5 Strategies for implementing an AI Chatbot in Healthcare
By Nora Osman, Associate Vice President, IT Service Management, Montefiore Health System
Recently I began implementing a strategy to bring AI to my healthcare organization. The reason was simple; to find ways to improve the customer experience while reducing the burden on the IT teams that deliver the service. The approach I took was to deliver the most straightforward and shortest solution that gets technology out of the way of the clinical provider, enabling them to be better at their jobs. This starts with eliminating all complexity for the customer while opening multiple channels for support to meet them where they are. While selecting an AI Chatbot to solve some of these issues, we ran into multiple challenges in Healthcare. Below I outlined both the challenge and the approach to take when addressing it.
The implementation of an AI Chatbot may be a technical approach, but it is much more of a cultural transformation than a technological one.
- Providers want to only talk to people and not a virtual agent.
The consensus was that providers (Doctors, nurses, clinical staff) don’t have the time to interact with anything but human beings, and they barely have time to talk to even the Service Desk.
Approach: Understand the top 20 issues or requests that likely comprise 80% of the need and simplify this delivery via the Virtual Agent. Not everyone will use this channel but remembering that different generations embrace technology differently is important. The AI Team (as we called it) set out to identify the top EMR-related how to questions and request types, and structured workflow to relay the step-by-step guidance in the simplest way possible. They used images from tip sheets, knowledge base content already published, as well as ticket resolution data.
- AI requires a wealth of knowledge content to ingest and learn. Fundamentally an AI Chatbot is only as good as the knowledge base content it is fed.
Approach: There’s no easy way around identifying needed knowledge base content and spending the massive amount of time necessary to create or update it. The best way to tackle this is to make it everyone’s responsibility, taking a divide and conquer approach. An exercise in the assignment of content to reviewers and contributors is necessary, and someone to manage the process of receiving and critiquing the content is also part of the process. Beyond the initial seeding of content for the AI chatbot to ingest, you also must have a model of curating new and needed content based on customer interactions on a regular (daily) basis to fill gaps, so customers continuously get refreshed content.
- AI must be integrated into multiple other tools/platforms as part of the design.
Any solution vendor will showcase the capabilities of their solution in its final, most optimized way; however, they won’t harp on the extent of the integration that will be needed for that solution to function effectively.
Approach: Coming to terms with what your strategy should be for how the solution is accessed by your customers is core to creating a solid plan for building those integrations. Understand what website, self-service portal, Teams, SharePoint, or mobile app access you’ll need and engage early to architect the design. Never underestimate the importance of a strong architecture exercise and having this ratified by the appropriate governance area not to slow the plan down.
- A core AI team with the right tools and mindset will need to own the platform and plan. Beyond having a strategy and a project, there needs to be a true champion in your sponsor as well as a technical team responsible for the day-to-day issue discovery and resolution.
Approach: From the onset, define a core AI team with a vision, mission, and spirit. Have a formalized project with a project manager, team manager, and several technical roles from areas that directly understand the customer’s pain. Beyond going live and learning what day 1 looks like, you must plan for day 2 and beyond. What’s the level of effort needed to engage with the customers and hear how they like/dislike certain functions in AI? Who will do what and how quickly? And how are escalations handled? A core team that works well together is critical to the success of the initiative, beyond bringing it to the finish line. As part of this area, I also suggest looking at changing roles permanently for individuals who will be taking on more of a customer experience “engineering” role, amassing both technical skills as well as a customer-centric charter for enhancing that experience daily.
- Shift-left is key to truly evolving and improving the customer experience with the AI Chatbot. Some customers will be ok with tinkering around with answers, and will even be ok with following instructions, but not all. Many will want the easy or easier button and will quickly lose interest unless you automate the solution.
Approach: Automation using RPA routines will be imperative to expediting adoption levels with this strategy. Start planning very early for what can be automated and either expedited or eliminated. Easier things like password resets can get customers to like the solution because it gives them speed and autonomy. Beyond that, software installation requests, hardware ordering and even account provisioning can all be factored into the plan. And once the adoption rates go up, the IT teams’ mindset changes the building to barricade future tickets from their queues. With time, focus on having a more self-healing and a self-delivery mode for customers. If something breaks, it should fix itself; if needed, it should order and deliver itself. Many of the tools to make this happen may possibly exist in your organization, and they just need to be integrated into a process.
As you think through the approaches outlined here, make sure you’re looking at this more as a marathon and not a sprint. The implementation of an AI Chatbot may be a technical approach, but it is much more of a cultural transformation than a technological one. It requires a mindset change that only comes with habit changing with time.