The nostalgia bug bit me when I heard "The Matrix Resurrections" was set to premiere at the end of December. To prepare for its release, I rewatched the prior movies. I was surprised to find that these movies captivated me in a wholly different way than when I first saw them. As someone who moved into the world of artificial intelligence (AI) and machine learning in the last few years, I found the original movies' themes of reality and choice inspired a lot of thought around AI – specifically in the context of conversational data.
"If real is what you can feel, smell, taste and see, then real is simply electrical signals interpreted by your brain." – Morpheus, Matrix (1999)
The basic definition of AI is computers or machines performing functions typically associated with intelligent beings. These computers use highly advanced mathematics and robotics to imitate human cognitive and advanced motor functions. The machines can take actions like shaking a hand if someone extends it or recognizing a picture of a four-legged animal as a dog or cat. These capabilities are surprisingly complex when you think about each task that goes into making those assessments. AI can be narrow—the ability to do one dedicated task. Or, it can be general—the ability to do multiple, disparate tasks.
For our purposes in insurance, let’s divide AI into four major categories:
- Vision: The ability to interpret and act on or generate pictures and videos
- Speech: The ability to interpret and act on or generate sounds
- Natural Language: The ability to interpret and act on or generate written words
- Decision-Making: The ability to approximate models of human choice and actions
For insurance, the one category that can upend our entire system is natural language. The power behind natural language is conversational data. Conversational data originates throughout the insurance lifecycle from policy acquisition to claim resolution and policy renewal in the countless emails, webchats, phone calls and text messages among the policyholder, carrier employees, agents/brokers and ecosystem partners. However, the industry has done little with this data for several reasons, including limited capacity and capabilities. We have spent significant capital and sweat equity to modernize and undergo digital transformation over the last decade. We have also allocated considerable resources to build data infrastructure and compete with other industries to acquire data scientists and data engineers.
See also: The Challenges of 'Data Wrangling'
These limitations will change in the next few years as the industry catches up. Carriers have started to unlock the power of conversational data through new conversational AI capabilities like chatbots. However, for most carriers, these capabilities have been primarily focused on cost reduction through self-service opportunities to battle the pressure on premium growth from alternative direct-to-consumer offerings from insurtechs and companies in adjacent industries such as automakers and smart home manufacturers.
"Have you ever had a dream, Neo, that you were so sure was real? What if you were unable to wake from that dream, Neo? How would you know the difference between the dream world and the real world? " – Morpheus, Matrix (1999)
In this transitional state with limited conversational AI capability, the bots and processes are more mechanical, presenting many points of potential customer frustration and friction if carriers are not thoughtful in their approach. During this period, carriers can minimize some of the impact by following a few principles:
1. Minimize the bot. Don’t overengineer the process for the sake of automation. Designing intelligent communication flows taught us that if a bot asks a policyholder more than five questions, they are likely to yell or frantically type “talk to a representative.” The carrier will start in the negative for that interaction. Carriers need to ask, “What is the smallest number of questions I can ask to get to the root of the issue to resolve?” Much of that insight comes from poring over the prior data to find the underlying patterns.
2. Don’t be afraid to tag in a human earlier through failover processes. Doing so will both prevent the interaction from starting poorly and give the bot another trainer. Before moving on to the next task, build in time for the operator to provide feedback from the conversation and policyholder behaviors to inform future training sessions with the bot. It is a small investment today that will pay enormous dividends.
3. Train, train, train. Before releasing any bot, make sure to run it through its paces. Map common journeys and understand policyholder intent and utterances so the bot produces the right response pattern. This requires a good data model with robust infrastructure to meet response speed. Also, having a dedicated research team for new natural language processing algorithms will help immensely, as new models are making it easier to understand context within the text (a good example is transformer models).
Remember these principles to reduce poor experiences during this transition. Eventually, the bots will substantially improve, becoming much better human proxies as machine learning algorithms improve at a breakneck pace. With continuing pressure from alternative options in the market, many carriers will likely acquiesce to the allure of moving many more human tasks in insurance completely over to bots to remain cost-competitive. The insurance industry has been dealing with this commoditization trend over the last decade as carriers underwent the current digital transformation of web, mobile and self-service. As a result, insurance has become a low-engagement, price-sensitive category for the common consumer.
The continuation of this trend will require carriers to constantly manage the optimization between perceived policyholder value for their products, carrier pricing power, cost of service delivery and internal change agility to gain even modest market share versus peers. Carriers will no longer be policyholder experience managers, they will transform into data and technology companies optimally transferring risk, especially as usage-based pricing and new competitors become more prominent across the consumer base. Only the most sophisticated, data-oriented companies survive in this future.
"Because you didn't come here to make the choice. You've already made it. You're here to try to understand WHY you made it." – Oracle, Matrix Reloaded (2003)
It’s crucial to pay attention to additional environmental changes that are creeping up. The recent backlash on Meta and other firms' use of machine learning and AI to drive customer behaviors may cause consumers/policyholders to revolt because they distrust companies that collect data, thinking they will misuse it or try to “manipulate” them. Additionally, as we discover insights into systematic bias in machine learning and its societal impact, regulators may step in with policies restricting its usage. Finally, we could make the choice ourselves as an industry – do we want to continue down the commoditization path by focusing on rote cost-cutting?
See also: Why Exactly Does Big Data Matter?
As Merovingian said in "The Matrix Reloaded," “Choice is an illusion created between those with power and those without.”
We should remember why we are in insurance — to create real value for the policyholder, the carrier and other stakeholders — and do everything we can to ensure the power of conversational data is used to reverse the commoditization trend. Our industry began with the dream of human empathy and fear reduction when people were the most uncertain about their future.
Rather than investing in artificial intelligence—teaching machines to work like humans—we can empower agents, adjusters and underwriters to keep human empathy in play by investing in augmented intelligence—using machines to help humans work better. We can use conversational data to create brand segmentation strategies that meet policyholders’ needs for who they are and what they want. And we can use conversational data to both cut costs and automate non-value-added tasks like phone tag while focusing on creating truly loveable experiences for policyholders. Ultimately, it is our rebellion and our choice to take the blue pill or the red one.
