How Medisolv decodes healthcare reporting with generative AI

Medisolv is a healthcare quality reporting software company that assists hospitals and healthcare organizations report on quality measures to the Centers for Medicare and Medicaid Services (CMS). Given the company’s deeply technical category, SVP of Regulatory Affairs & Advisory Services Erin Heilman wanted to find a way to more easily deliver precise interpretations of lengthy rulings, create compliant and accurate resources for customers, and enable employees to digest and comply with regulatory changes.

Medisolv chose Writer for its graph-based RAG technology, adherence to strict data privacy and security requirements, and ability to automatically enforce legal, regulatory, and brand rules. With Writer, Medisolv built an internal knowledge assistant that serves as a central source of truth and created AI agents to swiftly analyze and develop resources on complex regulatory topics, leading to 80% time savings in asset production.

Tell us about yourself and Medisolv.

My name is Erin Heilman, and I’m the SVP of Regulatory Affairs & Advisory Services for Medisolv. Medisolv is a healthcare quality reporting software vendor, founded over 25 years ago. We help hospitals and healthcare organizations ‌report quality measures to the Centers for Medicare and Medicaid Services (CMS).

CMS dictates what hospitals and other healthcare organizations must report to the government to ensure they are fully reimbursed for their claims. It’s a deeply technical space that requires interpreting a lot of rulings that come down from the federal government, and Medisolv helps our clients understand those regulations. Our software helps them track those quality measures and ultimately complete those submissions on their behalf.

What led you to pursue generative AI?

What really got me started on this AI journey was the fact that the space we’re in is highly technical. There is so much information, so many rulings that come down from CMS, and downloading that data and interpreting it in a way that makes sense to an end user is extremely difficult.

Additionally, like many other marketing teams, we’re resource constrained. It seemed like AI was at a point where we could use it to take that information, dissect it, then use it to create something that could get us 70% of the way there, and then we could add that remaining 30% to get us to a better spot.

Another reason we began to look at AI was to support legal and brand compliance. We’ve been around for 25 years, and we’ve experienced exponential growth. Medisolv is in a space that requires deep-level technical expertise. Getting new people onboarded so they can understand how to talk like us and adhere to our brand was a challenge. Even our employees that come to us with 15 years of clinical experience struggle to learn some of the nuances. Being in such a complex and technical industry, which requires a strong knowledge base, we wanted to look at how generative AI could enable the team to be more efficient and help us operationalize and enforce better legal and brand adherence.

Why did you choose Writer?

When I was considering what AI software to choose, there were a few different factors that went into consideration. Number one is transparency. It was important that we pick a generative AI solution that would be very transparent about how it was built and how it would use our data.

Other aspects I looked for were the ability to create a terminology database and to enforce our style guide. The healthcare space already has a lot of acronyms, but because we have that additional layer of working with the government, it’s an acronym nightmare. Because of that, it was important for us to find a tool that allowed us to define the specific acronyms and language rules so that it made sense to our people.

Writer Knowledge Graph, their graph-based retrieval-augmented generation (RAG) solution, was another huge differentiator for us. Due to the deep level of expertise that you need to have in the field, it’s difficult to train people. Having an internal source of truth for everything we do, that everyone could have access to, was a big differentiator and incredibly beneficial for us.

How has Writer usage grown at Medisolv?

We did a phased rollout when we implemented Writer. We started with marketing because I knew they would find immediate use for it. We saw instant efficiency gains in writing emails, first drafts of blog articles and webinar scripts, and introductory emails. We expanded next to the sales team, who use Writer for various workflows like drafting outbound emails.

We then rolled Writer out to our client services team, who tend to be highly technical people from clinical and medical backgrounds. They might not necessarily be the most skilled writers, but they know the information. We saw immediate pick-up in this team, because they were now able to more confidently write out a draft of the information they already know, and then have Writer come in and polish it up. Our client services team was so excited for Writer because it helped them overcome a roadblock we all face as writers and gain confidence in the materials they could produce.

Tell us a bit more about how you and the team are using Knowledge Graph.

Knowledge Graph will help us answer and ensure widespread compliance for these very complex regulatory issues. For example, if you search for “what year is the cesarean birth measure required to be submitted under a specific program” in Google, it is really hard to find an accurate answer.

With Writer, that internal knowledge is centralized and we no longer need to comb the internet to try and understand the information. Instead, information that has already been verified by our experts will be right there at your fingertips. Our teams won’t have to continually fact-check with our client services team. Instead, they can just use Knowledge Graph and ask their questions right there.

Tell us about your custom AI agents.

We worked with Writer to create a custom AI agent to draft highly technical blog articles. As I’ve mentioned, our space is very complex and can be difficult to understand. CMS has certain measures that hospitals and healthcare providers have to report to the government. But those measures are extremely technical, and while they’re called “human-readable,” they really have to be interpreted by a computer. These are called specifications, and all of the lines look like computer code, so it’s hard to understand how patients qualify for different things.

For example, if a mother is coming into a hospital for birth, she would fall into a particular population and qualify for specific things. But to‌ understand what population she’s a part of and what she’d qualify for, you have to “speak” the logic in these specifications. You need to have that clinical background because the specifications reference information that only someone in a clinical setting would understand.

For me to understand how someone qualifies for a measure and to disseminate that information, I’d have to go through the specification myself. I’d read through one of these documents and try to understand the rationale behind why they want to lower the rate of cesarean births and understand the classification of the population, and then meet with a member of the client services team to see if I interpreted the specification correctly.

