Intelligent virtual agents (IVAs) can help transform customer experience but present a massive risk-reward. Do it wrong and those patients will go elsewhere.
Every business or IT leader I talk to today has customer experience (CX) improvement as one of their top initiatives. My research shows that 95% of companies now compete on CX versus only 22% a mere five years ago. While there is no silver bullet to improve CX, AI is viewed as an enabler of change allowing companies to deliver differentiated experiences.
Use of Intelligent Virtual Agents in Healthcare
Healthcare companies are increasingly using AI powered intelligent virtual agents (IVAs) to improve patient and member experiences. While this AI tool offers significant benefits, it also introduces new challenges. According to new insights from Five9, healthcare companies must navigate complexities such as maintaining regulatory compliance, integrating AI systems with existing workflows, and ensuring security. They must also update their systems and prevent AI from learning incorrect information, which adds to the operational demands.
Five9 published the insights it collected from customers in an e-guide designed to help healthcare companies create AI-powered self-service channels. The data comes from over 60 organizations using Five9’s IVAs, including providers, payers, suppliers, and pharmaceutical companies. Let’s dive deeper into some of the key insights from the e-guide, AI in Healthcare: How AI Drives Value For Five9 Healthcare Customers.
IVA Use Cases
Businesses across industries are highly interested in IVAs. They have the potential to completely change CX as they can provide fast answers with little to no human labor costs. For regulated industries, such as healthcare and finance, ensuring compliance has been a barrier to entry. AI can play a huge role here as AI-powered solutions can ensure mandates are being met more accurately than people can.
The primary driver for adopting IVAs is to reduce frustration and improve conversations with people. Call volume varies greatly for healthcare companies: a large provider may receive millions of calls per month, while a local clinic may only get a few thousand. With an IVA, all organizations benefit from low abandonment rates (commonly 6 to 15 percent), average queue times of under one minute, and only about half of the calls needing an agent’s attention.
Five9 found that the top priority for healthcare companies is to enhance the patient experience, which accounts for 55 percent of the use cases. Common applications in this category include call steering, queue callbacks, short message service (SMS) follow-ups, and chatbots. Security, making up 16 percent of the use cases, is essential for tasks like identity authentication, activating and deactivating users, and password resets.
Revenue generation, comprising 13 percent of the use cases, facilitates scheduling appointments or rides, prescription refills, order processing, and bill payments. Lastly, reducing administrative work accounts for 12 percent of the use cases, covering status checks, frequently asked questions (FAQs), and information gathering.
For smaller providers, the scheduling aspect is a massive issue. In interviews with regional hospitals, clinics and other smaller medical facilities, humans having to call and confirm appointments is time consuming, error prone and creates many missed appointments every week. This creates a gap where a clinician could be seeing another patient had they been aware of an opening causing the facility to miss out on revenue. Over time, I believe scheduling and call backs will be one of the top “low hanging fruit” use cases for AI in healthcare.
The priorities are different for large providers compared to small clinics. Small clinics, which are often overloaded, prioritize queue callbacks (66 percent) and focus on personalization. In contrast, large providers, with more departments to transfer calls to, also seek personalization but have complex workflows. Large providers increasingly use IVAs for authentication, with call steering (35 percent), appointment scheduling (20 percent), and ID/authentication (15 percent) being major priorities.
There are also significant differences between healthcare providers and insurance payers. While both groups prioritize call steering and authentication, they face unique operational challenges and business pressures. For example, insurance payers deal with issues like annual member churn and seasonal work demands. They rely on IVAs to improve the member experience. They’ve also started using IVAs in new ways to train agents by simulating difficult customer interactions, known as “angry IVAs.”
Key Steps and Considerations for Deploying IVAs
Deploying an IVA requires thorough planning and coordination. Five9 suggests defining clear goals and success metrics, considering input from all stakeholders. Companies should identify the skills and resources needed, including technical expertise and integration with existing systems like databases and health records. They should also focus on essential functionalities like call steering and language support.
Five9 recommends implementing projects in well-defined phases. This includes seeking guidance from IVA vendors who have experience with similar deployments. AI, especially generative AI, can enhance an IVA’s capabilities. But it’s important to be mindful of privacy and compliance issues. Finally, companies should estimate the project timeline, considering team preparedness, feature complexity, and integration challenges.
My Take on Designing, Building, and Testing IVAs
At a high level, I agree with the recommendations from Five9 although trying to gather KPIs from multiple stakeholders can be a nightmare as everyone has different metrics on which they’re measured. But that should still be the overall goal.
The focus on success metrics can also be nebulous as we are so early in the AI cycle that there are few best practices to define success. As an example, is an error rate of 5% with IVAs “success?” Some might look at it and wait for perfection – but that may never come.
My recommendation is to understand what the human error rate is and then once the AI crosses that, put it into production. I’ve talked to organizations that strive for perfection only to have their testing phase drag on and on and that helps no one.
During the design phase, companies should focus on the end users. Listening to user calls can help identify common issues. I also recommend creating interaction flows and a logical diagram to map out user interactions. Companies should plan for various user scenarios and make sure the system is easy to update and maintain.
In the build and test phase, an obvious but critical step is to have the right tools to implement the IVA. Companies should be running functional tests to guarantee that the system works as intended. It’s advisable to arrange for vendor training and user acceptance testing, so admins can manage and fine-tune the system. Ultimately, the IVA will require regular monitoring and updating to ensure it learns correctly.
Summary
IVAs hold plenty of potential to help transform customer experience but present a massive risk-reward. Do it right, gain share, add revenue, and have happier patients. Do it wrong and that will drive customers away. An interesting factoid from my research is that 22% of consumers will switch brands because of a single, bad experience. With healthcare, that number is likely lower as switching doctors can be a headache but too many negative interactions will eventually lead to churn. As companies embark on their IVA journey it’s important to follow the best practices, vendor and partner expertise to navigate the complexities of deploying IVAs.
