Everyone expects AI to improve productivity and efficiency. AI helps organizations accelerate processes, lower operating costs, and reduce repetitive work for employees. Its value, however, depends on whether it solves the right problem. Automating a process that already creates friction simply scales existing inefficiencies.
Innovation is a new way of looking at an old problem
In a recent project for one of Slovakia's leading banks, we redesigned the software interface used across its branch network. During user research, we uncovered numerous UX issues. One finding stood out. Relationship managers had to complete a customer profile containing 116 fields during every consultation.
The bank introduced this process to improve the quality and accuracy of customer data. In practice, the opposite happened. Employees knew that no customer would spend an entire meeting answering more than one hundred questions. To move through the form, they often filled mandatory fields with dots or placeholder values, to save time. The result was predictable. Employees became frustrated. Customers had a poor experience. The bank collected data it could not rely on.
The system had been designed around technology rather than the people using it. When employees work in an inefficient environment for long enough, they naturally develop shortcuts and workarounds to get their job done.
That is why every automation initiative should begin with understanding user journeys, identifying friction, and improving the underlying process. Human journey first. Feature second.
When the problem wasn't the call centre but the invoice
Bratislava Water Company faced a high volume of customer calls. The initial assumption was that customer communication needed improvement. What remained unclear was where the problem actually originated. Was it the website, customer support, or the documents customers received?
Instead of optimizing the FAQ section or automating customer support, we conducted user research. It revealed that customers simply did not understand their invoices. We interviewed and tested the experience with real BVS customers. We also analysed call recordings and customer interactions. Based on these insights, we redesigned invoices and supporting customer documents. The impact was measurable. Invoice-related calls to the contact centre dropped by 21.6%, while first payment reminders decreased by 5.9%.
BVS could have deployed an AI chatbot to answer invoice-related questions and reduce pressure on the call centre. Without understanding the underlying issue, however, AI would only have automated responses to confusing communication instead of eliminating the root cause.
These improvements also created a stronger foundation for future automation. BVS can now optimize a process that already works, instead of scaling one that does not. Technology may become increasingly powerful, but its value depends on whether it solves a real customer or business problem.
Before introducing AI, identify the problem
1, Turn data into insights
Data never tells the whole story. It needs context. Demographic data tells you who your customers are. Behavioural and analytical data shows what they do. It reveals where customers abandon onboarding, which products they use, and how many calls reach your contact centre. Only research with customers showed why those behaviours occur. A good example comes from Slovak health insurer Dôvera. When collecting overdue insurance payments, the company achieved a recovery rate of only 19%. Every debtor received the same standard communication.
User research uncovered very different motivations and barriers across customer segments. Based on these findings, we tailored communication instead of automating a generic approach. More relevant reminders, supported by different messaging and arguments, increased the payment rate from 19% to 24% a relative improvement of more than 26%.
2, Validate assumptions with people
Insights remain assumptions until they are tested with real people, whether customers or employees. This requires professionally conducted qualitative research, often using ethnographic interviewing techniques. Experienced design partners or research agencies can help uncover the underlying drivers behind customer behaviour.
Many organizations facing a situation similar to BVS would have briefed an agency with a straightforward request: "Redesign our website FAQ because we receive too many calls." That solution would only treat the symptom. Without user research, the root cause would have remained hidden. Customer experience design helped reduce the overall number of calls while improving customer satisfaction.
3, Calculate the return on investment
Customer research often uncovers more opportunities than an organization can pursue at once. Not every problem justifies an expensive AI solution. Prioritize initiatives based on business impact and expected return on investment.
A strong product team or an independent design partner can help build an objective business case. If a customer problem occurs only occasionally, automation is unlikely to deliver positive returns. The investment may never pay back. When a problem affects a large share of customers or employees, AI becomes a powerful lever. The greatest value comes from following the right sequence. Understand people first. Design a better process. Then use technology to scale it.
Technology alone does not close the gap between customer expectations and reality. Organizations close that gap by listening, understanding, and responding to people's needs. AI amplifies good solutions. It accelerates them and extends their reach. But only when the foundations are already in place.
