Voice analysis: Reading between the lines to accelerate claims
Clearspeed’s intelligent voice-based screening technique aims to build trust, accelerate claims processing and minimise fraud. Alex Martin, Clearspeed CEO, explains the value of voice analytics.
Insurers want swift and accurate claims processing. Speed improves customer experience because legitimate claims can be resolved quickly and efficiently, leading to happy customers and better retention, plus lower processing costs. However, speed must be accompanied by accuracy: The aim is straight-through processing that also reduces errors and fraud. In other words, a twofold win.
It is a goal that Clearspeed says it has achieved with an AI-enabled voice analytics service that accelerates low-risk transactions while flagging risks. Big Tech has made significant strides with voice analysis in a wide range of fields, such as authentication and security, and Clearspeed has applied it to insurance risks with the help of a military breakthrough born of necessity.
Technology with military origins
Alex Martin, Clearspeed’s co-founder and CEO, says that the insurance use case was inspired by his time in the armed forces and his experience of working with talented voice technology specialists who pioneered a solution to build trust in hostile regions.
“The challenge in combat zones, particularly in the special operations field, is how to build trust quickly,” he says. “Different languages and cultures lead to uncertainty and mixed messages, which undermine trust, and trust is imperative in military situations.”
Martin says that when the military researched ways to establish trust quickly, universal voice characteristics proved very revealing. This was particularly important when liaising with local nationals in conflict areas in Africa and the Middle East, as there are few other ways to vet associates and form trusted relationships.
„Instead of trying to catch people out, like a lie detector does, you’re accelerating claims for the vast majority of low-risk, honest people.”
“You might think a lie detector is all you need,” says Martin, “but lie detectors only provide binary answers. A response is either true or false, and lie detectors are suitable only for one-to-one situations. We needed to develop risk identification and risk stratification techniques that could cover more ground at speed, which is where voice analysis comes in.”
Covering ground swiftly
Instead of following the lie detector route, voice specialists adopted a metal-detector approach, where voice screening could quickly identify low-risk individuals while highlighting those who need further investigation. Clearspeed, launched in 2016 in San Francisco, adopted the military model and refined it for claims handling.
“Like a metal detector, voice analysis allows you to scan and interpret large amounts of data quickly and efficiently,” says Martin. “Above all, you can clear the way for claims if nothing merits closer inspection. You dig only when needed, stopping if there is something unusual that needs more investigation. Instead of trying to catch people out, like a lie detector does, you’re accelerating claims for the vast majority of low-risk, honest people.”
Martin emphasises that Clearspeed’s approach has nothing to do with voice recognition or biometrics. It involves four steps, beginning with an automated phone questionnaire. The speech is then converted into what Martin describes as an “unbiased and language-independent data model.” Next, the data is analysed to identify vocal characteristics associated with risk, then a risk score is assigned and the results are provided via an API or an app.
AI-based risk evaluation
“We capture characteristics that are always present in everyone’s speech,” says Martin. “When you provide inaccurate information, your speech is slightly impacted. We apply artificial intelligence to generate low, medium or high-risk evaluations based on the distinctive voice characteristics in our data model. There is no risk of AI bias because the AI works more like a heart monitor, responding objectively to a voice signal that is either present or not present.”
Martin says the data-model questions prompt ‘yes’ or ‘no’ answers and are geared to the type of insurance cover, such as auto, home and property, or gadget. In what is an automated process, every customer has the same experience and the objective is to remove friction and settle claims swiftly.
“Insurers need a simple way to distinguish low risk from high,” says Martin, “and it must start with claims triage at first notice of loss. If you have the right voice-based insights, you can avoid delaying honest claims and focus your expert follow-up on questionable ones. Our philosophy is to make things easier for good actors and more difficult for fraudsters.”
Martin says that many insurers are challenged to provide straight-through processing at first notice of loss. While techniques such as Natural Language Processing are sometimes used to better understand customers, they involve large amounts of data. Not so with voice analysis, which Martin says can quickly deliver highly accurate risk assessments from limited but precise information to help inform workflows. Moreover, using Clearspeed’s questionnaire foundation and AI-based analytics, voice technology could also be used at the quote stage to accelerate onboarding and help determine premiums.