Transforming the Patient Journey and Reducing Costs

Transforming the Patient Journey with AI-Powered Chatbots

Hear from Harvard’s Boston Children’s Hospital Digital Transformation Officer, Bill Gagnon, as he shares how chatbots can streamline the patient journey and how to apply AI effectively within chatbot experiences. In addition, this discussion explores practical lessons from healthcare leaders who have already implemented these tools successfully. Whether you’re just beginning your transformation or already in progress, you’ll gain insights and suggestions you can apply immediately.

What You’ll Learn in This Transcript

This conversation covers:

• Why Boston Children’s Hospital adopted chatbots
• How chatbots reduced call center volume during COVID
• Why an open-source chatbot approach worked in healthcare
• How the team manages compliance, content, and escalation
• What it takes to maintain and scale the chatbot across departments

With that context in mind, let’s step into the conversation itself.

Video Transcript (Edited for Readability)

Opening and Context

To set the stage, Brian Anderson opens the session by framing where this conversation fits within a broader digital transformation journey.

 

Brian Anderson: All right, we are live on the webinar. So thanks thanks for joining. I want to kick this off. We’re gonna run our second part of our series around digital transformation. This is part two of our digital transformation series. This session focuses on transforming the patient journey and reducing costs with AI, chatbots, and open source. Our guest is Bill Gagnon, Director of Digital Transformation at Boston Children’s Hospital. Bill, welcome back. Could you introduce yourself?

 

Bill Gagnon: Thanks, Brian. I appreciate being here. Yes, I’ve worked in digital transformation since the mid-1990s. Over the years, I’ve worked at organizations like Citigroup, GE, and Cigna. More recently, I’ve focused on Boston Children’s Hospital. Today, I’m excited to share what we’ve learned from deploying our chatbot and related initiatives.

 

Brian Anderson: Awesome. Thank you, Bill. And I’m Brian Anderson. I’m the CEO of Augusto Digital. We focus on digital transformation and software solutions in healthcare. And so we’re working in this space with clients every day. So we’re looking forward to this webinar here. So let’s get this started.

 

Why Chatbots Became a Priority (Bill’s Background With Chatbots)

To understand the decision to deploy chatbots at Boston Children’s Hospital, it helps to start with Bill’s long history working with these tools.

 

Brian Anderson: So, hey Bill. Why don’t you start by just sharing some background on your kind of usage of chatbots over the years, and why you’ve decided to implement one of these at Boston Children’s?

 

Bill Gagnon: Sure, absolutely. I’ve worked in digital since the early days of websites. Over time, the work evolved into process mapping and understanding customer experiences. I learned to map workflows across departments and identify pain points. In multiple organizations, one consistent need came up: people needed information 24/7. In healthcare, those might be questions like:

• Where do I park?
• What time do I arrive?
• How do I find the right location or doctor?

Those questions create pressure on call centers. A chatbot can take that load off by giving answers quickly and consistently. I implemented chat at GE and Cigna, and later expanded it at Boston Children’s Hospital.

 

The Catalyst: Improving the Digital Front Door

Building on those early lessons, Brian asks what specifically triggered the chatbot rollout at Boston Children’s.

 

Brian Anderson: Great. Let’s talk a little bit about the catalyst of what led you to improve your digital front door for Boston Children’s, and what drove the chatbot implementation.

 

Bill Gagnon: When I started at Boston Children’s in 2015, I met with 18 service areas. We ran working sessions that included process mapping and opportunity identification. Over and over, people said the same thing: we needed to deliver more information to end users. Those users include patients, families, referring physicians, and researchers.

The chatbot stood out because it was a fast, high-value win. We could spin it up quickly, load in accurate content, and give users something that felt like a real conversation. Done well, it works like an interactive FAQ, but with more flexibility and a better user experience.

A Real Example: COVID-19 and Call Center Impact

To make the impact tangible, Brian asks Bill to walk through a concrete example.

 

Brian Anderson: That’s great. I have an example here, some information that you shared with us to provide an example scenario. Let’s walk through that together. On your website, you’re highlighting COVID-19, right?

 

Bill Gagnon: Right. We launched the chatbot in 2018, but we moved carefully at first. Healthcare requires compliance, and we had to avoid sharing PHI or medical advice. Then COVID hit in March 2020. Suddenly, people flooded the website because they needed answers immediately:

• Will my child’s appointment still happen?
• What do I do with a specific condition right now?
• Where can I find updated COVID guidance?

We launched a COVID chatbot within the third week. Call center volume dropped dramatically because the chatbot answered many questions right away. Because the chatbot is open source, we could update it daily. We could see:

• Which questions people asked
• What questions weren’t answered
• Where the chatbot needed improvement

The chatbot maintained about a 95% answer rate. For unanswered questions, we followed up and then added that knowledge back into the system.

