BRIAN KENNY: Welcome to Cold Call, the podcast where we dive deep into the groundbreaking ideas in Harvard Business School case studies with the faculty who wrote them. In healthcare, some of the most important innovations aren’t new drugs or breakthrough devices. They’re new ways of giving clinicians back the one resource they can never make more of, time. As artificial intelligence rapidly finds its way into hospitals and medical practices, leaders face a difficult balancing act. Can technology reduce administrative burdens without creating new risks? Will it strengthen the relationship between physicians and patients or subtly change it? And how should organizations measure success when the benefits extend beyond efficiency to include trust, wellbeing and the quality of care?
Today’s case explores a real-world effort to answer those questions inside one of the nation’s leading academic health systems. We’ll examine the organizational challenges that accompany technological change and the difficult trade-offs involved in evaluating investments whose returns are both financial and deeply human. My guest today is Professor Susanna Gallani, and we’ll discuss her case, “The AI Scribe: Enhancing Physician Presence and Curbing Burnout at Mass General Brigham.”
I’m your host, Brian Kenny, and you’re listening to Cold Call on the HBR Podcast Network.
Susanna Gallani studies performance management systems and explores the interplay between monetary and non-monetary incentives, and you probably study much more than that too, Susanna. Welcome back to Cold Call.
SUSANNA GALLANI: Thank you. Thank you for having me. It’s good to be back.
BRIAN KENNY: Yeah, it’s been a little while, so we’re happy to have you back on the show, and we were just chatting before we rolled tape here, and every conversation these days seems to turn to AI. It’s just top of mind for everybody, to the point maybe of AI fatigue, but there are some real-world implications for the case that you wrote and for how Mass General is trying to use AI to help physicians. So, thanks for writing it. Thanks for being here to talk about it.
SUSANNA GALLANI: Sure. My pleasure.
BRIAN KENNY: Maybe we just get started by asking you, how did you hear about this? Why did you think this would be an interesting case to write?
SUSANNA GALLANI: So there’s a combination of things that are happening in this space. First of all, the case is written with Rob Huckman and Suraj Srinivasan and also an external co-author, Asaf Bitton, who is at Ariadne Labs, and of course the conversation about AI in healthcare is everywhere. Everybody talks about AI in healthcare and what we can do or cannot do with that, and we were looking for our course, “Transforming Healthcare Delivery,” which is a second-year course here in the MBA program. We were looking to introduce more of this AI content in our course, because as you can imagine, the evolution of AI in healthcare is a big transformational drive, and we still don’t know exactly what it’s going to do, but we want to be up with it as it evolves, and so Rob had heard about this implementation, and I had known about that system. MGB, which is the organization that we study in the case, is not the only one that is implementing this technique and this technology. It’s pretty much everywhere right now, but MGB was doing it at a large scale, and of course, it’s a leading academic medical center. So it was interesting for us from a number of perspectives, because it layers up, in addition to the delivery of healthcare, there’s also the training and the research part, and so all of these aspects kind of converge into the problems that we address in the case.
BRIAN KENNY: MGB being Mass General Brigham, just for our listeners who aren’t from the Boston area, it’s an enormous enterprise. I don’t know how many physicians they have on there.
SUSANNA GALLANI: About 12,000.
BRIAN KENNY: Okay. So just enormous. So this is an effort to scale something up in a very significant way, and I think oftentimes, when we think about AI in healthcare, we’re thinking about how it could accelerate cures or treatments of some kind. This, I found to be very interesting, because it was really about how they’re doing their work and how they can do their work more efficiently. Can you talk maybe a little bit about some of the challenges that clinicians face?
SUSANNA GALLANI: Yes. So in a way, I have heard this described as the first innovation in healthcare that is improving care without asking the physicians to do more, and I want to extend this. This is not just physicians, it’s clinicians in general, because there’s nurses and there’s advanced practice providers as well. So this is not just a solution for physicians, narrowly defined, but this is the first time that we’re trying to do more and better in healthcare while taking work off the physician’s plate, the clinician’s plate, which is important, because we know we have a burnout problem. We have a wellbeing problem in healthcare. We have a turnover problem.
