There's a question I've been sitting with for a while now, and I suspect I'm not alone: at what point does AI stop being something building societies experiment with and become something they genuinely operate through?

We're closer to that inflection point than many leaders realise. And for building societies, the moment matters more, and the stakes are higher, than for almost any other type of financial institution.

The structural problem that AI has exposed

For the past few years, much of the conversation about AI in financial services has focused on pilots: intelligent chatbots, document summarisation tools, productivity copilots. These are genuinely useful. But they sit on top of organisations that were designed around something else entirely, which is functional specialisation. Finance, operations, HR, compliance, risk, each with their own systems, their own data, their own rhythms.

New research from the IBM Institute for Business Value, published this spring, puts a sharp number on this challenge. Across a survey of 2,000 C-suite executives globally, 82% said that functional silos are actively blocking the value that AI could deliver. Not a talent gap. 

Not a technology gap. A structural one

The problem isn't that societies lack ambition or capability. It's that the operating model itself was built for a different era.

For building societies, this lands with particular force. Many operate on core systems that are decades old, with data locked inside function-specific architectures that were never designed to speak to each other. When AI tries to automate across those boundaries, it hits walls. The result is AI that handles isolated tasks rather than driving outcomes, which is exactly where a lot of organisations find themselves today.

What the next shift actually looks like

The IBM research describes what comes next as the "interconnected enterprise," an operating model organised around end-to-end workflows rather than departmental functions. In this model, AI agents handle routine execution continuously and at scale, while human experts concentrate on judgment, oversight, and the decisions that genuinely require experience and nuance.

In a building society context, think about the mortgage journey

Today, that process typically involves handoffs between origination, underwriting, compliance, valuation, and completions, each with its own team, its own queue, its own bottlenecks. Agentic AI doesn't just speed up one of those steps. It can orchestrate across all of them, handling the data gathering, the rule-checking, the documentation, and the scheduling, while surfacing to a human the specific decisions that require judgment. A complex self-employed application, a borrower with a non-standard income profile, a potential affordability concern. 

These are exactly the moments where experienced colleagues add irreplaceable value.

The AI handles the rest.

This is not a distant vision. The IBM research found that 55% of organisations globally are already actively developing or deploying an agentic AI operating model. And organisations that have made the structural changes to support it are projecting significant performance improvements by 2027, including 45% gains in operational efficiency and 23% in revenue growth.

Why building societies are actually well positioned

Here's where I think the mutual sector narrative often gets the framing wrong. Building societies tend to talk about AI adoption as if it's a game of catch-up with larger banks, as if the constraint is resources or scale. But the IBM research points to something more interesting. 

The primary accelerant of autonomous workflow adoption isn't technology investment. It's change management readiness and governance. When those two capabilities are strong, the odds of successful adoption increase dramatically.

Building societies have something the largest institutions often struggle to replicate: organisational intimacy. Fewer layers of bureaucracy. Leadership closer to the operational reality. A culture, when it's working well, where change can be communicated honestly and staff feel genuinely part of what's being built. That's not a consolation prize for being smaller. It's a genuine structural advantage when it comes to the kind of cultural and operational change that agentic AI requires.

There's also a member trust dimension that matters here. 

The IBM research highlights data quality and governance as a critical foundation, with 77% of executives surveyed investing directly in this area. For building societies, where the member relationship is the entire point of the enterprise, the discipline of handling data responsibly isn't just a compliance requirement. It's a competitive differentiator. Members who trust their society with their financial data, savings history, mortgage affordability, payment patterns, are providing an asset that, governed properly, can power genuinely personalised, timely, and useful service. Not the kind of personalisation that feels like surveillance, but the kind that feels like being known.

The branch question

I want to say something about branches, because the conversation around them often slides into a false binary of digital versus physical, or modern versus traditional.

The most interesting application of AI in a branch setting isn't a customer-facing kiosk or a digital queue system. It's everything that happens in the background that currently takes a colleague's time and attention away from the member sitting in front of them.

If AI is handling the administrative load, the compliance checks, the document preparation, the case notes, the follow-up scheduling, then the branch colleague can be present in a different way. Not distracted, not working through a checklist, but genuinely listening. That's the version of the branch that members value most. And it's the version that technology, done well, can help to protect and strengthen rather than replace.

The governance question no one should skip

The IBM research is direct on something that deserves more attention in our sector. 

Autonomous operations require new governance structures, not just new tools. The research describes an orchestration layer that gives leaders continuous visibility into how AI is performing across workflows, where exceptions are occurring, and where human intervention is needed. This matters for building societies not just as a management tool but as a regulatory one. The FCA's evolving expectations around AI explainability and the treatment of vulnerable customers mean that "we trust the model" is not an acceptable governance posture. Societies need to be able to show what their AI is doing and why, at a workflow level, not just in the abstract.

The research also found that when six foundational capabilities are in place together, namely change management, AI governance, data governance, real-time data integration, system interoperability, and financial integration, autonomous workflow adoption is more than five times more likely to succeed. The effect is multiplicative, not additive. Investing in one or two of these in isolation, while leaving others underdeveloped, significantly limits the return.

What I'd encourage leaders to sit with

The IBM report frames the current moment as a leadership test as much as a technology one. I think that's right. The organisations that will get the most from agentic AI aren't necessarily those with the largest technology budgets. They're the ones whose leadership teams are willing to ask harder questions.

Which of our workflows genuinely create member value, and which exist because they've always existed? Where does human judgment add something irreplaceable, and where are we using people as expensive data-processing machines? How do we talk honestly with our colleagues about what's changing, and what the new shape of their roles looks like?

Building societies have spent 200 years answering to their members, not their shareholders. That orientation, long-term, relationship-first, community-rooted, isn't a constraint on what's possible with AI. It's the reason the opportunity is so significant. 

The question is whether we use this moment to deepen what makes building societies distinctive, or whether we let the technology reshape us into something that looks more like everyone else.

I know which I'd rather see.

________________________________________

The IBM Institute for Business Value report 'The Blueprint for Agentic Operations: How to build an interconnected enterprise' is publicly available and draws on a survey of 2,000 senior executives across 16 countries and 17 industries.