Marco Spagnuolo (pictured), did not plan a career in insurance technology, he entered the field by chance, stayed for the substance, and now leads Swiss Re’s customer-facing conversation intelligence proposition for insurers using call centres.
After studying economics, he joined Allianz as an underwriting assistant, later moving into brokerage at Aon and earning a Master’s in Insurance and Risk Management. He has been with Swiss Re for over 15 years, holding roles across life and health, including technical accounting, portfolio management, and in-force solutions.
The neatness of Spagnuolo’s current title rather disguises the accidental way he entered the industry. “It happened by chance,” he said, recalling the call that first brought him into insurance. Later, he became interested in reinsurance in particular because it allowed him, as he put it, to “conclude the value chain” after seeing the primary insurer and broker sides of the market.
That long progression matters, because his present role did not emerge from a pure technology track. It was shaped instead by years inside insurance operations and customer-experience work. At Swiss Re, his path through in-force solutions and behavioural-economics-based training for call-centre environments led him to a practical observation: customer-facing staff often lacked tools that could genuinely support their work. That, in turn, led to the idea behind conversation intelligence.
Spagnuolo now oversees a platform called CXaaS, (customer experience as a service), which is aimed not at Swiss Re’s own internal transformation but at insurers and other insurance players operating call centres.
The platform was developed with a California startup, combining AI capability with sector-specific insurance expertise. Spagnuolo’s argument is that generic conversation analytics are not enough for the market he serves. Insurance calls are shaped by very specific moments and needs: buying cover, preventing lapse, handling policy changes, or notifying claims. Swiss Re’s value, in his telling, lies in embedding those realities into the system.
At the heart of the proposition is the conversion of raw conversation into usable information. Spagnuolo described conversation intelligence as a bundle of AI techs including natural language processing and understanding, machine learning, sentiment detection and speech recognition, working together to transcribe spoken interaction and make it analysable.
His central claim was simple and striking: “That’s where the magic happens, because unstructured data becomes structured,” he said. Once a call is anonymized and transcribed, firms can analyse not only what a customer said, but how it was said. In a sector still wrestling with fragmented data estates and manual processes, that ability to structure speech is not merely a technical feat; it is a strategic one.
The system works off recorded conversations through telephony systems. Those audio files are fed into the platform, usually by API connection, and transcription can happen simultaneously. Depending on the client’s preference, the platform can support real-time functionality or retrospective review.
Spagnuolo’s strongest point was not that AI sounds impressive, but that many insurers still lack a precise understanding of why policyholders are calling in the first place. In many operations, he said, agents still use manual spreadsheets with pre-populated fields to summarise the reason for contact, a method that is both approximate and error-prone.
By contrast, machine categorisation can give firms a near-live view of customer demand and friction points. Spagnuolo said insurers can identify shifts in call topics over time and can detect digital failures quickly. For example, sudden login problems on a website or app that trigger a spike in inbound calls. Without such tools, he suggested, it might take weeks to discover the issue, “if ever”.
There is also a more prosaic prize: lower-cost servicing. Spagnuolo pointed to routine administrative requests such as changing addresses or updating beneficiaries, tasks that can now be shifted into digital self-service channels. In his view, better analysis of call-centre demand gives insurers a clearer basis on which to shape those digital agendas.
Spagnuolo grouped the business case into three parts: productivity, customer experience and compliance. Monitoring calls at scale can surface where agents are frequently placing customers on hold, struggling with claims conversations, or taking materially longer than peers, all of which can suggest coaching or training gaps.
On customer experience, the attraction lies in quantifying what previously sat in the realm of instinct. Managers may know which agent performs well with customers; conversation intelligence promises to show why. That matters in businesses where the difference between a retained policyholder and a lost one may come down to the quality of a single phone call.
The compliance case is equally plain. Instead of relying on random sampling, the tool can scan every call, whether that is a thousand or a million, and surface complaints, misconduct or other non-compliant behaviour. For regulated insurance operations, that sort of scale is not trivial.
Spagnuolo said transcription accuracy is “above 90%”, while acknowledging that results vary with accent, language, jargon and recording quality. In noisy conditions, he accepted, the audio quality and hence the transcript will be impacted.
The system follows privacy by design principles, in accordance with all relevant laws and regulations.
If the technological side has moved swiftly, the organisational side has been slower. Spagnuolo said one of his clearest lessons from the work is that digital initiatives are not won on systems design alone. “It’s 50% tech and 50% people,” he said.
He described the “human element” as the harder part of the job. Early on, he said, he was too focused on the technical dimensions of the platform and underestimated how much depended on adoption, patience and change management, particularly inside established organisations, where AI projects move more slowly than the underlying technology.
That issue extends to how the platform is framed for employees. A tool of this kind can easily be seen as surveillance if introduced poorly. Spagnuolo said the messaging matters: used properly, it should be positioned not as a control device, but as support for call-centre staff trying to do a difficult job better.
Spagnuolo is not in the camp that thinks insurance decisions will soon be handed wholesale to machines. In his personal view, “AI will not replace humans,” and he does not see it becoming the final arbiter of decisions in insurance. Rather, he expects it to support decision-making by contextualising large volumes of information and producing credible suggestions at speed.
He is, however, notably open to the idea that AI may become highly competent in emotional interaction. The current platform can already assess whether empathy has been displayed in a call, not through tone analysis, which he said is highly sensitive, but through the words used. Expressions such as “I’m sorry to hear that” can be identified and measured.
His most arresting observation may be that human beings do not automatically have a monopoly on empathy. “I’m not convinced that humans necessarily would always be more empathetic than what a virtual agent could be,” he said. In an industry that still trades heavily on trust, reassurance and judgement, that line will make some uncomfortable. But it is a sign of how far the conversation has moved.
For all that, Spagnuolo does not present as a technologist sealed inside the subject. Outside work, he reads widely, follows geopolitics, is learning German. He also swims frequently and, as a father of two teenagers, is mindful that family time is limited.
There was a dry note of realism in the way he described home life. He knows there is only “a time window” before children disappear into lives of their own. It was an unexpectedly human coda to a conversation about structured data, AI agents and digital adoption and perhaps a useful reminder that, in insurance as elsewhere, technology is still judged by what it does for people.
