Quick Wins, Lasting Impact: Pragmatic AI For Insurance

It’s impossible to read any industry publication without finding an over-saturation of content about AI. Just as the early internet promised to take just about every consumer experience online, and smart phones & wearable devices turned everything “smart” AI seems to be all the rage. 

Currently, we seem to be right at the “Peak of Inflated Expectation” phase in the Gartner Technology Hype Cycle. While we may be at or nearing the peak of the hype cycle, I’m here to tell you that this doesn't mean slowing down. 

While frivolous investments in making everything AI enabled presents a risk, right now is the right time to be laying solid groundwork to capitalize on a more stabilized AI ecosystem. Your goal should be to reap short term benefits without over extending your organization, while keeping your foundation rock solid for future developments. 

Sounds easy enough, right? 

I recently completed some course work on AI strategy at MIT and have been meaning to write on the topic, but haven't been sure where to start. An upcoming webinar on the topic of reinventing underwriting data intake finally prompted me to actually get some ideas on paper. Streamlining data intake was an effective prompt not only because it presents clear AI use cases, but because the process of intaking clean, accurate data is a core enabler of future integrations, insights, and, yes, AI implementation. 

When looking at your use cases for AI, whether it be your underwriting processes, claims management, or customer experience, there are a few things to consider in order to create an effective plan of action. Some core areas of focus in developing your AI strategy are change management, balancing the short and long term, and determining the true costs and benefits of using AI. 

Let’s dive in. 

Change Management

A common theme I hear when talking to insurtech companies is that sales are easy, implementation is hard. It shouldn't come as a surprise that the industry of risk management is adverse to the risks that come with process changes; the processes in place have worked well for the industry for quite some time now. The same challenges are likely to rear their heads with AI implementation, but you can ease that transition with the right groundwork laid. 

Successful change management that truly leads to buy-in should lean on a mix of rational, political and cultural strategies.  

The rational strategies include the formal design of the change. The goals, KPIs, business cases, and process redesigns involved in your AI implementation should be well thought out and not over complicated. 

Political change management refers to exercising power. This can take the form of placing accountability for results on key management or spreading it around the organization. Organizations should also consider how transparency on process & performance can ease the minds of your team. On the other side of this public or non-public pressure can be necessary to push change along at times.

From a cultural standpoint, you need to present a clear vision, engage with employees, celebrate wins, and provide proper incentives. Clarity of direction and a feeling of empowerment from the right incentives will help drive buy-in organically. 

Balancing Long Term Vision With the Need for Quick Wins 

Digital transformations can fail spectacularly in many ways. While there are countless prominent stories of companies not thinking about the long term (Sears, Blockbuster, Kodak, etc.), digital transformations can fail without enough short term focus. Quick wins are essential to maintaining momentum, buy-in from employees, and capital for your projects.

GE once set out to be a “top 10 software company,” investing heavily in Predix, a cloud-based IoT platform, taking on an overly ambitious goal that left them spread too thin to generate any short term successes. The British Broadcasting Corporation launched the Digital Media Initiative in 2008 to modernize production systems under a single digital platform, pouring resources into an all-encompassing system without delivering anything the staff could use in the near term. 

To balance your long term aspirations with short term momentum, you need to frame your initiatives properly. Long term planning requires accurate data and thoughtful data governance procedures. Your data is the foundation upon which AI can be trained, integrations can be powered, and decisions can be made. At every step of digitization, you should think about how data quality, storage, and accessibility can be improved. 

At the same time, companies must consider where quick wins can be achieved. Can redundant processes be removed as you look to reduce mistakes in your data? Can the speed of a process or efficiency of an underwriter be improved while increasing the accuracy of their decision making? These are the types of changes that help you stack quick wins into the masonry of your digital future. 

A good example of this I saw recently was from Feathery, who is building no code, customizable underwriting data intake workflows with AI powered document intelligence. The quick wins here are saving time by eliminating manual data entry and streamlining customer experience. The long term thinking pertains to the data quality and ability of the product to integrate with current or future systems. 

You can learn more about Feathery and more in the upcoming Reinventing Data Intake webinar I mentioned earlier. Click here to register.

Thoughtful Implementation of AI: a Framework 

With AI mentioned in everything you read, and every product being “AI powered” now, how do you decide what’s worth pursuing and what will be a waste of your time? It sounds like a messy undertaking to sift through the noise, but thankfully, a pretty simple and effective framework exists.

First, map out your objectives that satisfy both the need for quick wins and the requirement for long term value and flexibility (particularly around data quality and organization.)

Given your parameters and plan for implementation, use the Gen AI cost equation to determine if this is a viable area for implementation of AI. The Gen AI Cost Equation, developed by Dr. Rama Ramakrishnan at the MIT Sloan School of Management, is designed to help businesses easily determine the viability of successful AI implementation in different business processes. 

To use the framework you first need to disaggregate processes into discrete tasks; that is, break the process you are looking at down into granular steps, so each step can be looked at individually. 

For each task, roughly consider if the cost equation below is satisfied:

The ongoing cost of using AI

+ the upfront cost of adapting AI for the task

+ the ongoing cost of detecting and fixing errors in AI’s output on the task...

Should be less than the cost of doing the task without AI 

It’s as simple as that. Some cost considerations might be the cost of training a proprietary model versus using something out of the box. You also might consider the long term opportunity cost of not changing anything as an additional cost of doing the task without Gen AI. It’s also worth noting that the costs in this equation are constantly changing with technology and your business, so it’s worth revisiting somewhat frequently. 

Looking Into the (Near and Long Term) Future

In the end, it all comes down to recognizing the current moment—full of AI hype yet ripe with opportunity—and ensuring your organization doesn’t lose sight of what really matters. You want to plant seeds for the future, but you also want to harvest some benefits today. Building a solid data foundation, looking for clear quick wins, and applying the Gen AI Cost Equation (or a similar tool) are all concrete steps to stay grounded as you navigate this rapidly evolving landscape. There will always be more case studies and new AI tools coming out daily, but if you focus on creating lasting change through thoughtful planning and realistic action, you’ll find that you’re more than ready both today and well into the future.

If you are interested in learning more about AI enablement and use cases in underwriting, consider joining Insurance Nerds, Next Insurance, and Feathery for our upcoming webinar on April 22nd. You can register here.