European regulation is entering a more practical phase when it comes to consumer protection and automation. After the introduction of the EU AI Act in 2024, which set out high-level principles around transparency and fairness, individual countries are now beginning to translate those principles into enforceable national rules. The result is a patchwork of regulation, but with some clear themes emerging, particularly around access to human support, service quality outcomes, and accountability.

In the latest episode of Cavell Cloud Conversations, Patrick Watson and Finbarr Begley explore what this looks like in practice, focusing on developments in Italy and Spain, and what they signal for the future of voice AI, CCaaS, and customer service strategy across Europe.

From principles to outcomes

One of the challenges with regulating AI is that the technology moves faster than legislation. Rather than attempting to control how AI systems are built, regulators are increasingly focusing on outcomes. What experience is the customer entitled to, regardless of whether a human or a machine is involved?

This shift is particularly visible in consumer-facing industries where customers have limited choice, such as utilities. In these environments, poor service cannot be easily avoided by switching provider, so regulators are stepping in to set minimum standards.

Italy. Guardrails where choice is limited

Italy provides a useful example of a lighter-touch but targeted approach. Rather than introducing broad consumer AI legislation, regulators have focused on utilities, where customers are effectively captive.

Under guidance from ARERA, Italy’s utilities regulator, calls must be answered within three minutes. Crucially, any time spent interacting with an automated system counts towards that limit. Automation is still permitted, but only if it is effective. Bots that loop, repeat questions, or fail to route customers efficiently risk pushing providers out of compliance.

This has had tangible consequences. In one case discussed on the podcast, a utility provider was forced to switch off its voice AI system entirely. The bot worked well for FAQs and delivered useful information, but real customer conversations are rarely linear. Customers were realising partway through a long-mostly FAQ related call that they need human support, and that call was then being flagged as a customer service interaction and ruining the Time To Answer KPIs that the provider had to meet.

Italy has also mandated physical service locations in every town or city where utilities operate, ensuring access for non-digital customers. The message is clear. Digital-first is acceptable, but digital-only is not.

Spain. A more comprehensive model

Spain has gone much further. In December 2025, a new consumer protection law came into force covering large organisations and companies operating in protected sectors. Its focus is explicitly on service quality outcomes.

Key requirements include:

• 95 percent of inbound calls must be answered within three minutes

• Automated interactions must either resolve the issue or escalate to a human

• Customers have an explicit right to human support

• All customer service lines must be free

• No dark patterns that loop or deflect customers indefinitely

• Mandatory opt-in for outbound communications

Outbound calling is also tightly controlled. Companies must use different numbers for sales and support, making it clear to customers why they are being contacted. This also strengthens verification and reduces fraud, with telcos responsible for securing these numbers.

Perhaps most significantly, companies must maintain detailed customer service audit logs, available to the government on request. Annual independent audits are required to prove compliance with service KPIs. The burden of proof sits firmly with the provider.

Language support is another important dimension. Businesses must offer customer service in the relevant regional languages, reflecting Spain’s linguistic diversity.

What this means for voice AI

These regulations do not ban voice AI. Instead, they raise the bar. Automation is acceptable, even encouraged, when it works. When it does not, escalation must be fast and frictionless.

This creates strong incentives to invest in better AI rather than less. Systems need to accurately assess whether an issue is suitable for self-service and, when it is not, pass full context to a human agent. Re-authentication, repetition, and loss of context become even more costly in a regulated environment.

There is also a strategic implication. As automation reduces overall call volumes, each remaining human interaction carries more weight. A single poor experience can drive churn, particularly when customers rarely speak to their provider.

A signal to the rest of Europe

Spain’s approach is likely to be closely watched. European regulators have a history of allowing individual countries to act as test beds before broader adoption. Markets such as France and Italy could move in a similar direction if outcomes are seen to improve without stifling innovation.

For vendors and service providers, the message is not that automation is going away. It is that accountability is increasing. Customer experience outcomes, not cost savings alone, are becoming the benchmark.

As Finbarr notes in the discussion, consumers already expect better. Regulation may simply force consistency, ensuring that access to effective support is not a differentiator reserved for premium brands.

In that sense, this wave of legislation could act as a rising tide, lifting standards across the industry while pushing vendors and providers to build smarter, more integrated, and more human-aware systems.

Listen to the podcast with Patrick Watson and Finbarr Begley.