If you listen to the first Friday of the month job report, healthcare jobs are always at the top of the list in terms of job creation, because there’s so much churn, and in addition, the needs for healthcare are increasing. Populations are aging. During COVID, we’re still experiencing the tail end of the COVID effect where people didn’t take as good care of themselves as they should have or they would have in other times. So there’s a lot of demand of physician’s time.
In addition to that, we see that physicians don’t want to be physicians anymore, because it’s becoming a job that is away from their calling and it doesn’t let them really engage with the patients in the way that they would want, which is where this technology comes in, because it’s a way for them to really free themselves up from the monitor or the computer while they are sitting with the patient and really focus on the conversation like you and I are doing with each other.
Imagine that as I’m saying these things, you were typing up things on your computer. It would feel very unnatural to me to be talking to you, and the same thing is in the office visit, and think about the experience of a patient. I always say healthcare is one of the most human experience we can have, and having the physician look into your eyes or listening, actively listening to what you’re saying and engaging in your conversation instead of doing a lot of what we do in the classroom, which is absorb what you just said and just transfer it to another medium helps a lot in the relationship with the patient.
BRIAN KENNY: I hadn’t really thought about it until I read the case, but it occurred to me that my doctor does take a lot of notes while we’re having meetings, and that seems to me like a fairly recent thing, and I think it’s part of the administrative tasks of doctors have increased exponentially. We’ve got all these systems, the information needs to be in there. They’re trying to get better at sharing information across platforms so that your dermatologist can see what your primary care physician is doing and so on, but it reminds me a little bit of education. In elementary education, my wife is a teacher and the amount of administrative work that she has to do has increased quite a bit over the years. So it just sort of sparked that for me. Maybe you can describe what AI Scribe is and how it works.
SUSANNA GALLANI: Yeah. It’s actually very simple as a concept. Obviously very complicated to implement it or to develop it, but basically this is how it works. When you see the doctor, they will probably ask for your permission to record your conversation with them, and it will be through an app on their phone. There’s nothing magical. There’s no invasive technology of any kind.
BRIAN KENNY: They just put the phone on the table.
SUSANNA GALLANI: Just put the phone on the table, and if you agree to that, that conversation will be recorded, and the recording will be decoded by an AI engine behind the scenes, and it will automatically populate the notes that the doctor would otherwise take during the visit. That’s all it is in a nutshell. Now, this unlocks a number of other things. First of all, there’s billing attached to that. So as the AI decodes what was said in the interaction, they also understand that the AI understands or identifies services that are being provided and can attach a billing code to it. So there’s a revenue generation aspect to it, which, by the way, the doctor would have to do anyway. So it’s not that we’re increasing the billing, not necessarily at least, and this is still a task that the doctor would have to do, and oftentimes, they end up doing after hours, the so-called “pajama time,” which is so dreadful, and it takes physicians away from their family time.
BRIAN KENNY: So is it agentic in the sense that it just does it automatically? It takes agency?
SUSANNA GALLANI: So the agentic part is what happens after with the data. At the moment, the core of this technology is just simply recording, transcribing, and interpreting in a way, in a way that you can create billing codes. The other thing is prescriptions. So if there is a drug that is being discussed in the interaction, the AI identifies that drug and puts it as a possibility for the prescription that the doctor might want to sign off. Now importantly, none of this is set in stone until the doctor actually signs the notes. So there’s still a requirement for the physician to go through the note and make sure that everything has been done correctly and approve it, otherwise the process is not complete. Shortly after, one of the main safety measures of this technology is that the recording is completely deleted.
BRIAN KENNY: Okay, but the notes exist. The notes remain.
SUSANNA GALLANI: The notes remain in the EHR of the organization that implements…
BRIAN KENNY: Who was it at Mass General Brigham that decided they should try to do this? Were they trying to solve a particular problem or were they just aware? Was it the burnout that you mentioned?
