Voice Agents in Banking and Insurance: What They Really Deliver in 2026

Introduction

Voice agents are AI-driven assistants that can understand, handle and resolve customer calls in banking and insurance without human intervention – unlike traditional phone systems, they work in natural language and draw on live contract and customer data. They matter in 2026 because banks and insurers are cutting branch and broker capacity at precisely the moment customers still reach for the phone first. Voice agents help close that gap across four groups: end customers, advisers and agencies, brokers and sales partners, and the people who manage those partner relationships.

Why phone accessibility has become a board-level priority in financial services

Banks and insurers have spent years steering customers towards apps and self-service portals. Yet the moment it’s about money, a contract or a claim, people still pick up the phone – and that’s exactly where staff are thinnest on the ground: in branches, in agencies, in call centres. This article looks at why accessibility has become the real bottleneck in customer centricity, what the Lucerne University of Applied Sciences CX-TOM study reveals about it, and what Agentic AI Voice can realistically do for customers, advisers, brokers and their account managers alike.

The paradox: everyone digitises, everyone still rings up

Two figures sum up the situation. 61% of policyholders would rather call their adviser than use any other channel – that’s according to Continentale’s 2025 study of 1,200 respondents. Even at a Swiss neobank built almost entirely around chatbots, around 45% of enquiries still come in by phone, matched only by the bot itself, with the rest handled in writing. Lucerne University of Applied Sciences confirmed the pattern in spring 2026, looking at 61 Swiss banks: just two offer round-the-clock phone support, three have actually cut their phone hours since 2024, and only two have extended them. When something goes wrong, people want to speak to a human being – true in Zurich as much as in Cologne.

On the supply side, capacity is shrinking. The Bundesbank recorded 16,799 domestic bank branches at the end of 2025 – a 6% drop on the year before, following an 8.4% fall in 2024, and nearly half the number of a decade ago. In insurance, the number of registered intermediaries fell by 2,970 to 178,791 over the course of 2025. Customer numbers haven’t shrunk in either sector. What’s shrunk is how many staff are available per customer.

There’s also a pattern every service centre manager will recognise: customers expect their issue resolved there and then. No ticket number, no promised callback, and certainly no phone system that eventually puts them through to someone who’s stuck in a meeting. They want it sorted before they hang up.

How banks assess their own customer centricity

In 2026, Lucerne University of Applied Sciences examined how 43 banks across Germany, Austria and Switzerland structure their approach to customer centricity, with msg as study partner. The findings are candid – and uncomfortable reading.

86% of institutions cite changing customer expectations as the single biggest pressure on their target operating model, ahead of competitive pressure and regulation, both at 65%. Asked what happens if they fail to adapt their operating model over the next three years, 84% expect to become less attractive to new customers, 67% expect to lose competitive ground, 63% expect costs to rise, and 58% expect to lose existing customers.

Yet 53.5% of the banks surveyed sit in the bottom two of five maturity tiers, and only 9.3% run a scaled, AI-enabled CX enterprise operating model. The feedback loop breaks down at the crucial point: 74% analyse free-text customer feedback, 28% tell customers what came of it, and just 12% track whether anyone internally actually took ownership. Exactly one of the 43 institutions has a Chief Customer Officer.

Customer centricity is clearly on the agenda – execution is where it falls down. And the most basic test of execution, the one customers notice first, is whether anyone picks up the phone at all.

Beyond the hold queue: what voice agents can actually do

Everyone’s dealt with a phone bot: asking for a policy number, mishearing „yes“ as „no“, and eventually putting you through to a person anyway. Agentic AI Voice works differently. It understands what’s being asked in natural language, pulls up live contract and customer data, and either resolves the matter there and then or hands over to a colleague with full context attached. It can take calls, and just as easily make them – booking appointments, sending reminders, chasing things up.

Two things decide whether customers accept it. First, conversational quality: speech processing with no noticeable lag, so the caller doesn’t feel they’re talking to a machine. Second, transparency – EU AI legislation requires people to be told when they’re speaking with an AI, and that openness is part of what builds trust. Bitkom’s figures from May 2026 show a market split down the middle: 56% of Germans see AI in finance as an opportunity, 40% as a risk. 27% would be happy to hand most financial decisions to an AI, while 49% reject the idea of AI in financial matters altogether. Acceptance, in practice, comes down to whether the agent actually solves the problem.

Insurance: where voice agents make the biggest difference

The case for voice agents in insurance sales breaks down into four roles, each with an inbound and an outbound side.

End customer

• Inbound: claims, checking contract status, updating an address or bank details, querying a premium, requesting a certificate – exactly the calls clogging up hold queues today, and exactly what an agent with access to the policy system can resolve in one call, at 7am or 9pm alike.
• Outbound: proactive contact when a contract changes, a claim is approved, or an annual review is due. Continentale’s study shows the gap clearly: 58% of customers with an adviser want a regular annual review, but only 42% of advisers actually offer one.

Agency

This is where the biggest opportunity lies, because the agency-customer relationship, in both directions, sits entirely within the insurer’s control.