 

Beyond the Website: Internal Use at Scale

As usage grew, the chatbot evolved into an internal tool as well.

Brian Anderson: So it expanded beyond patient use, right?

Bill Gagnon: Exactly. It became useful internally, especially during COVID. Staff were working remotely and didn’t have a single place to find critical information. Some had paper lists of numbers or fax details. Others had separate references, so we installed the chatbot on staff desktops. It became a centralized knowledge tool for frontline agents. That reduced call handling time and improved consistency.

 

Accessibility and Unexpected Gains

Looking beyond efficiency, the team discovered additional benefits through accessibility improvements.

 

Brian Anderson: So you’re able to adapt and learn. Whereas a lot of times, people will go to your website, and you don’t know exactly what they’re doing. You have the analytics, but this is much more informed because they’re asking questions. Right?

 

Bill Gagnon: Absolutely right. We also improved ADA compliance. We added speech-to-text so users could interact more easily. That created an unexpected benefit. Call center agents could repeat a caller’s question to the chatbot. The tool would return the best answer instantly. As a result, call time dropped, and the agent experience improved.

 

AI and ChatGPT: Why Controlled Chatbots Still Matter

With newer AI tools in the spotlight, Brian asks how this approach fits into the future.

 

Brian Anderson: Let’s dive a little deeper into the AI. This isn’t ChatGPT. You guys were using this before ChatGPT even came out. With that in mind, how do you see AI evolving here?

Bill Gagnon: I’m aging myself on using this back in the mid-nineties when it came out. I saw that there was a big use for this, and it’s always that. How do we actually reduce people’s time online and get them to either acquire or to engage? These tools were used when I was working at GE International, where we had many different languages, and they had the ability to select their choice and come in and talk to people about different things, from different finance tools or insurance tools. And at Cigna, the same thing.

These chatbots are based on heuristic logic and keyword/sentiment matching. It pulls from a controlled database. That’s important in healthcare because we can ensure:

• Content is vetted and compliant
• Responses do not provide medical advice
• PHI is protected

ChatGPT can be powerful, but this tool remains valuable because it’s controlled. We know what’s in it, and we can verify every answer.

 

Why Open Source Was the Right Choice

The previous emphasis on control also influenced a key technical decision.

 

Brian Anderson: Why did you choose open source instead of a SaaS chatbot?

Bill Gagnon: Two reasons. First, budget matters (especially for nonprofits). Second, customization matters. With SaaS tools, you often get 80% of what you need. The remaining 20% can take months to request and may never get prioritized. With open source, we can invest directly in improvements. Augusto helped us enhance the tool quickly in ways that mattered for our workflows.

 

How Content and Maintenance Work

Once the system was in place, audience questions were asked on governance and maintenance.

 

Brian Anderson: “Is chatbot content centrally controlled or updated by departments?”

Bill Gagnon: Departments contribute content. However, it’s centrally managed by my team. We also partner with legal, compliance, and clinical leaders. That ensures accuracy and safety. We also write questions in multiple forms because patients and clinicians use different languages. In addition, we’ve expanded content into French, Portuguese, and Spanish. Over time, we may integrate translation tools for even faster scaling.

Brian Anderson: “What does maintenance look like or require?”

Bill Gagnon: At the beginning, it took a group effort. We had 5–10 people involved for the first few months, gathering common questions and building the knowledge base. Now, maintenance is much lighter. One person on my team oversees it, along with reputation management. Daily review takes about 30 minutes. Most work today is monitoring feedback and routing concerns appropriately.

 

Handling Sensitive or High-Risk Inputs

Even with strong governance, sensitive situations still arise.

 

Brian Anderson: How do you handle sensitive information, like suicidal ideation?

Bill Gagnon: Because the chatbot is always available, some users share urgent or sensitive messages. We treat that seriously. We added keyword triggers in collaboration with legal, compliance, and clinical staff. If certain terms appear, the chatbot escalates with immediate guidance and directs the user to appropriate help. We also review logs to identify new terms and improve our response library.

 

Wrap-Up and Next Webinar

With those safeguards in place, Brian begins to wrap up the session.

 

Brian Anderson: If you’d like a copy of the presentation, email me. We’ll also publish the session on our site. Our next webinar continues the series and focuses on building digital transformation teams that adapt to changing priorities, technologies, and security needs. That will be on August 16.

Thanks again for joining. And thank you, Bill, for sharing your experience.

Bill Gagnon: Thanks, Brian. Appreciate it.

 

Conclusion

Transforming the patient journey requires more than deploying a chatbot. It takes the right strategy, the right operating model, and the discipline to scale what works while protecting trust and experience. The insights shared here offer a real-world view of what successful teams prioritize and how they move from pilots to outcomes.

 

If you want help applying these lessons to your system, Augusto can support you from planning through implementation and rollout.

 

Schedule Meeting with an Augusto consultant.