The evolution of agentic AI in banking, Stefanos Athanasiadis, MD Investment Solutions, CEO, Profile Centevo, Profile Software

By Robin Amlot

Stefanos Athanasiadis was named Managing Director, Investment Solutions by Profile Software in February 2026. He is also Chief Executive Officer of the company’s Nordics-based Profile Centevo, which offers asset and fund management solutions, delivering digital services for more than 65 clients.s. Athanasiadis tells us why agentic AI is becoming a strategic priority for banks:  “Agentic AI is the evolution of GenAI, moving from insightful information to execution and autonomous processes, as banks are under pressure to do more with speed, accuracy and control, while responding to increasingly complex customer and regulatory demands.

“The shift now is from a first wave of assistive AI, answering questions and drafting documents, to a second wave that executes complete processes end to end. That second wave is what changes the operating model, and it is where banks are least prepared.

“What makes it especially relevant now is its ability to support high-value banking priorities such as liquidity management, forecasting, customer servicing and operational automation. It allows institutions to turn large volumes of structured and unstructured data into actionable insight, improve decision-making and reduce manual effort across the organisation.

“At the same time, banks are moving beyond isolated pilots. They are looking for AI that is embedded into core platforms, backed by governance and designed to scale and execute. That is where enterprise deployment becomes a real differentiator.”

What does agentic AI mean in practice for banks?

“In practice, agentic AI means embedding AI into the bank’s daily operating model across front, middle and back-office functions related to execution, with the ability to autonomously perform tasks while others require human approval. This allows banks to:

• automate repetitive, knowledge-intensive workflows such as reporting, document handling and customer communications;
• improve liquidity visibility and forecasting through faster access to contextual insights;
• support staff with intelligent assistance for decision preparation, exception handling and service execution.

“The real value comes when AI is not treated as a standalone add-on, but as part of the systems banks already use. That is what makes it more consistent, secure and scalable, while freeing teams to focus on higher-value work.

“This is not theoretical. In a live fund-administration operation, an agentic process now runs daily reconciliation on real trading data with full accuracy, escalating only genuine exceptions to a human.”

What business value can agentic AI deliver to banks at enterprise scale?

“At enterprise scale, agentic AI can create measurable value across operations, decision-making and customer engagement. Banks can:

• accelerate forecasting and planning processes with faster synthesis of internal and market information;
• reduce operational friction by automating content-heavy and process-heavy tasks;
• improve service consistency and responsiveness across customer touchpoints;
• strengthen internal productivity while supporting governance and auditability;
• automate and accelerate tasks and operations through the autonomous execution of processes.

“For Profile Software, this value goes well beyond a single banking function. It spans treasury and core banking operations, while also extending clearly into investment management, covering the full spectrum from wealth management to custody. With Axia Suite and its AI-related capabilities, we support a unified investment management environment covering private and institutional wealth, custody, asset and fund management, while also addressing the needs of retail and mass affluent segments. This is important because it allows financial institutions to combine efficiency with more personalised and scalable client service across multiple business lines.”

Why are banks moving from AI experimentation to enterprise-wide deployment now?

“Many banks have already tested AI in isolated scenarios. What is changing now is the need to turn those pilots into scalable business outcomes, measurable savings and performance optimisation.

“Economic pressure, rising service expectations and growing operational complexity are pushing institutions to invest in technologies that can deliver enterprise-wide impact. Agentic and generative AI support multiple functions at once, from workflow automation and content generation to insight extraction and execution.

“However, sustainable deployment requires more than innovation. It requires the right architecture, trusted data, clear governance and business alignment. That is why banks are increasingly prioritising solutions that are enterprise-ready from the outset.”

What are the main challenges banks face in adopting AI?

“The main challenges are typically related to integration, governance, data quality, and operational trust. Banks need to ensure:

• secure access to high-quality, governed data;
• alignment between AI outputs and regulatory or internal control requirements;
• seamless integration with existing banking platforms and workflows;
• accuracy and reliability of results when the system is asked to execute, not just advise.

“Without these foundations, even the most advanced AI models cannot deliver reliable business value at scale. This is why successful adoption depends on combining innovation with control, so that AI becomes a trusted operational capability rather than an isolated experiment.

“One lesson from production is counter-intuitive: most of the value does not come from heavy AI reasoning. In our own deployments, the majority of steps are deterministic rules or single, simple model calls; only a small fraction requires complex reasoning. Treating every step as an ‘AI problem’ is how projects become slow, expensive and hard to govern.”

How does Profile Software differentiate its approach to delivering AI in banking?

“We focus on enterprise-ready solutions where AI is connected to real banking and investment use cases, rather than presented as innovation for its own sake.

“Being on the frontier of agentic process automation requires three things at once, and few providers have all three: the systems, the domain expertise, and the agentic technology to make AI execute reliably, under governance, with humans in control. Intelligence layered on top of someone else’s stack is the easy part; the hard, defensible part is combining all three so AI executes inside real systems and processes. Specifically, we:

• bring the systems: the core platforms where banking, investment, wealth and custody data and processes live, anchored by Axia Suite and our broader platform footprint;
• bring the domain expertise: people who run these financial operations day-to-day and know where the real exceptions and risks are;
• bring the agentic technology: the architecture to make AI execute reliably, under governance, with humans in control;
• combine all three so AI executes inside each institution’s real systems and processes, not as a generic assistant bolted on top.

“We build tailor-made agentic processes on each client’s own systems and workflows, not a generic assistant bolted on top, but processes designed around the institution’s actual data, controls and operating model. This matters because the highest-value processes depend on internal systems that external, off-the-shelf AI tools cannot reach.

“This approach allows institutions to deploy AI with confidence, scale it and create a more connected and responsive operating model.”

What trends will shape the future of generative and agentic AI in banking?

“We see three trends shaping the next phase of agentic AI adoption in banking:

•  Embedded AI across enterprise workflows, where intelligence and execution capabilities are built directly into banking platforms and user journeys.
•  More personalised servicing models, particularly in wealth, advisory and affluent banking segments where contextual engagement matters most.
•  Unified operating environments, connecting banking, treasury, wealth and custody functions through interoperable, AI-enabled platforms that support execution at scale.

“As adoption matures, the institutions that will stand out are those that combine innovation with operational discipline. Reflecting this investment, Profile Software operates a Stockholm-based centre of excellence for agentic operations and R&D, close to the clients we serve.

“Generative AI, combined with agentic AI, is becoming a practical enterprise capability for banks, provided it is deployed with the appropriate human approvals. The institutions that win will not be those that deploy the most AI, but those that deploy it where it actually pays, with the discipline to keep humans in control of the decisions that matter.”