Ireland has transposed the EU distance marketing directive through the European Union (Distance Contracts for Financial Services) Regulations 2026, giving consumers the right to request human intervention rather than having to rely solely on an automated system when they buy a financial product or service online or over the phone.

Directors Club News asked Liam Dunne, Co-founder and CEO at Klearcom, what happens when the escalation path itself is broken…

New rules in Ireland give consumers the right to speak to a human when they are dealing with an automated system. From where you sit, what does that change?

On paper it looks like a small change, but in practice it brings the escalation path into audit territory. Think about what used to happen. Somebody got stuck with a bot, gave up, and then hung up. That turned into a report as an abandoned contact, everyone treated it as a service issue, somebody’s CSAT dipped that month. Now that same hang up starts to look like a customer being denied something they’re entitled to. It’s a different conversation, and gives the caller much more leverage than before.

I’d be careful not to overstate the scope, because there’s potential to do that. The rules go into a new part of Ireland’s Consumer Rights Act 2022 and they cover contracts closed online or over the phone. A&L Goodbody’s read is that the entitlement kicks in where someone is dealing with a firm through an online interface before they sign, and in justified circumstances after. Nobody has legislated an uptime target for your phone lines, let’s be straight about that.

The bit that might catch people out. The human agent that is handed that customer to is usually sitting on a phone line, so how that’s achieved consistently is the question. Keeping it working is something you must do every day, on every number, in every market you sell into. That’s a very different job.

You have said the handoff is where things break. What does that look like in practice?

A few things, on repeat. First, the transfer can just fail. The bot say that it’s putting you through and then the call drops, or you get silence, or a busy tone. Most of the time that’s a carrier or a trunk, which has nothing to do with the bot.

The second break would be the context can reset. You’ve spent four minutes giving an assistant your account number, date of birth, why you rang, and so on. The human agent picks up and asks you for all of it again. On paper that transfer worked fine. Ask the customer how it went!

The third one is the authentication loop. The assistant verifies you, hands you off to the human agent, and the agent’s desktop doesn’t recognize the session. So you do it again, or even a third time.

The last one is the sneaky one. The path quietly stops existing. Someone edits an IVR menu, retires an option, reroutes a number, and the branch that used to reach a person starts dropping people back to the main menu instead. Nobody spots it, because nobody tested that branch after the change went live.

Why do failures like that go unnoticed inside large organizations?

You’re monitoring the things you own. Your platform will happily tell you a transfer was initiated, it’ll give you a handle time, abandonment on the queue, all of that. What it can’t see is the stretch of the journey sitting outside your voice estate. Things like the carrier, the number translation, the route from a mobile network in one country into a contact center in another.

You’ll be getting the situation where your dashboards are green, and meanwhile a customer in Spain rings your toll free number, hears a recorded message in the wrong language, and has no way through to anybody. That call never touches your platform, so the platform has nothing to report.

Does the move to agentic AI make this harder?

It certainly raises the stakes. With a scripted IVR you’ve got a fixed set of paths and you can write every one up on a whiteboard. An agentic system decides at runtime, so the escalation might fire in a situation you never planned for.

Agentic AI systems will be good enough now that customers interact with them longer before asking for a person. So by the time somebody does ask, they’re usually deep into a query with the Agentic AI and fairly invested in finishing it. A broken handoff at that point costs you a lot more than it used to.

If this is becoming a compliance matter, what counts as evidence?

The test can be straightforward. Show me, for a given date, that a customer ringing from a particular country got to a human in a reasonable time. Few organisations can show this. The evidence means records over time, such as this number over time, this result, this timestamp.

What should contact centre voice leaders actually do about it?

Start by listing every number a customer can ring you on. Sounds trivial – it’s not! It almost never is. Then walk each one through to a human, from outside your own network, on a real call, and from the country the customer is actually calling from. A call off a Dublin desk phone and one off a Portuguese mobile take different routes and they break in different ways.

Run all this via a tight schedule rather than as a once off. Carriers reroute things and numbers get ported. It’s not glamorous but absolutely necessary. It’ll be the difference between a right that exists on paper and one that works when somebody picks up the phone.

Liam Dunne is Co-founder and CEO at Klearcom.