Zurich Global CX Head On Why AI's Insurance Opportunity Lives Between Customer Interactions

Most people hear from their insurer only a handful of times across a decade, usually at renewal or after something has gone wrong, and those few conversations end up carrying nearly the entire weight of the relationship. An industry with so little contact and so much riding on each exchange makes an unusual test case for AI, because the most visible place to deploy it, directly in front of the customer, is also where a misstep costs the most. For insurers, the better place to aim AI may be the years between those calls, when there's still time to prepare the agent, clean up the claims record, and make sure the next conversation starts with the customer's history already in hand.

Tom Voirol, Global Head of CX Strategy and Design at Zurich Insurance, was previously Chief Experience Design Officer at Ogilvy and spent a decade leading user experience globally at Adelphi Digital Consulting Group. At Zurich, he’s focused on what happens behind the customer conversation. The company is exploring AI across its operations, but Voirol is wary of putting it directly in front of customers, where automated underwriting can create costly errors and a failed handoff can damage trust. "I don't know whether people forgive companies any more when the AI didn't inform the human," he says. "It's about as bad as when a human didn't inform another human."

Personalizing the experience

Voirol describes insurance as a "sea of sameness," where carriers offer broadly similar products and customers rarely have reason to notice the difference. "So what makes you stand out?" he asks. "We believe that that is the brand delivered through the customer experience." Efficiency alone won't get an insurer there, and he worries the whole financial services sector has been tuned for it. "Frictionlessness is great, right? But if everything's frictionless, that doesn't make you you. It makes you good, but it doesn't make you you."

Personalization would be the obvious way to stand out, but insurers run into two obstacles. The first is data. A bank sees every card transaction and account movement, often spotting an upcoming home purchase before the customer mentions it, while an insurer can go years without hearing from a policyholder. "Insurance companies, we never hear from each other," Voirol says. The second is the product itself. Insurance works by pooling risk, so that "those who don't need a claim pay for those who do." The more an insurer tailors pricing and coverage to each individual, the less pooling is left, and the less the product works like insurance.

Zurich is applying personalization to the interaction instead. An internal effort to build empathy into every channel starts with recognizing how differently people absorb information, since some want every detail and suspect something is being hidden without it, while others want the top line and nothing more. "Can we actually in the moment serve you appropriately so you don't yawn through a five-minute call where everything gets explained to you, or you don't feel like we're keeping things from you?" AI can read those signals and guide the employee handling the conversation, and Voirol sees it as IP Zurich needs to own.

Knowing when a conversation matters

Most of the industry's AI talk assumes customers buy insurance directly, Voirol says, even though agents and brokers still handle most sales. "Most insurance gets sold rather than bought, and it mostly gets sold through humans," he says, adding that "the human intermediation channel has been pronounced dead over and over and over again." The vendor pitches, and LinkedIn posts, that reach him all seem to skip past that. "It always smells like direct sales," he says, "and direct sales are so small that they're almost negligible."

The agent or broker is where AI has the most to work with, starting with timing. When an insurer only hears from someone every few years, knowing when to call matters as much as what to say. Voirol wants AI to flag the moments that change what a customer needs, such as the arrival of a child, so an agent can walk in ready to talk about how they plan to pay for university rather than opening a brochure. Contact centers everywhere are shifting toward proactive service, and in insurance it would give a relationship built on bad news a reason to start somewhere better. Voirol doesn't rule out AI handling those conversations eventually. "Maybe AI can have those at some point, but they're deeply personal," he says, which leaves the nearer opportunity to the people already having them.

The infrastructure customers never see

A customer has no idea whether two internal systems share compatible data fields. They only notice when they have to repeat their story every time a conversation moves between teams or channels, and Voirol argues AI can make that worse. "If you apply AI to shoddy processes, you get shoddy AI," he says. "You have to explain yourself four times already. Let's slap AI on top. You're gonna have to explain yourself five times." Without fixing what sits underneath, "you just do the same crap more efficiently," since automation tends to remove the workarounds staff had built to keep a flawed process running. What's left is the hardest work landing on people anyway, only later in the conversation and with a more frustrated customer.

Voirol calls it "development debt that's accumulated over the years, sometimes decades." Until that gets resolved, whatever goes on top inherits the same problems. At Zurich, that work varies by market, since products, distribution models, and even the meaning of "agent" differ from country to country. "We can equip and convince and guide them to the degree that we can," he says of local teams. "But it's very much locally executed."

He also sees AI playing a role in the repair itself. Traditional integrations depend on precise field-to-field mappings, while models, he argues, can sometimes recognize that two slightly mismatched sets of metadata refer to the same thing. "You can actually apply human-like judgment to the connecting of systems," Voirol says, which he believes "can and does and will accelerate the build-out of improved infrastructure."

Measuring resolution the way customers live it

Zurich runs dozens of contact centers, and no two markets measure them quite the same way. Time on call still shows up as a KPI in many of them, and Voirol questions what it captures. A customer who phoned three times for a total of four minutes scores well and leaves frustrated, while one nine-minute call that settles everything may leave them happy.

Voirol would rather measure how long a problem felt unresolved, whether the company updated the customer without being asked, and whether the policyholder understands the coverage before a claim arrives. Each of those measures resolution as the customer defines it rather than as the company logs it. Every proactive update is also a call the customer doesn't have to make, leaving them "ever so slightly" more likely to renew.

AI changes what a CX team can learn from all of it. "I care slightly more about the quality than I do about the quantity," Voirol says of NPS, and open-ended feedback from hundreds of thousands of respondents has become something teams can analyze. Audio goes further, letting teams study the conversation itself instead of asking a customer to summarize it two days later. "You can actually hear their frustration and then how it switches into delight in the call or vice versa," he says.

Compliance review asks whether an agent followed the rules, but Voirol wants something harder to score. Zurich's agents typically aren't working from scripts, so the company can examine whether an interaction reflects the brand values, then spot the patterns across thousands of calls. That review, he says, "wasn't possible before in any meaningful way."

The fundamentals of insurance have barely moved in a century and a half. "You pool money, you throw a little money every month into something, something bad happens, we've got you back," Voirol says. What he sees shifting is everything built around that promise, including Zurich's partnership with Empathy, which gives families a person and an app to handle the practical aftermath of a death, along with a Life Vault where policyholders can store passwords and instructions ahead of time. An insurer that rarely hears from its customers can still be useful to them.

The views and opinions expressed are those of Tom Voirol and do not represent the official policy or position of any organization.