It has become much easier to experiment with AI.
Teams can test new tools quickly, build prototypes and explore applications that would have taken much longer to develop only a few years ago.
But for banks, proving that something can work is very different from making it work across the organisation.
On 25 August 2026, CFTE welcomed Daniel Minarik, Managing Director, Data and AI, Financial Services at Accenture, for the webinar From AI Experimentation to AI as Usual: How Banks Scale AI Across the Enterprise.
Drawing on his current work with financial institutions and his previous experience as Chief Data and Innovation Officer at Tatra banka, Daniel shared a practical view of what happens after the experimentation phase.
His central message was simple: the technology is only part of the challenge.
Banks have been using AI for years. What has changed is how people interact with it.
AI itself is not new to banking.
Machine learning, predictive analytics, scoring and risk models have been used across financial institutions for many years. Much of this happened behind the scenes, without customers or employees necessarily knowing that AI was involved.
Generative AI changed that.
The introduction of conversational tools brought the technology directly in front of users. Employees could interact with it themselves, while customers gained access to powerful tools outside the traditional banking environment.
For Daniel, this shift in the interface matters because it changes expectations.
Employees increasingly expect to have useful tools available at work. Customers have more options for where they ask financial questions and how they access information.
Banks therefore need to think not only about the technology they deploy, but about how the experience of banking itself may change.
Scaling requires more than a list of use cases
One of the common starting points for organisations is to ask for use cases.
Where can we apply it?
What can we automate?
What can we show quickly?
Daniel suggested looking more broadly.
He outlined four areas that need to develop together:
Governance and technology
Banks need the right tools, security, processes and guardrails to use AI safely.
Change management and education
Employees need to understand what these tools can do, where their limits are and how they apply to their own work.
Business value and use cases
Projects should support the wider priorities of the bank rather than exist simply because the technology is available.
Data readiness
Banks may have large amounts of valuable information, but it needs to be organised, accessible and appropriately governed before it can support useful applications.
These areas are closely connected. Better education, for example, can help employees identify stronger opportunities based on the work they already understand.
The best ideas may come from the people doing the work
Education was one of the strongest themes of the session.
Daniel explained that when employees understand the technology better, the quality of the ideas coming from the organisation can improve.
The people working within a process every day often know where the delays, repetitive tasks and difficult decisions are.
They are therefore well placed to spot where a new tool may genuinely help.
This creates a different approach from a central technology team trying to invent every use case.
Instead, employees create demand based on real business needs, and the organisation can then decide which ideas are worth taking further.
Daniel described this as a more demand-led approach to building the pipeline of opportunities.
A good demo does not mean you are ready to launch
This was perhaps the most practical point from the webinar.
It is now possible to produce an impressive demonstration very quickly.
But a demonstration mainly proves the experience.
Once a bank decides to put that idea into production, a much longer list of questions appears.
Is the right data available?
Who owns it?
Can the solution work with existing systems?
Does it meet internal policies?
What needs to be reviewed by security or compliance?
How will it be tested and monitored?
Who will own the process once it is live?
Traditional software development, integration and control processes do not disappear simply because the front end was quick to build.
Daniel noted that, in more complex cases, the AI component can represent only a relatively small part of the total work involved. The harder part is often getting the solution safely into the existing banking environment.
This creates a new challenge for banks.
The problem is no longer generating enough ideas.
It is deciding which ideas should move forward and making sure the organisation has the ability to deliver them properly.
Clear ownership matters
Scaling also becomes difficult when responsibility is spread across too many teams without clear accountability.
Technology may own one part.
Data another.
HR may lead education.
Business teams may own individual applications.
Risk, security, legal and compliance all have roles to play.
Daniel’s view was that these responsibilities do not all need to sit within one department, but there should be one clearly accountable owner for the overall strategy.
That owner then needs a way of bringing the right people together.
Instead of involving security, compliance or legal only after a project has already gained momentum, these functions can be brought in earlier so important questions are addressed before they become delays.
What does “AI as usual” look like?
The title of the webinar points towards an important idea.
Today, organisations often talk explicitly about whether employees are “using AI”.
They measure logins, usage, licences and adoption.
Daniel suggested that the longer-term destination may look quite different.
If the technology becomes properly embedded into everyday tools and processes, employees may eventually stop thinking about whether they are using AI at all.
It simply becomes part of how the work gets done.
That is what AI as usual means.
Not endless experimentation.
Not adding another standalone initiative.
But gradually making useful capabilities part of ordinary business.
The bigger challenge is changing the organisation
Daniel closed the session with a point that goes beyond technology.
Many of the same models, platforms and tools are becoming available to organisations across the market.
The difference may increasingly come from how organisations respond to them.
Can they educate their people?
Can they organise the right teams around a problem?
Can they make good decisions about what to pursue?
Can they adapt their processes as the technology changes?
Daniel described the ability to change how an organisation operates as one of the most important forms of innovation.
For banks looking to move beyond pilots, this may be the real challenge.
The technology can help create the opportunity.
The organisation still has to make it work.
Thank you to Daniel Minarik for joining CFTE and the Global Women in AI community, and for sharing his experience and practical perspectives on what it takes to move from experimentation towards everyday use across a bank.
