At The Banking Scene, we’ve long argued that the future of banking will be shaped less by technology itself and more by the way banks integrate intelligence, both artificial and human, into their core. That belief was strongly reaffirmed during our recent roundtable, “Banking on AI: Exploring the AI-Accelerated Future of Finance,” hosted in collaboration with Cognizant and Microsoft.

The idea of the Agentic Enterprise took centre stage, sparked by Cognizant’s David Fearne, Global Director of Generative AI's framing, and brought to life by diverse, candid perspectives from industry practitioners. But this wasn’t a conversation about revolution. It was about evolution: how financial institutions are gradually transitioning from isolated AI deployments to orchestrated, intelligent systems that reflect, not replace, human decision-making.

During the discussion, Alok Chaurasia, Cognizant FS Lead Belux, mentioned that they believe in "Being AI-powered" rather than "AI Everywhere," focusing on contextualised problem solving and reimagining the business journey through use cases across three vectors: Vector 1 – Tech for Tech, Vector 2 – AI for Existing Business, and Vector 3 – Agentification & Service-as-a-Service.

This concept finds a compelling complement in the Cognizant–Microsoft white paper, “Intuitive AI: Invisible Intelligence, Visible Impact”, which outlines how banks can embed AI into the very structure of their organisations, not just to automate tasks, but to transform how decisions are made and delivered.

From AI Tools to Agentic Thinking

The agentic enterprise reframes how we think about deploying AI. Instead of standalone copilots or bots, imagine an internal ecosystem where AI agents operate across three levels: task-level agents perform specific actions, mid-level agents coordinate business domains, and leadership agents synthesise strategic objectives.

Some bankers described this as a “mirror organisation”: a digital twin of human governance and intent. These agents interact through natural language, reason based on context, and collaborate to make decisions, supported, not replaced, by humans.

This layered structure aligns directly with the “Intuitive AI” framework, where systems are organised around intent, cognition, information, and presentation. It’s no longer enough to automate a workflow. Banks must start designing AI systems that understand, communicate with, and collaborate with—both internally and with customers.

The Real Foundation? Information, Not Just Data

One of the most important shifts in thinking came in how we define AI readiness. Traditional models treat data as fuel: raw material for predictions. But GenAI operates differently. It starts not from datasets, but from information: well-documented processes, contextual business knowledge, and properly structured content.

Several participants observed that “GenAI is about information in, decision out” - and this changes the game. Large datasets aren’t always required. In fact, smaller organisations, with simpler and more easily documented processes, may have a competitive advantage.

That said, challenges persist. Banks still wrestle with:

• Poor data classification
• Fragmented repositories
• Unclear access controls

Tools like Microsoft Purview and Entra can help, but as one banker noted, “You don’t need better AI; you need better metadata.” And often, the barrier isn’t technical. It’s architectural.

Enterprise Architecture: The Silent Enabler

The agentic model only works if it is built on a coherent enterprise architecture. Process maturity, governance clarity, and system integration become preconditions—not afterthoughts.

This is not something IT alone can fix. As discussed, the métier of ICT may need to evolve toward enterprise design to support AI orchestration. Business architecture becomes the bridge between strategy and intelligent automation.

Without this foundation, agentic initiatives risk becoming islands of intelligence in oceans of entropy.

Three Domains of Opportunity

Across all the insights shared, use cases began to fall into three primary categories:

1. AI for Clients

This includes chatbots, virtual agents, and personalised services. Some banks already use GenAI to summarise portfolios, anticipate needs, and recommend actions, especially in retail and wealth contexts.

2. AI for Processes

Here we saw impressive traction. AI is helping automate KYC, AML, and compliance reporting. Some agents can now perform regulatory checks and even rewrite customer documents in line with Consumer Duty requirements, while maintaining full audit trails.

3. AI for Employees

This is about augmentation, not replacement. Internal copilots support onboarding, policy navigation, and knowledge retrieval. In some cases, they now serve as productivity interfaces across tools such as Excel, Outlook, and internal CRMs.

And this augmentation-first approach matters. As one banker pointed out, “Use AI to do all the things we want to do but never had time for - before replacing people.”

Orchestration: The Missing Middle

Perhaps the biggest theme to emerge was the need for orchestration—connecting siloed agents into a functioning system.

While many banks have pilots in place, few have built coordination layers between them. Yet this is where real transformation lies. Without orchestration, we’re just layering intelligence onto fragmented processes. With it, we create a responsive, adaptive enterprise that can reason, decide, and act at scale.

This is where the white paper’s “Intuitive AI” lens becomes essential. Intent, cognition, and presentation aren’t just design principles, they are orchestration rules. They ensure agents don’t just work, they align.

Ethics, Trust, and the Human Compact

As AI becomes more embedded, the ethical lens sharpens. The conversation shifted from abstract concerns about bias to very real fears of displacement, particularly in junior roles across legal, compliance, and operations.

To address this, several banks have introduced AI charters: internal governance documents that define how AI is used, who remains accountable, and what principles guide adoption. These charters serve as trust mechanisms, but also as change management tools, helping staff understand not just what is changing, but why.

There was also talk of AI joining the boardroom, not as a voting member, but as a reasoning engine, surfacing insights or challenging assumptions. In time, we may see AI not just executing strategy, but informing it.

Reframing the Strategic Risk: Attention, Not Intelligence

A powerful closing thought came from one participant who flipped the dominant narrative: “If the cost of intelligence goes to zero, the customer becomes the bottleneck.”

In a world where every provider can generate insight, summarise terms, and optimise offers, the scarcest resource isn’t intelligence, it’s attention. Differentiation won’t come from how smart your AI is, but from how clearly it communicates, how well it earns trust, and how meaningfully it connects.

In this light, go-to-market becomes a strategic application of AI, and the human touch becomes a competitive edge.

Final Thought: Same, But Different

As I closed the session, I returned to a simple question: what becomes of banking when intelligence is abundant and always-on?

The answer I received has stayed with me:

“The same, but different.”

The core values: trust, value creation, and customer centricity, will remain. But the organisations that deliver on those promises will evolve. They’ll be flatter, faster, and more agentic.

They’ll rely on systems that understand intent, process information, reason through complexity, and present it with clarity, just as the Cognizant–Microsoft white paper so elegantly describes.

But above all, they will need to stay human.

Because when intelligence is free, the banks that win will be those that remain worthy of attention.

🔗 You can read the full Cognizant–Microsoft white paper, “Intuitive AI: Invisible Intelligence, Visible Impact,” here.