Genuine transformation using AI demands that those at the centre, that is, organisations, people, and societies, engage as architects, not merely as spectators or passive beneficiaries. In a recent episode of The Cambridge Executive Business Insights: Rethinking AII podcast, Oscar Barlow, Head of AI Advocacy at Starling Bank, explores how digital-first banking can embed AI not as a distant, opaque tool, but as an agent of purposeful, accessible change, for all.
Purposeful AI in an era of lean innovation
Rethinking AI in banking is not about making systems ever bigger or more powerful; it is, instead, about making them smarter and more purposeful. Barlow’s conversation with podcast host Jaideep Prabhu places this mission squarely in the wake of the financial crisis, a time when the established models of banking stood exposed and vulnerable. What emerged was a new wave of ‘challenger’ banks (Starling Bank being one of them), conceived from inception as digital, agile, technology-led organisations, a marked departure from their asset-heavy, legacy-bound progenitors.
As Oscar Barlow puts it, Starling Bank defines itself as “a technology company that is also a bank”, the order of priorities telling. The agility encoded into Starling’s DNA is not accidental; rather, it is a deliberate response to the conformity seen in traditional banks, where complexity and inertia stymied both innovation and genuine customer focus. For Barlow, the critical insight is that being frugal or lean need not be an exercise in compromise; instead, it is a structural enabler, allowing organisations to rapidly respond to change and to centre customers in ways that traditional incumbents struggled to achieve.
Digital DNA: laying the groundwork for AI
The digital-native nature of Starling means that AI is not so much an afterthought or cosmetic utility as it is a natural extension of the bank’s foundational systems. This marriage of cloud-enabled architecture, agile culture, and relentless customer focus means that any layer of intelligence, including AI, integrates not as an awkward add-on, but as an organic evolution of capabilities.
In describing Starling’s approach, Oscar Barlow draws a compelling contrast with previous waves of enterprise transformation such as the move to become more data-driven. Those earlier initiatives, though impactful, often demanded laborious centralisation of data, new workflows, and a degree of cultural change that taxed both systems and the professionals using them. AI, by contrast, presents a striking accessibility: “You talk to it, and it talks back to you.” The conversational interface of generative AI lowers barriers, making advanced technology feel approachable, unthreatening and, crucially, empowering for staff across all disciplines.
From AI anxiety to AI empowerment
Organisations, no matter how advanced, are ultimately communities of people. The journey from AI anxiety to AI empowerment is thus as much about culture and trust as it is about coding and algorithms. Jaideep Prabhu and Oscar Barlow dissected the social dimensions of transformation: while internal teams display growing curiosity and even excitement about AI’s potential, broader public sentiment often veers toward caution and even fear, shaped by previous, disruptive waves of technological change.
Starling’s answer to this anxiety is twofold. Internally, the bank cultivates a dispersed, deliberately peer-led community of “AI champions”- trusted colleagues embedded within every division, tasked with sharing good practices, troubleshooting problems, and, crucially, making it safe for others to admit gaps in understanding. In environments where AI feels unfamiliar or intimidating, this approach turns learning into a communal, trusted endeavour, rather than a top-down imposition.
Externally, Starling are deliberate in putting tangible, customer-facing AI tools into users’ hands. For example, AI-powered spending intelligence and marketplace fraud protection tools that speak plainly to real-world needs, demystifying AI by virtue of its obvious utility. By “earning public trust” through authentic, everyday value, Starling secures its license to push further innovation.
Building an AI-fluent culture: decentralised enablement over centralised control
Perhaps the most illuminating insight is Starling’s philosophy on organisational enablement. Rather than centralising expertise in a distant centre, AI fluency is cultivated as a living, distributed attribute, which is owned by everyone. The “AI champions” network, comprised of 50-60 individuals across all areas, acts as a catalyst for grassroots innovation by marrying domain expertise (e.g., legal, finance, marketing) to technical understanding of AI. As Oscar notes, it is rarely effective for a tech specialist to dictate practices to, say, a team of experienced lawyers. It is far more productive to equip those lawyers with one of their own, fluent both in their unique professional context and in AI tool usage.
This peer-embedded model does much to lower the psychological barriers that often accompany rapid change; people are far more willing to express confusion or seek help when the “expert” is a peer, not an outsider, making learning safe, candid, and incremental.
The training itself is notable for its rigour. Every employee undergoes foundational education, not only in the technical characteristics of AI, but also in its societal risks, for example, bias and hallucinations. Starling has even implemented internal audits of AI outputs to check for bias, finding relatively little so far, but using lessons learnt to further refine practice and vigilance.
Accountability in the age of automation
As AI systems absorb more responsibility for both customer-facing and operational tasks, the question of accountability looms large, especially in a sector as tightly regulated as banking. Here too, Starling’s stance is unambiguous: the use of AI is always the responsibility of the human who employs it. Delegation to a tool does not absolve one of judgement or oversight. Cases in which, for example, a member of staff might seek to attribute a decision solely to an AI output (“the AI said this customer was suspicious, so we acted”) are explicitly disallowed.
