Before COVID-19, healthcare worker burnout had reached crisis levels, research from the National Academies of Medicine shows. More than 330,000 U.S. healthcare professionals left the industry in 2021, with many remaining workers considering leaving, reported commercial intelligence company Definitive Healthcare.

For healthcare organizations, burnout leads to decreased productivity and increased recruitment and training costs. This also the problem that a healthcare organization can’t deliver a superior patient experience if employees are burned out and dissatisfied.

In today’s hypercompetitive market, where retail disruptors such as CVS Health and Walmart Health are leveraging strong consumer-centric business models, substandard patient and member experiences will cost payers and providers in reduced revenue and diminished market share.

Patty Hayward is general manager of healthcare and life sciences at Talkdesk, a healthcare customer-service technology and services company. She points to generative AI as one tool to help solve some of these problems, specifically ChatGPT.

Talkdesk is using ChatGPT now to make healthcare contact center agents more efficient and effective, she said. Combining AI with automation eases the agent burden by removing tasks from their plates and helps create a frictionless experience for patients, she added.

We interviewed Hayward to talk about ChatGPT in healthcare, how Talkdesk as an early adopter has incorporated ChatGPT, the benefits of this generative AI tool, and where healthcare organizations must be cautious.

Q. What is it about ChatGPT that makes it particularly interesting to your corner of the health IT industry?

A. We all know there are challenges for generative AI and ChatGPT in clinical settings today. There are accuracy issues, and the “hallucinations” that large language models can have make the idea of a “ChatGPT Doctor” really far from today’s reality, and maybe an impossibility.

In healthcare, we already have seen AI works best when supporting human beings, not necessarily replacing their decision-making and training. But even with decision support, there are patient privacy issues and lack of governance that health IT leaders are currently navigating.

What is interesting about our area – the contact center – is it is a natural space to both experiment with generative AI and give healthcare organizations real value on day one while doing so. Contact centers deal with a lot of volume, and much of it is highly transactional – checking on a claim, changing an appointment or requesting a prescription refill.

Healthcare has been trying to automate self-service in those sorts of areas for years, and generative AI gives us an opportunity to deliver more complex and relevant answers for patients, and free up staff time for more complex interactions – because the conversations on the other end of the spectrum in healthcare are highly complex and require a lot of knowledge and empathy from staff.

A poor contact center experience with a retailer can be frustrating, but one in healthcare can impact health outcomes, care plan adherence, and even be traumatic to patients or members calling or chatting in a moment of real need. So generative AI can act as a copilot for staff to get answers, and can be trained to offer suggestions on tone and wording to ensure staff that might be new – or who might just be overwhelmed – have the support needed.

Making these interactions more efficient, accurate and empathetic helps patients, and it helps the organization.

Q. You have incorporated ChatGPT into your contact center workflow. What do agents do differently?

A. Generative AI is at its best when it’s taking a specific input like a transcript or a draft of a message and using its knowledge of language to reformat it in a different output. One of the most delightful and entertaining parts of ChatGPT can be asking it to take a concept and write a song or a poem about it.

In a healthcare contact center setting, there are actually a lot of places where we need to manipulate written information – summarizing, synthesizing and personalizing it. Generative AI is really helping us there.

The first place we started was actually after the call or chat ends – agents have to do work after the interaction to submit their notes and enter a disposition for the patient or member. We have supported live transcription of calls and chats for years, so now we give them a one-click summarization of any call they finish.

The agent instantly gets a uniform summary they can edit if needed or send instantly over to the system of record with another click. Our AI can also recommend a disposition for the caller based on what was said. It’s already saving agents really valuable after-call time.

We also are using AI to help with the management and administration of healthcare contact centers. Even in smaller provider or payer organizations, you can have a lot of complexity around facilities, departments, teams and similarly complex flows for a patient or member to go through to get to the best person or resource.

So, we are using generative AI to dynamically recommend templates and support content for administrators who want to make updates without having to either read lots of technical support documentation or get in touch with our team. We’re also using generative AI to suggest chat responses based on context and knowledge base articles.

In all cases, an overarching theme is using generative AI to turn staff into experts instantly – whether it’s during or after an interaction with a consumer or in administering the contact center.

Q. How does ChatGPT help? What are the benefits of using this type of AI in this way?

A. Generative AI can go beyond rules-based configurations and static sets of variables to present personalized answers and information to consumers and staff in the best way. For example, most self-service flows today start with taking a set of data, like patients’ names and the dates and times of appointments, and building different templates for presenting to them.

This can take time to set up and optimize, and often results in patients having to “learn” the bot: They need to ask questions in specific ways and enter information in a particular order. With generative AI, the same data can be used to build a relevant, personalized patient answer immediately and without staff having to manage large numbers of templates.

For patients, it means we can deliver on the promise of true natural language understanding and that answers to their questions can be truly personalized to them – their information, health-literacy level and their entire history of interactions with the healthcare organization.

Patients want to feel known and cared for at every step of the journey, not just when face-to-face with a clinician. Healthcare organizations too often deliver impersonal and formulaic experiences and expect patients to self-navigate and constantly re-explain their needs, despite all the insight and information that they have gathered from previous interactions.

Generative AI can translate that information for staff or self-service instantly, so that it can finally be leveraged in conversations with patients to save them time and frustration.

Q. What do you have to be careful about when using ChatGPT? Many experts in the industry have concerns about AI, for example, its accuracy or biases.

A. As I mentioned earlier, the current accuracy issues plaguing ChatGPT are a giant red flag in the clinical realm, and rightly so. This technology is evolving rapidly, and I expect many of the accuracy issues to diminish in the next few years, or even sooner.

However, the technology itself already is getting ahead of ethical and moral conversations regarding its use. Are there inherent biases in ChatGPT? Does generative AI add to data security and patient privacy concerns? It will be interesting to see, if and when a regulatory framework emerges.

Generative AI algorithms have their limitations, but they will only get better and more accurate. Still, there’s no point in delaying implementation of transformative technology just because it hasn’t attained perfection. Healthcare organizations can put ChatGPT to excellent use today as a tool to improve the patient/member experience while increasing operational efficiency and reducing contact-center agent burnout.

Follow Bill's HIT coverage on LinkedIn: Bill Siwicki
Email him: bsiwicki@himss.org
Healthcare IT News is a HIMSS Media publication.