Roundtable Highlights: Turning AI into shareholder value

Roundtable: Turning AI spend into real business value

Artificial intelligence is rapidly climbing the strategic agenda across financial services, but many lenders remain in the early stages of turning investment into tangible returns. At a recent Credit Strategy roundtable, industry leaders discussed how their organisations are approaching AI adoption, where the most practical opportunities lie, and the barriers preventing wider deployment.

Despite strong interest from boards and investors, participants agreed that most institutions are still experimenting rather than scaling. Terms such as “embryonic”, “foundational” and “testing” were repeatedly used to describe current strategies, reflecting the cautious pace at which regulated firms are integrating AI into their operations. 

From experimentation to proof of concept

Across the discussion, AI initiatives were largely confined to pilot projects and internal tools. Several participants highlighted early success with contained use cases that help employees navigate internal knowledge or documentation.

One organisation has developed an AI-powered assistant capable of answering questions about company policies by retrieving and summarising information from internal documentation. While relatively simple, the tool has demonstrated the value of AI for reducing administrative overhead and improving access to information.

Elsewhere, teams are exploring both generative AI and traditional machine learning techniques. Predictive models are being tested to support lending insights and customer contact strategies, while generative AI tools are being evaluated for knowledge retrieval and automation across operational processes.

However, participants stressed that deployment remains cautious as firms balance innovation with their responsibilities as regulated lenders.

Governance and trust slow deployment

A key theme throughout the roundtable was governance. Unlike many other sectors, financial institutions must ensure that any AI-driven decisions remain transparent, fair and compliant.

This is particularly important when models influence lending decisions or customer treatment. Participants noted that while AI may uncover new predictive signals, organisations must carefully assess whether acting on those insights aligns with customer outcome principles and regulatory expectations.

As a result, many firms are running AI tools alongside existing processes rather than replacing them outright. This parallel approach allows organisations to test reliability and build internal confidence before integrating AI outputs into operational decision-making.

Efficiency gains lead the way

While fully automated decisioning remains some distance away, operational efficiency is emerging as the most immediate opportunity for AI adoption.

Customer service functions were highlighted as a particularly promising area. One lender is exploring AI tools that analyse historical customer interactions to provide call agents with summaries of previous conversations. The same technology could also generate structured notes following long customer calls, significantly reducing administrative work for frontline teams.

Compliance processes are another area attracting attention. Activities such as periodic know-your-customer reviews often involve large volumes of data gathering and documentation. Participants suggested that AI could support these processes by assembling and summarising information ahead of human review.

Starting from a clean slate

The discussion also highlighted the difference between organisations modernising legacy processes and those building new operations from scratch.

For newer banking initiatives, AI can be embedded from the outset. One participant involved in launching a new retail banking proposition noted that designing processes around AI from day one creates opportunities to rethink traditional operating models, potentially reducing the need for large customer service operations in the future.

For established institutions, however, legacy systems and entrenched processes make transformation more complex. Changing workflows that employees have followed for years can prove as challenging as implementing the technology itself.

Avoiding the automation trap

Participants also warned against simply applying AI to existing processes without rethinking how those processes should work.

Automating inefficient workflows risks amplifying problems rather than solving them. As one contributor noted, organisations must avoid the temptation to accelerate flawed processes through automation rather than redesigning them for the AI era.

A gradual path to value

While enthusiasm for AI is growing, the roundtable concluded that meaningful value will emerge gradually.

Most lenders are currently focused on experimentation and internal capability building rather than large-scale transformation. Early gains are likely to come from efficiency improvements before more complex applications like AI-driven decisioning become viable.

For now, the industry appears united around a cautious but optimistic view: AI will reshape lending operations, but doing so responsibly will require time, experimentation and careful governance.