SUSANNA GALLANI: Yes. Predominantly that was the driving force. In fact, they actually stated very clearly in one of our interviews that that was exactly the reason. The main reason why they wanted to try this technology was to reduce the burnout among their clinicians.
BRIAN KENNY: What were the sort of organizational challenges that they had as they tried to implement this? How was it received by the physicians?
SUSANNA GALLANI: Variation as you can imagine. So initially they piloted with about 800 people that were self-selected, and so that went quite well. Although it did spark some controversy and some concerns. There are concerns on all sides. The physician, for example, sometimes is uncomfortable because they have to actually say more things out loud.
BRIAN KENNY: Interesting.
SUSANNA GALLANI: Because the AI will not transcribe something that it doesn’t hear. So it’s important that the physician kind of verbalizes their thinking, the drugs that we just mentioned. So if I don’t say, “I’m thinking about prescribing you whatever milligrams of this particular drug,” the AI has no idea that you’re thinking that. So in order for it to do the work for you, you have to actually verbalize actively during the visit, which is very different.
BRIAN KENNY: Could be better for the patient, though, because you’re hearing a lot more than you might otherwise hear.
SUSANNA GALLANI: It could, to the extent that the physician uses accessible language, and that’s one of the barriers that now the patient might actually be sitting next to somebody who’s speaking gibberish from what they can understand, because they are using technical terms and classification of stages of disease and things that might be obscure or maybe somewhat terrifying for the patient that doesn’t understand what’s being said.
The other pushback was, “Well, when I read this note, this is not how I would write the note.” So there’s a level of comfort and acceptance and identity that is kind of missing when the note is automatically generated by a machine, and these notes are there to be read later on, and so if the physician doesn’t recognize their code, for lack of a better term, they resist to that. Say, “Well, this is not the information I would want to have in the note.
Of course, you can imagine some physicians said there’s too much information, because the fact that Mrs. Jones recently lost her husband and feels lonely might be very significant for the primary care physician or for somebody in behavioral health, but maybe for the surgeon that is doing their appendectomy, is not important. So how do we tailor the amount of additional information that is non-clinical in nature? There’s variation in how much people desire to have in the note. So there’s a number of those additional challenges, and plus there’s change. There’s change and there’s trust that needs to happen. So the physician also doesn’t know what this information is going to do, and so there’s a little bit of resistance upfront. From the patient’s perspective, the most prominent type of challenge is accepting that that conversation is being recorded. So where is my data going? What is this data going to be used for? Is it still my data? Do I feel comfortable speaking to you naturally when I know it’s being recorded?
BRIAN KENNY: Yeah, yeah, and I could see, with all the rules and regulations that we have in place to protect people’s information and identity, I could certainly understand that being a concern. There also might be the cynics out there who say that the hospitals are only doing this because it’s in their best self-interest from a legal perspective or a productivity perspective. Were there certain things that Mass General was sort of measuring as they went through this process? Sort of key metrics that they had in mind?
SUSANNA GALLANI: Yes. They measured productivity for sure. That was one, but mostly what they were most interested in was their burnout measures, and during this pilot program, they actually noticed a significant drop in burnout.
BRIAN KENNY: Really?
SUSANNA GALLANI: People that had embraced this technology felt that they had more time for themselves, that they didn’t feel they had to be on their electronic medical records the whole time during the visit, that they could engage with the patient, but also their family time was freer and more under their control.
BRIAN KENNY: Yeah.
SUSANNA GALLANI: So they definitely noticed an improvement there.
BRIAN KENNY: Okay, and are they also using it…? A lot of people are doing telehealth visits now. That’s a way that people are seeing their doctors without having to go in person. Do they still use the technology even in a telehealth type?
SUSANNA GALLANI: That was not in the case, but it would be very difficult for me to imagine that that doesn’t happen, because it’s even easier. We have recordings of our Zoom meetings, so why wouldn’t we use this in a telehealth situation?
BRIAN KENNY: Were the physicians at all… And this may not have come up in the case either, but it makes me wonder if physicians were concerned that it would be used for review purposes or my performance is being gauged by what’s coming out in these notes.
