Veradigm develops self-service healthcare chatbot "Billerbot" with Azure AI

Ensuring accurate and timely billing for medical services is a complex and time-consuming process. The Veradigm Practice Fusion Billing Services team is on point to manage this critical operation, freeing up its customers’ billing staff to handle other critical business. However, there can be a learning curve for small and independent practices that are not familiar with medical billing workflows and technology. 

“We had just released the Practice Fusion Billing Services solution when we started to see an influx of questions coming from these (smaller) practices,” says Ian Maurer, Vice President of Development at Veradigm. “They were reaching out to our billing staff with basic questions about billing terminology, workflow, and process. It was really bogging down our billing team.” The company recognized it needed a way to support its small business customers that wouldn’t overburden the billing team. But how? Executive action would help Maurer find the answer. 

An executive mandate leads to AI innovation

In early 2023, Veradigm leadership issued an executive mandate to begin exploring ways to embed AI enhancements and capabilities into existing products. “ChatGPT had just come out, and there was a lot of buzz in the industry,” Maurer says. “[Leadership] wanted us to figure out how we could take advantage of this type of technology within the products and services that we deliver today.” 

A small leadership committee was formed to brainstorm potential generative AI innovations. During the initial meeting, the team focused on identifying the types of AI capabilities that they felt could be implemented and deployed across one or more products that year. Each team member chose real-world issues that could potentially be resolved using an AI-based solution. For example, Maurer visualized adding a chatbot to the Practice Fusion EHR. 

“I had been thinking about developing a support process to handle the types of requests the team was receiving,” he says. "Many customers were asking the same basic questions about billing, related terminology, and process. I thought, maybe if we insert conversational AI into the Practice Fusion product, we could deflect a portion of these questions that our billing staff encounter day to day,” Maurer adds, thus freeing up billing staff to focus on other tasks. 

Maurer and the other committee members spent a month independently honing their ideas, then they reconvened to vote on next steps. “We reviewed all the ideas, and mine was the least complex,” Maurer says. “It was also the least risky because we wouldn’t be dealing with claims data or clinical data. It was just a generic knowledge base. Ultimately, everyone agreed that my idea would be a great one to pilot.” 

Developing a new feature with Azure AI Search, Azure OpenAI Service, and Zammo.ai

Although he’s not a developer by trade, Maurer personally worked on developing the generative AI feature in the initial phases. “Usually, I’m managing teams that do that type of work. But this was a pet project, and there really weren’t people available to jump in and do the work,” he says. Veradigm chose the Azure OpenAI Product Group’s GenAI Accelerator, based on the Zammo.ai SaaS solution, to deploy the proof-of-concept solution. 

“Practice Fusion is a very modern, sophisticated SaaS product. We had to make sure that any conversational AI enhancement would meet that standard. We couldn’t release something that’s unresponsive and provides a poor user experience,” Maurer says. “We liked how Zammo.ai solution resides privately and securely on our Azure tenant. They’ve worked with many healthcare organizations, so I knew they understood our space. If you can demonstrate as an organization that you can meet healthcare regulatory requirements, that goes a long way allowing us to be able to use your solution.” 

Maurer selected an out-of-the-box solution from Zammo.ai that took advantage of Azure AI Search and Azure OpenAI Service within a ChatGPT model for the development work. “I modeled my solution on it,” he says, adding that he was able to customize it to meet the specific needs of his project. 

The AI was trained in a closed system based upon a comprehensive billing knowledge base to create a new chatbot dubbed Billerbot. Maurer and his team curated data from internal documents and external resources as a starting point. Then, a web crawler was sent to scan and retrieve data from the Practice Fusion online product documentation on an ongoing basis. Maurer estimates the web crawler updates the knowledge base with the latest data about every five minutes, ensuring customers get the very latest information. 

“Microsoft has given us better control outcomes so we can put guardrails up to prevent the agent from accessing content outside of the knowledge base,” Mauer says. This enabled them to limit the nature of questions that can be answered without speaking to an agent. Next, the chatbot was trained to generate normalized, real-time responses to common Practice Fusion billing questions, including common onboarding issues.  

Reducing support calls with Billerbot

Billerbot made its debut in 2023, less than a year after the idea was pitched. The chatbot has been available to all Practice Fusion clients since day one. However, Veradigm opted for a quiet rollout, alerting a limited number of users to its availability. As of early 2024, approximately 200 of Practice Fusion’s 20,000 practices were aware of the feature. More customers are being introduced to Billerbot every day, with a target of about 700 practices by year’s end. In the meantime, Veradigm is monitoring the results and honing the tool. 

As of February 2024, Billerbot appeared to be deflecting 5 percent of billing queries. Veradigm has a long-term goal of deflecting 10–15 percent of support requests. With additional users coming online daily, Maurer expects to hit that target sometime in 2024. In the meantime, he’s thinking about yet more ways to help billing customers get the help they need. 

“We’re thinking of this as a foundational product that we can build on in a couple different ways,” he says. “There are smaller enhancements planned, but there are also some more powerful capabilities that I’d love to implement at some point. We’ll see what’s possible.” 

Find out more about Veradigm on Twitter, Facebook, and LinkedIn.