How AI, ML can be agent superpowers behind the scenes

The decision to purchase insurance is high stakes for a consumer. In the case of life insurance, it could mean the difference between securing their family’s financial future or not. Buying a policy can feel overwhelmingly complicated and, at the same time, like something one can’t get wrong. It should come as no surprise, then, that many consumers still prefer to talk it through with a live human.  

This is why rather than debating whether AI will replace life insurance agents, it’s more productive to consider ways in which AI could be their superpower. 

Emerging AI and machine learning technologies have the potential to transform how agents work, for the better. Already, machine learning helps agents deliver more personalized insurance product recommendations to their customers, saving them time and improving conversion rates. And we’ve only just reached the tip of the iceberg: AI could help agents cross-sell other lines of insurance, quickly answer consumers’ policy questions, and beyond.

However, agents have years’ worth of knowledge about insurance and experience building personal relationships. Building technology to support them isn’t quite as simple as hiring a few engineers. Rather, a nuanced collaboration between technical developers and agents is critical to ensure the tech is helpful vs. cumbersome. After years of building machine learning technology for agents, my team has learned three valuable lessons. 

1. Transparency in AI decision making is critical.

Would you let AI make decisions for you, if you didn’t know how it was making those decisions? Many, understandably, wouldn’t. When it comes to AI-driven recommendations for insurance products, we quickly learned that if agents don’t understand what’s behind them, they are less inclined to use them. 

For example, our Plan Decider tool is designed to help agents be more efficient by quickly surfacing a list of recommended insurance policies to present to a consumer. In an early iteration, while the tool presented a ranking to agents, there was little explanation of the “why.” This made it incredibly difficult for agents to explain the trade-offs between different policy options to a consumer without doing significant manual research. 

This illustrates the importance of bringing transparency into algorithmic decision-making. Better yet, you can allow agents to customize inputs to the algorithm based on their conversation with a customer to compare policy options quickly and easily. 

AI isn’t a crystal ball. If you want smart, experienced agents to use it, you need to explain to them how it works and how it can help them help consumers. 

2. Start collecting agent feedback in the prototyping stage.

In many tech-driven industries, you’ll hear the advice, “fail fast.” An engineer working on a consumer app might have an idea for a new feature, launch it in beta to a random user group, and then scrap it if it doesn’t work out with little consequence. 

In the insurance industry, however, failure can be painfully slow. Developing new technology is incredibly resource intensive due to the industry’s complexities and regulation. A new feature in agent software likely requires intensive hands-on training and support to enable success. Failure could come only after significant investment is made. 

An alternative approach requiring a lighter investment is to bring agents into the prototyping stage. A prototype is a mockup of an experience, sometimes with basic functionality that can be demoed. It enables you to collect valuable insights about how someone might use the technology before ever writing a single line of code. This allows you to make big changes before development begins, so that when you do get to that beta phase, the product is close to final, only requiring minor optimizations and tweaks. 

3. AI doesn’t have to be flashy. It just has to add value.  

Agents are paid on commission. They are incentivized to sell policies that meet their customers’ needs, because commission is contingent upon policy retention. AI that helps them make the right sale to the right customer is, therefore, incredibly useful. 

However, it’s easy to get caught up in thinking AI has to be flashy and exciting, like a branded chat bot or fully designed app. Instead, what if AI could go behind the scenes and help agents before they even talk to a customer? 

There are many ‘hidden’ use cases of AI that could bolster agent productivity without requiring them to learn a new tool or platform. For example, AI could help route prospective customers into different personalized journeys. The journey to buy life insurance is often long and winding, in some cases, with years in between a consumer’s initial intent and an actual purchase. AI tools could help point a customer who is just getting started to educational materials, track their intent over time, and eventually, match them to an agent who specializes in the type of policy they end up gravitating toward. 

Agents are at the forefront of digital transformation in life insurance, and it can clearly be their superpower. However, developers in the insurance industry can benefit from these three lessons before jumping on the AI bandwagon: provide algorithmic transparency, collect feedback early and often, and build features that deliver clear value.