Spanish banks have a retention problem, and the numbers are only getting worse. According to a 2026 Nickel study, in fact, 82% of Spaniards are seriously considering switching banks this year.
This would be cause for alarm across any industry, but for Spanish banks, the driver behind the exodus are the bigger concern. Among the top reasons to make the change, 41.4% of respondents cited poor customer service – a jump of more than 17 percentage points from 2025 – and a further 31.1% pointed to the difficulty of getting help without a prior appointment or simply lacking in-person attention.
Unsurprisingly, banks are turning to technology in a bid to speed up and smooth out customer servicing. The default line of action has been to implement AI, with Spanish banks cutting more than 8,000 jobs.
Why another chatbot won’t fix this
Luis Guillermo Pardo, CEO of AI infrastructure and cognitive agents company Kognia – which works with leading banks throughout Spain and Latin America – these efforts and adoption won’t pay off if banks don’t address the infrastructural woes that continue to plague them.
“Technological fragmentation and data isolation are the real culprits behind friction in the customer experience, not simply a lack of personnel,” Pardo noted.
In other words, another chatbot won’t fix this impending crisis; if a bank’s answer to slow service is a faster-talking chatbot plugged into a fragmented backend, it will run into the same wall consumers already say they’re tired of. It is clear customers don’t want a system that can answer simple questions but fails to actually resolve anything, with no clear path to a human when the issue needs to be escalated.
In Spanish banking specifically, that fragmentation shows up time and again: rigid legacy core systems mired in outdated and poorly formatted data that don’t communicate with each other, scattered customer information, manually-dense process around claims handling and fraud detection, demand spikes that degrade response times, and convoluted resolution paths.
The result, Pardo revealed, is that bank staff end up spending up to 40% of their time digging for information – previous time wasted, which could otherwise be used to help the customer resolve their query quickly and smoothly.
The data is there; accessing it is the problem
And the raw data is already there. Banks have the information to resolve the case, but accessing it and acting on it accordingly are the biggest hurdles to ensuring customers’ concerns are promptly resolved.
When a customer queries an issue or lodges a complaint with a chatbot or customer care representative, for example, they usually also share crucial information including the nature of the problem, when it occurred, their location, and their account or file number.
But a disconnected system means that information gets lost to the void – or worse, mishandled, becoming the source of hallucinations and erroneous recommendations. Even when a human representative steps in to address a complaint, they are not given the full picture, and have to start from scratch to find a resolution in as little time as possible.
The result is a cycle of escalation and reset, with most of the burden falling on the customer to keep pursuing their case while the bank’s customer care team scrambles to retrieve scattered details.
Cognitive AI as a “digital worker”
Pardo drew a sharp distinction here between cognitive AI, systems that are designed to think like a human, and the traditional technology. Kognia’s product, the Alan Agentic System, which functions as a dynamic digital worker unifying real-time data access across a bank’s systems, automating complex back-office tasks through secure integrations, and orchestrating a case from start to finish rather than handing the customer a scripted partial answer and leaving the rest to chance.
The aim, as per the CEO, is to free up human staff for the conversations that genuinely need a person.
Compliance and trust by design
It’s worth noting that data protection and fraud security weren’t a minor concern in the Nickel survey; 62.8% of respondents cited stronger protection against fraud and better safeguarding of personal data as a reason they’d consider switching banks.
Any AI system handling financial data in Spain should thus address these anxieties explicitly – not as an afterthought, particularly as the EU’s data and AI laws continue to grow stricter.
This regulatory compliance, Pardo said, is an architectural starting point rather than something to be retrofitted:
“We build the Alan Agentic System around a privacy-bydesign-and-by-default approach, which in practice means native alignment between the EU’s GDPR, rigorous data classification and end-to-end traceability, and automated anonymization of sensitive financial data.”
“Our system is also audit-ready and explainable, ensuring human-in-the-loop requirements are met for any decision made with Alan,” Pardo added.
AI tools shouldn’t be designed to simply produce a correct answer. Rather, they should be able to show why they reached that answer and what data informed it.
“In the financial sector, it’s not enough for an AI to give a correct answer. We must audit why and with what data it made that decision,” Pardo emphasized, positioning this auditability as a way to avoid the black box problem pervasive across AI models.
Overcoming resistance on both sides
None of this, however, means adoption is automatic. The executive acknowledged that some resistance to AI in banking is simply natural in a traditionally-conservative sector, clarifying that the resistance looks different depending on who you ask.
Banks themselves tend to hesitate over compliance risk, the fear of AI hallucinations, cybersecurity exposure, and the practical difficulty of integrating new systems with legacy infrastructure without disrupting day-to-day operations.
Consumers, on the other hand, worry about losing personalization, have lingering doubts about the privacy of their financial data, and carry real frustration from years of dealing with rigid first-generation chatbots.
Three principles: quick wins, transparency, escalation
Kognia’s answer, based on its work in both Spain and Latin American markets like Colombia, rests on three principles: quick wins, radical transparency, and friction-free escalation.
Rather than attempting to transform an entire bank overnight, the company starts with low-risk, high-impact use cases – such as internal support tools for advisors or handling frequent customer queries – where improvements in satisfaction and resolution times are easy to measure.
Furthermore, customers are always told when they’re interacting with the Alan Agentic System, with a clear sense of what the system can and can’t do.
And finally, if a query is highly sensitive or emotionally charged, the AI hands it off immediately to a human advisor, carrying the full case context with it so the customer never has to repeat themselves.
“When banking committees see tangible evidence of governance and security, and users perceive 24/7 availability with real resolution instead of generic and unhelpful answers, resistance quickly turns into adoption,” Pardo explained.
Put the two halves of this story together, and a clearer picture emerges: Spanish banks aren’t losing customers because they have too little – or too much – technology.
Churn is mounting because systems aren’t keeping up with what modern banks actually need to do: see the whole customer at once, resolve a case without bouncing it between departments or resetting the query, and looping in a human the moment a conversation calls for one.
Featured image: Chris Boland via Unsplash+
Disclosure: This article mentions clients of an Espacio portfolio company.