SUSANNA GALLANI: So the notes are not technically part of the performance evaluations. They are for the trainees, which is another important piece of the case, but the concern that the physicians voiced with us was, “You are now giving me back time. What are you going to ask me to do with that time?”
BRIAN KENNY: Right. There’s the cynical piece I was trying to get out before.
SUSANNA GALLANI: That was the cynical piece, exactly, and that is a valid concern, because in the case, we describe one physician remembered when they were given human scribes, but this human scribe had a cost, and so they had to produce more, so see more patients, in order to pay for the human scribe to be assigned to them, and so the same type of concern was underlying this implementation saying, “Yeah, that’s great. Now you’re giving me back time, but how are you going to take that time back in other ways?”, and so if we end up doing this, obviously the purpose of reducing burnout will just go out the window.
BRIAN KENNY: Yeah. You want to try and create some more life balance for these people who are working crazy hours all the time. You mentioned there were some early adopters, the people who sort of signed up for it voluntarily. What happened when they tried to do a more general rollout of it? Were there people who were refusing to use it?
SUSANNA GALLANI: Yes. First of all, it was voluntary. So nobody is compelled to use it, but there is some sort of an expectation, because the second order effect, one of the second order effects of this technology is that it generates better data, and so imagine anything you want to do with clinical data, whether it’s research, whether it’s improving care, you want to have enough adoption, enough mass, to have that mass of data so that it’s consistent across an organization. That would make sense. Well, when they open it up for a larger group, which they still haven’t rolled it out to everyone, but they’re progressively getting there, one of the largest groups, which is the group that historically has the highest level of burnout is primary care. Once they opened it up to that group, it was opt-in, but for those people that chose to use the license, many of them, a significant amount did not pursue it. So they had the license, they just didn’t do it, and in fact, there was a rule that if you didn’t use it for three months, they would ask you why and potentially take it away, because it has a cost. There is a monthly fee for this license, which is person by person.
BRIAN KENNY: Yeah. I mean, we’re finding that the human side of AI, the getting people to adopt the tools and to use them, is highly variable. Some people are comfortable doing it. They think, “Oh, this is great. This is going to make my life easier.” Other people are thinking, “This is something new I have to do. I have to learn how to use this and what if it breaks or what if I don’t do it right?” What would you say to people who are listening, who are maybe going through the same kind of a process in their own organization? What are the sorts of things that you have to be thoughtful about as you’re trying to scale up this kind of platform?
SUSANNA GALLANI: The first thing is trust. We talked about it. It’s really difficult for people to just embrace this. When this technology changes daily, it is, from a user’s perspective, an enormous black box. You put things into it, and you don’t really know what happens to those things. Things being information, sometimes sensitive, sometimes personal, and so having this expectation that everybody will jump at the opportunity to use it is probably misguided.
BRIAN KENNY: Yes.
SUSANNA GALLANI: So understanding what do you need to do to make it so that people trust what you’re going to do with that technology. Generally speaking, security of that technology, it is a black box for the leadership too. It’s not that we understand this infinitely well.
BRIAN KENNY: We’re all learning. Yeah.
SUSANNA GALLANI: Yes, and it’s changing so much that we tend to understand less and less instead of more and more in certain ways of what it does and what it’s capable of doing.
So the first thing would be trust. The second thing would be use it to bring up the humanity of the workers and not to replace them, because there’s a huge fear. Now, this is less prominent in healthcare, I will say, than other industries, but we all are thinking about, “Is this thing going to take my job?”. So understanding that that is a concern that people have, even if it is buried in the back of your psychology, but especially if you’re bringing this technology in to help people—make it help people instead of making them afraid of it.
