Helping organizations engage people and uncover insight from data to shape the products, services and experiences they offer

Learn More

Contact Us


We'll be in touch soon!


Refer back to this favorites tab during today's session for access to your selections.
Refer back to this favorites tab during today's session for access to your selections.x CLOSE


How Insurers Can Overcome AI’s Steepest Obstacles (Part 2)


By addressing challenges — from data preparation to workforce readiness — carriers can join early adopters on the AI journey.

Second in a two-part series. 

As daunting as AI may seem, working examples abound. Forward-thinking insurers are already using artificial intelligence today:

  • Swiss Reinsurance Co. is working with IBM to develop a range of underwriting solutions to achieve accurate risk pricing.

  • enables customers to generate price quotes by texting a photo of their license plate. The company’s virtual insurance agent Evia then uses machine learning, natural language processing and an examination of public records to generate a quote. 

  • Buyonic Insurance Agency in Austin, Texas, greets its customers not with a human but with Siber, a robo-advisor that can rate, bind and issue policies on the spot while also answering and placing robo-calls.

  • Microsoft and GAFFEY Healthcare have partnered on a pilot to deploy machine learning in GAFFEY’s claims automation and processing engine at hospitals.

  • USAA recently added Nuance Communications’ virtual assistant Nina to its existing mobile customer service apps to enable speech recognition, text-to-speech and voice biometrics.

To join the list of early adopters, insurers must first equip themselves with the ability to gather information from various sources. They also need a highly integrated and digitized environment, and a workforce ready for innovation. Wherever you fall on the maturity spectrum (see Part 1), here’s a roadmap of specific AI challenges you’ll need to overcome:

Building the foundation.

For AI platforms to solve business problems, intelligent machines need access to lots of information to feed their continuous learning mechanism. This includes huge volumes of both domain- and customer-specific information covering all possible business scenarios. For example, if an AI solution is aimed at replacing the contact center, the bots need a wide variety of customer query and response data to inform their behavior. In collaboration with technology and consulting partners, insurers will need to build an ever-evolving, accurate and comprehensive knowledge repository of domain information, customer demographics and psychographics, as well as information on how these changes affect customer interactions with insurers. 

Dealing with the glitch risk.

Technologies such as speech recognition and machine learning require significant human oversight to ensure they resonate with interacting humans. Natural language processing (NLP) and voice recognition systems must be able to cope with diverse accents, background noise, distinctions between homophones (such as “buy” and “by”) and the speed of natural speech. Otherwise, AI failures can introduce unpredictable, bizarre and — in the case of Microsoft’s rogue chatbot — sometimes offensive errors. When that happens, AI systems require significant human investment in retraining and reconfiguring before resuming their work. Ensuring teams are equipped with the talent required to pilot and nurture such systems will be a big challenge, and making provisions for regular manual interventions will require time, money and energy.

Stakeholder readiness.

Insurers implementing AI solutions will need to manage the threat their workforce might feel by redesigning tasks, jobs, management practices and performance goals, particularly among advisory, operations and contact center teams. On the customer side, insurers will need to deal with trust issues, such as “Can I trust a machine in making long-term financial decisions?” and “Will bots be able to deal with my emotions during complaints and claims?”

Regulatory hurdles.

AI solutions require every interaction to be recorded for machine learning. Since most AI solutions are likely to reside on the cloud of a third-party technology provider, insurers will have to battle data privacy concerns. Keeping a close watch on regulatory changes will also be a challenge, considering the time lag for regulatory bodies to embrace technological innovations.

Technology refresh.

Implementing AI will require a good deal of supporting technology (see below). Insurers will need to evaluate the existing state of their technology landscape, integration among systems and applications, their own level of digitization and data availability. Building the necessary infrastructure and integration will require sizable amounts of time, resources and buy-in.

Figure 1

Race to the Future

The following guidelines can help insurers get a strong start on the AI journey:

  • Assess readiness: AI decision-makers must spend time socializing the concept and experience of AI solutions with their executive teams, peers and functional business leaders. They must also assess their internal technology landscape, the extent of historical data available, and the ways and means to scale and support AI applications.

  • Start small: A proof of concept model should be created that can safely be tested and adapted in a risk-free environment. Since AI machines excel at routine tasks and their algorithms often learn over time, insurers should focus their early efforts on the processes or assessments that are widely understood and add incremental value. Once that is done, insurers need to identify the right technology partner and AI solution to transform the use case from concept to reality.

  • Manage change: An effective and thoughtful HR strategy must be established because of AI’s ramifications for the workforce. Full communication and staff retraining will go a long way toward minimizing resistance and encouraging acceptance. The more decisions machines make, and the more data they analyze, the better equipped they are to undertake increasingly complex tasks and deliver more accurate and appropriate actions.

To learn more, please read our white paper How Insurers Can Harness Artificial Intelligence and our Cognizanti article Intelligent Automation: Where We Stand — And Where We’re Going, or visit the insurance section of our website.

Related Thinking

Save this article to your folders



Why Insurers Can No Longer Ignore AI...

What was once science fiction is fast becoming business reality. Here’s...

Save View

Save this article to your folders



Gaining Momentum: U.S. P&C Insurers...

U.S. property and casualty (P&C) insurers are belatedly embracing mobile...

Save View

Save this article to your folders



How Code Halos Are Changing the...

By finding meaning in the data that accumulates around people,...

Save View