Let’s say you woke up with a 102-degree fever, chills, and a small rash on your right shoulder.

Like so many of us, you would normally hop onto your computer to research the symptoms to make sure you don’t have some life-threatening disease. However, what if you could tap into an online 24/7 chatbot resource provided by your health plan or employer? You describe your symptoms — and instantly you receive an answer. Would you trust the recommendations provided by the chatbot, or would you prefer to get an actual doctor’s opinion? Most consumers today would likely prefer a human doctor.

But advancements in artificial intelligence (AI) technologies are pushing us closer to a world where technology will go beyond simply supporting care teams. Within a certain capacity, chatbots will become actual members of that care team. But before consumers can become comfortable with medical chatbots, the health care industry itself must be comfortable with and understand exactly where and how chatbots can provide value.

In other markets such as retail, chatbots have already taken hold. In 2016 alone, 1.8 billion interactions were exchanged between approximately 20,000 chatbots. Health care is the next industry that stands to benefit from this technology. We’ve seen speculation regarding how to apply chatbots to patient care, but so far it’s just that — speculation. Here’s what the industry should and shouldn’t expect in the coming year.

Chatbots won’t be replacing doctors anytime soon

The human touch will always be necessary in health care. But this doesn’t mean that health care systems won’t use chatbots as supplemental resources, given their ability to quickly process various types of data in high volumes. For example, chatbots can function as a triage service for patients who want to share their symptoms instantly without having to wait for a nurse or provider to get on the phone. By the time a live medical provider does connect with the patient, the context will already have been made available to them, which can accelerate treatment.

Depending on their use, chatbots can provide tremendous value for patients, providers, and health plans. However, technologists need to make sure that chatbots don’t make diagnoses without human intervention; that can become very risky, very quickly. Also, the human service layer addresses the health care needs that chatbots cannot, like reaching those patients who will be most receptive to a live conversation with another person.

Call centers will remain, but responsibilities may shift

Chatbots won’t completely take over the customer service representative role either. However, we will begin to see a shift in job responsibilities. For example, call center teams, which health care organizations spend millions of dollars on, can provide more value by addressing callers’ more complicated questions that require human touch and sophistication.

Straightforward questions like “is my knee surgery covered under my plan” or “is my provider in the network” can easily be answered by chatbots before they reach a live agent. But complex questions like “what changes can I make for my care management plan to be more effective” should be handled by a live person. By offloading basic and mostly generic questions to machines, health plans, providers, and employers can focus more on training customer service teams to create more meaningful conversations with consumers.

Also, remember that for chatbots to be functional, the technology will always need some form of training. We need people to do that training. Because they are on the front lines and have the most interaction with consumers, call center teams represent one of the best resources for training and testing these technologies.

Only implement chatbots if you can do it right

Companies should start leveraging chatbots to help streamline the consumer experience and better utilize their resources, but there are a series of considerations and initial steps they must take first.

First, decide what the scope of knowledge will be. Chatbots are supposed to make things easier for the user, so focus on having an exceptional user experience that doesn’t cause frustration. Implementation of proper responses will be key, along with having good linguistic and conversational approaches. An effective chatbot needs to be able to know you, understand what you need, and engage you in a natural conversation while providing responses, guidance to resources, and personalized and intelligent recommendations. Chatbots also need to know what they don’t know. The most effective chatbots can provide deep knowledge responses to some questions and be smart enough to guide the user to where they can find an accurate response to others.

Next, to launch an effective chatbot resource, health care organizations should also:

  • Utilize leading edge technologies and machine learning that gets smarter over time as it receives more data and information.
  • Make sure the technology has the right level of personality, tone, and sentiment, and can understand context.
  • Consider partnering with a third-party vendor with a supplemental database to help create rich content.

Finally, for chatbots to be successful, you need to leverage existing data. Organizations should have a large set of commonly asked questions based on data from online forms, emails, and call centers. It’s important to look at past consumer interactions, along with input from agents on the front lines, to assess the types of comments to expect and to understand how questions are asked in everyday language. The more data organizations have to start with, the more robust their implementations will be. Over time, as more and more consumers interact with them, chatbots will become smarter in answering questions and evaluating conversations. The formula is simple — the more you feed it, the more its vocabulary grows and the more capable it becomes.

Health care organizations are often perceived as technology laggards, but now they have the potential to lead the pack in applying AI. Companies like IBM Watson have demonstrated value in the fields of oncology and clinical trial placement. On the insurer front, companies are leveraging AI for on-demand support for health plan members. In addition to increasing member satisfaction and efficiencies, AI is also setting the stage for more meaningful experiences when a live agent is required. Supplementing human interactions with bots can extend customer service teams’ knowledge and reach. So, while chatbots may sound futuristic, they are just a natural progression to improve how the industry provides care and interacts with consumers.

Jeff Cohen is the cofounder and vice president of cognitive innovation services for Welltok, a population health services company.

Above: The Machine Intelligence Landscape. This article is part of our Artificial Intelligence series. You can download a high-resolution version of the landscape featuring 288 companies by clicking the image.