After all the prep work I’d done, it would be at least an hour-long interview with client services to get an understanding of the technical backend information. Then we could finally write the blog from scratch. The blogs we write need to make it easy for the reader to understand the information without having to read these specifications. It all just takes an extraordinary amount of time.

Writer helped us alleviate the burden on my team of going through this whole process. We created an AI agent that took those specifications — the deeply technical, essentially computer code — and translated it into something that an actual human could read. This was a huge benefit for us, because within two minutes of adding the link, the agent gave me the rationale that I wanted.

This AI agent gave us a 70% complete blog article that we could use and pair with the review of the specification. It took away that upfront burden of trying to understand the specification and made the whole process exponentially faster. With all of the people that were involved, it used to take us 10 hours to produce one of those articles. But now, we’re looking at two to three hours from start to finish.

We also created a regulatory agent with Writer. In the healthcare quality reporting space, there are rulings that come down from CMS every year, multiple times a year, that are thousands and thousands of pages long. These rulings dictate what a hospital or healthcare organization must report on each year to the government, but they change each year, and you can only know what’s changed by reading this thousands-of-page-long document. If anybody has ever read a ruling, you will know that it’s very difficult to digest because of all the specific references and pages of information. It’s not intuitive for somebody who’s just trying to understand what’s changed.

For us at Medisolv, we are very good at reading these documents, interpreting what they say, and then translating them into simpler information for our clients, but it’s a huge pain point. It takes a ton of time. But with our AI agent, we can now input this very long government document and quickly dissect it. The agent generates a table of contents that indicates where information related to specific regulatory issues are located and actually pulls that information so you can see how a specific issue has changed.

The ability to take this 10,000-page document and distill it into something readable is a huge benefit and time-saver. We can take this information and repurpose it into blogs, articles, or help us create first drafts of different types of assets.

What other Writer features do your teams rely on?

We found immediate gains from the pre-built agents, like recaps which takes a recording and generates a summary of it. When we were first testing out different use cases, I was at a conference and I recorded each session that I attended. When I got back, I uploaded all of the recordings into recaps and got summaries for all the different sessions. I could then create this entire narrative from the team’s notes and the recaps output, and turn it into a 40-page summary that shared great information for the rest of the company. That example was great to show us the tip of the iceberg in terms of what Writer can do.

Our marketing and sales teams also uses the outbound emails agent, which allows them to generate emails when they’re short on ideas or if they need a starting point. For example, if they’re going to a trade show, they can just put in some basic information about the event and our booth, and then it generates a series of emails for them. This has been incredibly helpful in jumpstarting the whole process.

Another feature we use all the time is Ask Writer. It’s incredibly helpful to bounce ideas off or in summarizing and extracting information from different sources. Since we have such technical documents and resources, Ask Writer is a great way to get quick answers to questions someone may have or get the right reference.

The style guide feature is extremely important to us as well, and we love that it can be incorporated right where we work. This takes away that burden of needing to reference an external PDF or continuously train team members to adhere to our style guide. Instead, we can meet them wherever they work.

What was your AI adoption strategy?

The implementation process was really smooth due to the great customer support we’ve had from Writer. I met with the team each week since generative AI was new to me, and I wanted to make sure adoption was done right. The Writer customer support team enabled me to share with my executives why generative AI was the right decision for us, as well as helping me understand what the day-to-day users would need.

The Writer team was great at teaching us how we could best use generative AI for ourselves. They helped us distill myths about AI, show us different kinds of use cases, and how we could implement Writer in our daily tasks.

Executive buy-in from the beginning was very important. Making sure our leaders were involved in the decision-making, understood why and how we would use generative AI, what the use cases could be, and ultimately how we could operationalize teams and improve efficiencies was key to smooth adoption. This top-down approach kept everybody motivated and at the forefront of their minds.

What outcomes are you looking to measure?

Our current goals are really around adoption. I first wanted our teams to focus on the value of Writer, and demonstrate how it improved our workflows. We saw almost immediately that 70% of users were using Writer for 30 days straight after deployment. For us, recognizing that 70% of people were using Writer immediately indicated great value and how we could be saving time.

Many of the teams who previously weren’t comfortable creating written materials are now doing so thanks to Writer. For example, having my client services team send me an article that has everything written out in complete paragraphs with correct brand compliance and everything, that’s a huge win.

What advice do you have for other AI leaders?

The best time to plant a tree is 20 years ago, and the next best time is today. They said the same thing when blogs first were published. This is really true, too, for generative AI.

It’s a monumental shift in the way we process and how we use the internet. Getting on board now allows you to be at the forefront and make changes to your team or company that will bear fruit in the future. If you are part of a niche industry like ours, the ability to establish and easily share that internal domain knowledge will differentiate your ability as a company to grow.

Additionally, for me, by rolling out generative AI early, I put myself in a strategic position of being a leader that brought AI to our organization. First is a great place to be whenever you’re in a strategic position.

What’s next for your team?

We’re currently experimenting with Writer AI Studio and the no-code environment. I took the Writer AI Studio certification course, which helped me understand how to build my own agents. I was able to immediately build out an agent that helped me dissect a hospital’s government results report and identify measures that they performed poorly on. It’s a bit of a magic trick for clients to take a complicated, hard-to-read CSV file and easily discover the specific measures that brought down their overall score in just 15 minutes. It blows everyone’s mind, especially mine. I’ve got hundreds of ideas that I can now easily test out, all because of the no-code environment.