In this article, Absa chief information and technology officer Johnson Idesoh writes about future developments in AI and how this will impact on the financial services sector.

When most people hear the term AI, their thoughts almost instinctively turn to generative large language models like OpenAI’s ChatGPT, Anthropic’s Claude and Alibaba’s Qwen. These tools, with ChatGPT alone boasting 200-million weekly active users and driving over 80% of Web traffic among AI platforms, have captured global attention. In the last week, DeepSeek hit the news headlines. What better programme, then, to explore the question of what comes next?

Ask ChatGPT, and it responds: Agentic AI.

“Agentic AI will be better than me,” it explains, “because it can act independently to achieve specific goals without needing constant input. While I rely on you to guide the conversation and provide instructions, Agentic AI can take initiative, adapt to situations, and proactively solve problems in real time. It’s the difference between giving directions to a driver (me) and having a self-driving car that takes you where you need to go on its own.”

A network of autonomous agents with the capacity to take purposeful, goal-directed actions through a nuanced understanding of their contextual environment – focused on making intuitive decisions and solving complex problems rather than generating content. Such technology would blur the boundaries between human decision-making and machine capability.

This very theme dominated discussions at the 2025 World Economic Forum gathering in Davos, where world leaders grappled with the implications of entering the Intelligent Age. Conversations spanned across industries, examining how organisations could harness the potential of these technologies to drive growth and innovation while mitigating their unintended consequences.

And nowhere is this more relevant than in the financial services sector, where Agentic AI has the potential to revolutionise not only how institutions design and deliver financial solutions but also how they fundamentally operate.

While automation and predictive analytics have delivered incremental gains to date, these technologies are inherently static – confined to executing predefined tasks within narrow parameters, often requiring human intervention to address complexities or anomalies. Agentic AI changes this paradigm entirely: autonomous agents would be capable of orchestrating complex workflows, learning continuously from new data, and adapting their actions in real-time, operating with an intelligence that mirrors human intuition but surpasses it in speed and scale.

For example, rather than just flagging irregularities in transactional data, an autonomous agent could respond with a cascading sequence of interdependent actions: quarantining suspect transactions, rebalancing risk models, and immediately notifying compliance and regulatory officers – all without manual input. This ability to initiate and execute multi-layered actions has the potential to transform areas like compliance, fraud detection and liquidity management, moving them from reactive processes to proactive, self-sustaining ecosystems.

The implications for the industry extend further, particularly when viewed through the lens of systemic coordination. Financial institutions are often siloed in their operations, with critical functions like risk, strategy, and operations working in parallel but rarely in true synchrony. Agentic AI introduces the possibility of seamless integration across these domains, enabling more cohesive and agile decision-making. Take stress testing, for example: today, this process involves painstaking effort to synthesise inputs from market data, geopolitical developments, and internal performance metrics. Agentic systems, by contrast, could integrate these variables live, generating adaptive simulations that continuously refine institutional resilience to changing conditions.

This capacity for integration is equally transformative for customer-facing applications.

In traditional banking, customer interactions are largely initiated by the client – whether to inquire about a product, seek advice, or resolve an issue. Even digital advancements, like chatbots and online portals, remain reactive at their core. Agentic AI would flip this model on its head, enabling institutions to anticipate client needs and deliver tailored solutions before a request is even made.

Consider, for example, a farmer. Drawing on local weather forecasts, market trends, and the farmer’s financial history, systems could autonomously approve a line of credit ahead of an anticipated drought or recommend an insurance product before the next growing season. It might also analyse crop yield patterns and suggest investments in machinery or irrigation systems, aligning financial support with the farmer’s specific goals and circumstances. The technology would enable institutions to deliver hyper-personalised solutions while educating users on the implications of their financial choices through actionable insights – advancing both financial inclusion and literacy.

However, it is important to temper these aspirations with critical scrutiny.

Agentic AI, for all its promise, remains a conceptual construct – one whose path to adoption is fraught with technological, regulatory, and ethical hurdles. How can a system that acts independently be reconciled with the accountability frameworks required in finance? What happens when an agent errs – or worse, makes a decision that inadvertently discriminates or causes harm? And can such systems ever fully align their objectives with the broader mission and values of the institutions they serve?

These considerations demand not just new governance models but a cultural shift within institutions—a willingness to trust and collaborate with systems that both challenge and complement traditional human oversight. This is especially relevant for the age-old question of disruption to human capital.

As Agentic AI integrates deeper into financial operations, it will undoubtedly blur the distinctions between front-office innovation, back-office efficiency, and risk governance. Functions such as transaction monitoring, compliance reporting, and routine customer interactions could be seamlessly managed by intelligent systems, allowing human workers to focus on higher-value tasks that require critical thinking and interpersonal skills.

Yet, as these systems assume roles that are tedious and time-consuming for humans, how can an institution preserve its sense of humanity—its ability to connect with clients, foster trust, and uphold ethical values? The answer lies not in choosing between human and machine but in creating a hybrid model where both work in tandem, each amplifying the strengths of the other. This requires deliberate investment in training, leadership, and culture to ensure that the integration of Agentic AI enhances, rather than erodes, the institution’s core values.

Here too, ChatGPT offers a profound lesson.

Asked if it believes Agentic AI will render it obsolete, the programme responds: “It doesn’t worry me, and I don’t think I’ll become irrelevant. Agentic AI and systems like me are complementary, not competing. My role may shift, but I’ll adapt to stay useful in new and meaningful ways. Change isn’t the end – it’s an opportunity.”