• Inbound: customers can reach their agency even while the adviser is in a meeting – 44% of working people want to reach their adviser before 9am, 28% after 6pm, yet only half of advisers are actually available before 9am.
• Outbound: scheduling within the existing book of business, reminders about expiring policies, targeted cross- and up-selling. And afterwards, a debrief the adviser records by phone on the way to their next appointment, which lands in the CRM as a complete record.

Broker

For the insurer, only the relationship with the broker matters, the broker’s relationship with their own customer runs through separate broker-management software, usually supplied by a network, and sits outside the insurer’s reach.

• Inbound: the broker calls about commission, application status, product details or policy status – and every call that doesn’t get answered is an argument for switching to a provider that does pick up.
• Outbound: the insurer calls about new product launches, webinars and activation campaigns, or because the broker’s submission pattern has changed.

Broker account manager

Managing anywhere from dozens to hundreds of relationships, with no time to do much beyond following instinct.

• Inbound: quick answers about a specific broker, on the road to a meeting, on commission statements, open cases, recent submissions or complaints.
• Outbound: the system flags something a person would likely miss, a broker whose submissions have been falling for three months, one who’s still productive but hasn’t been contacted in half a year, or one with an unresolved complaint. Submission behaviour is the earliest warning sign that a broker relationship is going cold, long before business actually moves elsewhere. Tracking it systematically, and acting on it, protects commission income that would otherwise only show up as a loss at year-end.

How the same logic plays out in banking

The same structure applies almost exactly to banks, building societies and fund managers – only the job titles change.

Customer

• Inbound: blocking a card, querying a transfer, booking an appointment, checking on a mortgage application, checking a building-society allocation.
• Outbound: a fixed-rate mortgage deal ending in twelve months, a savings plan that no longer matches the customer’s income. And from 1 January 2027, a genuinely new trigger that few service centres are staffed to handle: Germany’s new retirement savings account replaces the old Riester pension. Existing Riester contracts keep their protections, but switching to the new scheme is an option. At the end of 2025, the Federal Ministry of Labour and Social Affairs counted 14.66 million Riester contracts – many of those customers will be asking the same question in spring 2027: stay put, or switch? Call them first, and you control the conversation. Wait, and the hold queue does it for you.

Adviser

In branch, in private banking, as a mobile wealth adviser, or in a building society’s field sales team – the adviser here plays the same role as the agency does in insurance.

• Inbound: customers try to reach them mid-meeting, and an agent picks up the request, resolves the simple ones, or books a fully-briefed callback straight into the diary.
• Outbound: scheduling within the existing customer base, prepping customers ahead of a meeting, chasing missing paperwork. Afterwards, the debrief: the adviser talks through the outcome by phone, and the agent turns it into suitability documentation for the CRM. The compliance record still gets done – the evening spent at a desk doing it doesn’t.

Sales partner

The broker’s equivalent in banking is a more varied, and growing, group. As of 1 July 2026, Germany had 41,388 registered investment advisers and 58,014 registered mortgage brokers, the latter at an all-time high. Around 45% of new private mortgage lending in 2024 went through the two big broker platforms, up from just 19% in 2014. Add to that multi-line distribution networks, the local partners of building societies and fund managers – savings banks and cooperative banks – and independent private banking distributors. The same principle applies as with insurance: the relationship with the intermediary can be managed directly; their relationship with the end customer can’t.

• Inbound: a mortgage application is stuck, paperwork is missing, a rate needs clarifying, or a commission statement looks wrong. Outbound: rates change, a new product launches, or a partner hasn’t submitted any business in months.

Sales partner account manager

Key account manager in third-party distribution, regional lead for intermediaries, network coordinator at a building society, the exact same problem as the broker account manager: too many relationships, too little time, and decisions made on instinct rather than data.

• Inbound: quick briefing on a specific adviser or partner firm before a visit.
• Outbound: the system flags falling submissions, a dropping approval rate, or a partner suddenly doing noticeably less business. The early-warning signal that works so well in insurance broking works just as well here – it’s just that almost nobody in banking has thought to build it yet.

How to measure whether voice agents are working

The business case for voice agents isn’t just the minutes saved on calls. The metrics that matter:

• General service: accessibility and service levels throughout the day, first-contact resolution rate, cost per contact, post-call satisfaction.
• Outbound: appointment conversion rate, reach rate within the existing customer base, proportion of customers receiving an annual review.
• Partner business: reactivation rate for dormant brokers, submission rate per partner, share of wallet, how quickly account managers act on an early-warning signal.

One condition underpins all of these: a voice agent is only as good as the data behind it. Without a full, 360-degree view of the customer and partner relationship, it will end up answering questions it can’t actually answer, or calling the wrong people entirely. Voice is one part of customer centricity – the foundation is a target operating model where marketing, sales and service all draw on the same customer history.

Conclusion

Voice agents don’t fix a digitisation problem – they fix an accessibility problem. They close the gap that opens up when customers keep calling while branch and broker capacity keeps shrinking. The biggest opportunity sits where the customer relationship is entirely within the company’s own control – the agency and the adviser – and the same logic pays off again in partner management, where voice agents surface early-warning signals that simply don’t get picked up today for lack of time. Either way, the precondition is the same: a proper 360-degree data foundation. Without it, the agent ends up answering questions it was never in a position to answer.