Insurance fraud is on the rise. The tumultuous nature of the last two years means that there are increased opportunities for fraudsters to try and exploit the disruption.

Insurance call centres have always been seen as an easy target for fraudsters because of the volume of calls they have to deal with daily, the speed at which operators have to resolve issues and now the fact that many insurance agents are working remotely. Agents are, of course, aware of the risks of potential fraud but must also be focused on ensuring customer satisfaction – meaning that sometimes fraudulent activity can escape their notice. But this is where Artificial Intelligence and speech recognition can help to keep both businesses and customers safe.

According to a 2021 survey, approximately 31% of insurers have already deployed the use of AI in their companies, while a further 23% said that they had plans to start using the technology in the next year. There are a variety of benefits to implementing AI and voice recognition systems into business practice, improving efficiency and eliminating the risk of human error.

Voice AI, such as Intelligent Voice’s LexiQal system, can detect patterns in the behaviour, speech and language that are frequently displayed in fraudulent calls, which means that companies can better develop an algorithm to filter out these phone calls without them ever reaching an agent in the first place. This makes processes altogether more efficient and is especially useful for large companies dealing with a very high volume of calls on a day to day basis.

There are four common characteristics in fraudulent calls, which AI systems can be programmed to detect. These characteristics are; negation, hedging, latency, and emotion. A fraudulent caller often avoids directly answering questions and has more frequent pauses in their speech as they try to avoid inconsistencies that could reveal their intentions. Fraudulent calls are likely to react emotionally to being confronted, or when they think that they are close to being detected. Understandably, when a call handler is dealing with large volumes of phone calls a day, it can be difficult for them to pick up on every little inflection or clue – as well as continue to provide topline customer service. AI technology fills this role, detecting these common tells as the call goes on.

A caller will rarely speak to the same agent twice and this plays perfectly into the hands of fraudsters who will look to exploit this and find a way to slip through the system, with inconsistencies in the fraudulent story remaining undetected as they are passed from person to person. But AI can be programmed to detect recurring patterns and with every phone call monitored it is easier to note these inconsistencies somewhere central, so that call handlers are able to cross-reference and instantly flag suspicious activity.

AI systems are always improving and evolving via machine learning programs, which means the more calls that are logged – the better the system will be a recognising potential fraud. All features that are commonly detected in fraudulent calls can be tracked and monitored. If any particular speech pattern or trait increases in frequency, the AI will learn of its increased importance through the machine learning process. This constantly evolving system and detailed history mean that potential fraud can be detected from the very first call.

Fraud prevention is the top priority for insurers – and while the incidence of fraud might be on the rise, AI is an effective way to combat this, saving businesses money, time, and effort and eliminating the risk of fraudulent activity.

Nigel Cannings, CTO and co-founder at Intelligent Voice