BRIAN KENNY: We’ve done a couple of other cases that are similar in nature to this one. One was Pernod, the spirit company, and they were rolling out a platform, an AI platform, to their sales organization, and the benefit that they touted was essentially the same thing, which is, “You don’t need to worry about what’s going on underneath the hood with your customers. We’re giving you an opportunity to really develop a relationship with your customer. We’re giving you information that’s going to be helpful to them and help your relationship with them,” and this is reminding me of that, but even at that, there were a lot of salespeople who felt like it was going to cramp their style and people were going to be understanding their customers better than they were and then they wouldn’t need the salesperson anymore. It sounds like Mass General is really just trying to help these clinicians free them up of some of the tasks that they’ve had to do that are not patient focused, and really give them an opportunity to focus on the patient, which sounds like a win-win.
SUSANNA GALLANI: Yes. It’s exactly that.
BRIAN KENNY: Yeah. As we think about this particular case, which is sort of at the intersection of technology operations and management control, are there lessons that you can think of that this could offer more broadly outside of a healthcare setting, let’s say, like we were just talking about, who are considering AI for sort of thought intensive work, knowledge intensive work?
SUSANNA GALLANI: Yeah, for sure. The fact that this is happening in healthcare is just the context. It is, as you were mentioning, happening in many other industries. So it’s not specific to healthcare. Obviously in healthcare, there is maybe a specific type of risk because we’re talking about healthcare data or healthcare information, which is a little bit different than maybe your purchasing information, which it still has to be treated with respect and the appropriate safety precautions, but I think healthcare is one level deeper than that. So in terms of generalizable lessons, well, one thing that is clear to me is that we don’t have to jump into this all at once, and especially when we’re not ready. This technology is changing way too fast to make it profitable or valuable if we jump in when we’re not ready, because by the time we get the organization to be ready, then that technology will be obsolete.
So one of the things that they are pondering at Mass General Brigham, but I’m sure in other organizations too, is when is the right time to introduce this technology, and for whom? I mentioned the trainees before. So this, again, it’s a problem that we see in other industries as well. I actually had a recent conversation with a legal firm in England that is thinking about introducing AI for their work, but the problem they share with Mass General Brigham is, what are you going to do with the trainees? When you have a junior partner or an associate in a law firm, as well as a resident at Mass General, are you going to teach them to use AI or are you going to be AI independent, should that technology go away or not work? Or what are the opportunities that we’re missing in not training our trainees to use AI, but at the same time, what opportunities are we missing if we train them just to use AI? So either way, you have to consider the pros and cons of either choice.
BRIAN KENNY: Right, and we know that one of the ways that AI learns is by consuming mass amounts of information. So it almost seems like when you scale this up, you really do need to get a high percentage of adoption, or you’re not going to get the full benefit of what the AI is capable of doing. It’s not going to learn as much as it could, and not as fast. So it’s pretty complicated to think about all the implications of that.
This has been a great conversation, always is when you’re on the show. So thank you for being here. Just one last question would be, if you think about maybe one thing that our listeners should take away from this case, what would it be?
SUSANNA GALLANI: I go back to the point I made earlier, which is think about the human that is using the technology, and use the technology to solve a problem that the human has before thinking that it’s just going to be great.
BRIAN KENNY: Yeah.
SUSANNA GALLANI: We have to really be specific of what problem we’re trying to solve, and I think that Mass General Brigham did a great job in really pinpointing what was the purpose of that implementation. It was not just about… Certainly, they’re not unhappy to improve productivity, but the purpose, the driver of the decision was we have to reduce burnout, and that was their priority.
BRIAN KENNY: Yeah, that’s great. Susanna, thank you for joining me on Cold Call.
SUSANNA GALLANI: Thank you. Thank you for having me.
BRIAN KENNY: If you enjoy Cold Call, you might like our other podcasts: Climate Rising, Coaching Real Leaders, IdeaCast, Managing the Future of Work, Skydeck, and Think Big, Buy Small. Find them wherever you get your podcasts.
If you have any suggestions or just want to say hello, we want to hear from you. Email us at coldcall@hbs.edu. Thanks again for joining us. I’m your host BRIAN KENNY, and you’ve been listening to Cold Call, an official podcast of Harvard Business School and part of the HBR Podcast Network.
