Unlocking AI

It is amazing how AI enables us to be more productive, organized and creative. We constantly think about how we can better leverage it to help our customers and employees, and have rolled out a few new projects that are already having an impact.

AI and Design
The pace at which AI tools are moving is just astounding. My team has fully embraced Figma Make—Figma’s AI feature that uses Anthropic’s Claude agent, and we have not looked back. It allows designers and non-designers alike to create coded prototypes using just prompts. We even trained Product Managers to use this tool as well, which allows them to creatively explore concepts without needing a designer's help. We have been using Figma Make to quickly prototype ideas and user test hypotheses on interactive user experiences without needing developers.

We are also partnering with our development team to connect our Figma designs to a repository of design tokens and our coded React components so that eventually Figma Make can code our designs using our exact code components allowing developers to implement new pages and elements in much less time.

There are numerous other ways we leverage AI for our design work:

  • Analyzing a screenshot and providing design feedback.
  • Suggesting edge cases we may not have considered.
  • Recommending ways to better engage users.
  • Creating slides for presentations
  • Image manipulation
  • Assisting with research plans
  • Analyzing research findings
  • Persona development
  • Content writing
  • Brainstorming concepts
  • And many others I am sure

I have seen weaker designers become much stronger by their embrace of AI tools enabling them to provide more value ultimately increasing their own value. It’s exciting to see how quickly AI tools are improving and how these new capabilities are allowing us to do our work at a faster pace. What does the future hold for design teams?

Customer Support
One place AI is already having a material impact is in customer support roles within company call centers. We recently rolled out an AI Call Screening agent that can answer every phone call even when we’re at our busiest, like during the Medicare Annual Enrollment Period. In past years, eHealth has pulled folks from all over the company to screen calls before transferring customers to licensed insurance agents, and even with extra help thousands of calls go unanswered. With this new AI call screener we have the capacity to answer every call and the AI is able to do it just as well, if not better, than a real person. We have done user research to see how customers have received their AI screeners and they are both surprised that it’s AI, and pleased with the experience. We have been able to find the right balance of having the AI agent show the right amount of empathy, but keep the user on task.

We’re now training an AI Sales Agent to field customer calls and answer simple questions that are frequently asked by callers. If the question is more complicated the AI Sales Agent will transfer the customer to a real agent. Everyone is really excited to see how this goes.

Content Loading
Another project we are piloting that is a great use of generative AI is loading insurance plan content from our insurance carrier partners. Each year hundreds of carriers send us plan data for thousands of plans, and there is no standard format, so up to this point we have had to manually process and review each plan to parse the data. This process takes hundreds of hours with dozens of people, and although they do a great job it is very slow and sometimes mistakes are made.

By leveraging AI we are able to import the plan data using two different AI models and then use a third to check the work and identify where the data differs. Instead of taking hours per plan, the data can be consumed and processed in minutes with more accuracy. Plus, we are able to utilize most of these people to do something else.

Helping users online
After launching a few customer service related AI projects, we really want to find a way to help our online users find and enroll in health insurance plans. Our primary customers are shopping Medicare plans during the Medicare Annual Enrollment Period. As you would guess—Medicare is an overly complicated program that puts a serious cognitive load on our users. In order to find a plan that actually meets their needs, they have to add all of their doctors and drugs. On average, our users add six doctors and ten medications, after which they still need to analyze dozens of different plans. It’s a tall order for sure, especially when you consider cognitive capacity, motor skills, eyesight and other accessibility factors of the average Medicare beneficiary.

We already have chat for users to engage a sales support advisor, so we didn’t want to create an AI version of that. We want to make something really helpful that allows the user to stay on the website, so we created an AI Shopping Assistant.

It turns out that right now AI sucks at reading websites and struggles to know what is going on, even after it scrapes the webpage html and reads the DOM. What AI is really good at is talking to APIs, so we have created an experience that allows the user to add all of their doctors, drugs and any filters, all through the AI. Our initial concept was a chat interface, but we plan to expand that to conversational voice chat soon. Imagine a user just being able to talk to an AI on the website… “I need a plan that covers Dr. John Smith in San Francisco… and it needs to cover these medications… Oh and I really like Sutter Health, so only show me plans from them.” What a frictionless way for users to narrow down their plan options.

It feels like with AI, we are only scratching the surface, so it is exciting to see where things go from here. I know it’s going to allow us to build some amazing digital experiences for our customers though.