Why the European AI Regulation is the best argument for design in banking
The European AI Act requires that AI systems be transparent, explainable and controllable by the user. For the design, that's not a restriction — it's the best argument it's had in years. Organizations that design with that criterion now have an advantage over those that wait for the regulator.
Why the European AI Regulation is the best news design has had in years
When the European AI Act came into force in 2024, most organizations read it as a compliance issue. A new regulation, a new cost, a new layer of bureaucracy on projects that were already complex in themselves.
We read it differently.
The European AI Regulation is, among other things, a document that says that AI systems that make decisions that affect people have to be transparent, explainable and controllable by the user. That the person who receives an automated decision — about credit, about a public benefit, about insurance coverage — has the right to understand why that decision was made and to challenge it.
That's not bureaucracy. That's user-centered design with the force of law.
The problem that design has been trying to solve for years
For years, design teams have argued that transparency and explainability in digital systems are not just ethical principles — they are usability factors. A user who does not understand what a system does does not trust it. A user who does not trust it does not use it. And a system that no one uses is of no use, no matter how technologically sophisticated it is.
That argument has convinced some product teams. It hasn't always convinced management committees, legal teams, or product managers who prioritize speed of release over quality of experience.
The AI Act changes the conversation. Now transparency is not a design recommendation — it is a legal requirement for any high-risk AI system. And in banking, insurance and public administration, almost all AI systems that matter are high risk.
What does it mean to design for the AI Act
The Regulation establishes, among other things, that high-risk AI systems must:
— Be transparent enough so that its users can interpret the results and use them appropriately.
— Allow human supervision effectively: the people responsible for using the system have to be able to understand what it does and intervene when necessary.
— Provide understandable information to people affected by the system's decisions.
Translated into design language: you need interfaces that explain what the AI is deciding and why. You need control mechanisms that are real, not decorative. You need human referral flows that work well when the system fails or when the user doesn't understand the decision.
This is not resolved by the compliance team. It is resolved by the design team, with judgment and time.
Why this is a competitive advantage, not a burden
Organizations that start designing their AI systems with these criteria now have an advantage over those that leave it until the regulator comes knocking.
The reason is simple: designing transparency later is much more expensive than designing it from the beginning. An AI system built without thinking about explainability does not become explainable by adding warning text at the end of the flow. It requires rethinking the architecture of the experience — how information is presented, when it appears, in what format, with what language.
Organizations that do it well will also have a trust advantage with their users. And in sectors such as banking or insurance, where trust is the basis of the relationship with the customer, this translates directly into adoption, loyalty and business.
What we are learning at EGGS
In the AI projects we have been working on in recent years — in banking, insurance, public administration — we have seen that the most difficult problem is not technical. It's experience design.
How do you explain to a customer that an automated system has reviewed their credit application and reached a conclusion, without that explanation sounding like a corporate excuse? How do you design the moment when the user needs to talk to a person, without making it seem like the system has crashed? How do you build trust in a conversational assistant in a context — insurance, banking, healthcare — where the user has a lot at stake?
These questions do not have a generic answer. They have a response in the specific context of each organization, with research with real users and with a design process that treats transparency as an experience objective, not as a legal requirement to comply with with minimal effort.
It's exactly the type of problem for which iX exists — our methodology and knowledge platform for designing intelligent experiences in complex sectors.
The AI Act has not changed what we do. It has given us better arguments to explain why it matters to do it well.
