hristova (pictured), Director of Product Marketing at ABBYY
Advancements in AI-driven automation will have profound influence on financial institutions, their services, and customers’ expectations in 2025.
In fact, banking leaders need AI as much as consumers expect it. A recent survey revealed that 69% of IT decision-makers in the financial services sector fear that foregoing AI adoption will cause them to fall behind their competitors, while 64% reported that customers already expect AI integration in their services. This shift from uncertain investments in digital transformation to near-status quo across industries has left business leaders little choice when it comes to AI: learn how it can responsibly and sustainably deliver value to your organization and its customers, or your competitors will.
For the financial services sector, AI systems’ ongoing advancement most notably expands possibilities for custom-tailored service and risk reduction. Organizations in this space should have a plan to meet customers’ AI-driven expectations for personalized experiences, countermeasures against fraud, and transparency into ethical guardrails.
Hyper-personalizing services
As of June 2024, two-thirds of IT leaders in financial services were already reporting use of generative AI tools such as chatbots and digital assistants. These tools enable quicker, smoother communications with customers that can help meet the rising expectation for service at the speed of “now.”
More primitive chatbots were often limited in their functionality, only able to communicate with pre-determined prompts and keywords that would overlook complex or niche queries. Advancements in language models have enabled conversational AI chatbots that can respond more effectively to plain-language prompts, improving their ability to accurately address customers’ concerns.
Perhaps more impactful, however, is the ability of AI to analyze massive amounts of financial data to gain key insights and offer personalized recommendations accordingly. By deploying AI tools that are specialized to accurately process, extract and classify data from customers’ financial documents, financial institutions can deliver dynamic advisory services at a larger scale and lower cost.
In 2025 this will be increasingly accomplished by integrating the reliability of intelligent document processing (IDP) functionalities like natural language processing, optical character recognition and machine learning with the analytical and reasoning capabilities of large language models (LLMs). This augmentation accelerates decision-making and problem-solving steps that have traditionally been performed manually by humans, such as identifying and summarizing issues and discrepancies and recommending appropriate solutions.
AI agents and agentic AI systems have also emerged as a tool to accelerate these processes through autonomous decision-making, leveraging IDP and LLMs to maintain reliability while free of human oversight.
IBM’s watsonx is an example of a commercial generative AI used in tandem with IDP to excel in document-centric workflows. When augmented with IDP, watsonx can reliably uncover and analyze deep insights that enable more swift and informed decisions without heavy manual effort.
Beyond recommendations, this convergence of AI and big data analytics empowers institutions to make proactive, data-informed decisions that not only further enhance customer satisfaction and engagement but also improve their own bottom line. Institutions can develop a deeply granular understanding of key processes to identify bottlenecks and pain points that dampen customers’ journeys and negatively impact revenue cycles.
By tapping into the data previously sealed within financial documents or processes and using AI models purpose-built to quickly and accurately contextualize it, financial services institutions can make meaningful improvements to operational efficiency and the customer experience by way of data-driven process improvements and recommendations for services made with real behaviors and needs (of both customers and the business) in mind.
Heightening focus on AI responsibility
While customers are increasingly likely to expect the speedier and more personalized service enabled by AI, they’re also wary of its implications on their security and privacy.
As global regulation shines a spotlight on AI’s ethical quandaries, financial institutions are incumbent to be transparent and responsible in their use of AI to meet compliance with emerging standards and maintain customers’ trust. Strict focus on security and AI ethics, with the support of dedicated organizations such as ForHumanity, will further promote trust in AI as institutions adopt more stringent governance and compliance frameworks to safeguard against potential biases, breaches, and data misuse.
The Australian Securities and Investments Commission (ASIC) recently published a report detailing AI’s adoption and use in financial services, noting that AI adoption is outpacing development of risk and compliance frameworks in many cases. This misalignment creates risk of adverse impacts to sensitive decisions like financial scoring and prolongs a lack of transparency of deployed AI models.
Global guidance on AI in financial services echoes these concerns. SR Letter 11-7, for example, was issued by the Fed as guidance on AI model risk management for banking organizations. The letter holds these institutions to attentiveness to possible adverse consequences of AI-influenced decisioning including financial loss, poor business and strategic decision-making, or organizational reputation. Cooperating with AI ethicists and auditors can reveal banking institutions’ vulnerability to these outcomes and enable them to proactively mitigate the risks of AI bias and autonomous decisioning.
Financial institutions will find that prioritizing responsibility and transparency in their AI strategy will steer them further towards smaller, purpose-built AI models that are more explainable in nature and less likely to yield biased or inaccurate outcomes than their larger, more general counterparts. These models are trained to understand specific types of customers’ documents (such as bank statements, income documents and tax forms) and feed accurate data into an LLM as a basis for more grounded and reliable responses, reducing the likelihood of errant outputs.
While these measures bring certain benefits to customer experience and cost efficiency, they also contribute to the more complex and longer-term challenge of maintaining compliance with a constantly shifting global regulatory landscape.
Using AI to protect against AI-driven fraud
Recent research revealed that every Australian dollar lost to fraud costs financial institutions an average of AUD$4.21 in associated expenses. As AI continues to expand possibilities for committing fraud without human detection, AI also becomes increasingly crucial as a countermeasure against such malicious uses.
AI-enhanced fraud detection, with real-time AI models identifying and mitigating suspicious activities as they arise, will be the antidote to accelerated rates of fraudulent activities fueled by misuse of AI.
Agentic AI systems will also contribute to the real-time recognition of fraud in 2025. AI agents’ ability to make autonomous decisions with reliable accuracy reduces the need for manual human oversight in flagging potentially fraudulent activity, shifting the role of humans to confirming these instances instead of initially identifying them.
Additionally, automation strategies incorporating document AI can flag vulnerabilities and suspicious information in paper-based processes (where fraud is most likely to occur). Document AI leverages capabilities like intelligent character recognition to identify indicators of falsified documents like typos and formatting errors, while auditing and validation against KYC/AML rules further protect the integrity of document-based workflows.
In order to take full advantage of these AI innovations, institutions must ensure that they’re handling data (of their own and their customers) with best practices in mind, protecting against breaches and maximizing the value and reliability of AI deployments. Furthermore, it is recommended to focus on education, organizational buy-in, and incremental implementation to gradually introduce AI for long-term success.
If approached correctly, AI will not only elevate customer experience and operational agility in 2025 but also position the financial sector as a frontrunner in secure, responsible AI deployment.