Structural safeguards underpin this principle: access to AI tools is closely controlled depending on the level of exposure to sensitive customer data, and where such exposure exists, enhanced controls are mandatory. Rather than relying on external tools, Starling prefers to build bespoke, tightly integrated solutions within their own platform, allowing for nuanced control and traceability of AI-assisted workflows.
The returns on AI: hard metrics and human experience
A persistent challenge for many organisations is the quantification of AI’s return on investment. Starling’s approach is refreshingly pragmatic. Not only do they cite significant wins in areas such as fraud detection and transaction monitoring, but internal tools have delivered hard efficiencies: for example, by having AI generate written summaries of customer service calls, Starling has managed to save around 8,000 hours per month. These improvements are not abstract; they directly translate to reduced customer wait times, improved staff productivity, and a knock-on effect on workforce morale and attrition.
But the benefits are more nuanced than mere efficiency. As Jaideep Prabhu points out, roles at the frontline of customer service are notoriously high-stress and high-turnover; any tool that meaningfully enhances the daily working experience creates value not only for the balance sheet, but for the wellbeing and engagement of the people at the heart of the bank.
Democratising financial intelligence: AI for customers
The external, customer-facing applications of AI at Starling illustrate a glimpse of what “banking for everybody” might look like in an AI-powered future. In simple, approachable ways, users can interrogate their spending using everyday language; “How much did I spend on coffee in the past three weeks?”, and receive clear, actionable answers, helping to foster better financial habits and control. Starling’s upcoming launch of the UK’s first AI assistant, Starling Assistant, promises even greater empowerment through natural language interfaces: saving for travel, managing budgets, and even automating transfers can all be achieved through conversational prompts.
Significantly, even as the interface becomes “smoother”, the principle of human control is preserved. No automated action is performed without explicit confirmation, balancing convenience with safety. AI, in this vision, is not about displacing the customer’s agency but about making complex tasks intuitive and accessible.
Towards agentic AI: next-generation workflows
Looking to the future, Oscar Barlow acknowledges that the most fascinating questions concern the evolution of work itself with AI as a ubiquitous teammate. If the last wave of digital transformation saw organisations moving clumsily from paper to “digitising” processes (sometimes amounting to nothing more than uploading PDFs online), the genuine goal now must be a deeper, more integrated transformation where distributed AI agents become part of workflows, where they are seen not as mere additions, but as collaborative, adaptive contributors.
The challenge is profound: how to design teams where human professionals coordinate seamlessly with not just one, but potentially many instances of intelligent agents? This is not simply about automating menial tasks; it is about reimagining roles for a world where complex judgement, creativity, and adaptation are shared between people and machines. The answers, Oscar Barlow suggests, will emerge only as organisations experiment, learn, and remain willing to adapt in the face of continual change.
Compliant by design: innovation within the regulatory landscape
Innovation in banking can never be divorced from regulation, a fact that Starling has turned to its advantage. The principle-based regulatory regime in the UK provides both clarity of standards and the flexibility to respond to novel developments. As Oscar Barlow explains, every new feature goes through rigorous privacy and compliance assessments, with special attention to the prohibition on activities such as unlicensed financial advice.
Moreover, industry-wide AI leadership is now formalised: Starling’s CIO, Harriet, has been appointed as the UK financial services industry’s first official AI Champion, an accolade reflecting both the bank’s progress and the regulator’s appetite for well-governed innovation. Initiatives like regulatory “sandboxes” and innovation labs flow from the central insight that guidance should not be about stifling new ideas, but about stewarding them safely into reality.
Rethinking organisational change: one size doesn’t fit all
In seeking lessons for other banks, corporates, and indeed academic institutions, Barlow’s reflections are instructive. The way an organisation approaches AI will, and indeed should, mirror its broader culture. If that culture is experimental, empowered, and peer-led, then so too will be the most effective AI adoption journeys. The practical advice: start with deep organisational introspection. Ask: “What kind of organisation are we, really?” Build on those truths rather than chasing formulaic, external blueprints. Listen to and amplify the voices within who are already doing innovative work; let them become internal teachers and multipliers. Ensure that leadership not only gives permission for experimentation but signals its urgency and value.
The cautious optimism of a digital future
The journey to an AI-embedded future is neither linear nor without its challenges. Starling Bank’s experience, richly discussed in this episode, suggests that the best results flow not from grand, monolithic transformations, but from careful cultivation of a learning culture, bottom-up experimentation, and a relentless focus on human responsibility and trust. AI, for everybody, is not a technical project alone, it is a project tied to leadership, culture, and people.
For more insights on building meaningful, inclusive, and responsible AI in banking and beyond, listen to the full episode of this episode of Cambridge Executive Business Insights: Rethinking AI, available wherever you get your podcasts.
