Autumn-Kyoko Cushman — Fixing Broken Healthcare with AI and Human-Centered Design

Welcome to the Humans and AI podcast, powered by AI Frontier Network. I'm your host, Flora La Orga. Each episode we dive into the human story shaping artificial intelligence. Through the journeys of entrepreneurs, builders and leaders, we explore the insights, challenges and visions for the future of AI and humanity. Let's uncover the human side of AI together. Hi Ardun, how are you? I'm happy that you are here and thanks for joining the Humans and AI podcasts. So yeah, let's start with your journey, which actually I found it very interesting. You started your journey as a hospital courseman like so in the Navy. I found out such a big ship towards entrepreneurship and now be the founder and CEO of ShiftRx. So before going on to all that details, I would like to if you could share with us and take us back to that early experience and how that helped you to shape to. Yeah. The way that you see challenges into this healthcare system. Sure.

Thank you for having me floor. It is a pleasure. I always appreciate connecting with other female podcast hosts. I did start my career as a Navy corpsman. I was enlisted for almost five years stationed in Washington D.C. and the way that the corpsman title works is that you are expected to do everything within the scope of what the physician tells you to do. And so this role is super versatile. You spend time in EMT and paramedic lpn, you're able to do sutures, you get to get multi clinic experience. And it's not on the civilian healthcare side, but you still work with a lot of civilians. So I was stationed at Walter Reed in Bethesda, Maryland and I would say over half my colleagues were probably civilians and they, I largely had seen what they had experienced on the civilian healthcare system side and we still felt the ramifications of that on the military side as well. And so there's the same complexities when it comes to seeing patients, the workloads, the admin backlog, and not having enough time in the day to actually have patient and face to face interactions.

Okay, how did that take you to your journey to entrepreneurship and as well building a comp, building a startup with your best friend. For what I can understood, you're building this startup with your friends actually. So I find it very interesting that part too.

Yeah. So Leanne and I actually we met in high school originally when we were 16 and we just stayed in tandem in our career path. So she was a healthcare provider, she was a pharmacy technician at cvs. She went to school to be a pharmacist. She ended up burning out as well. And Then taking a job on the corporate side working at Evelyn Health and in medication therapy management. Just totally switched to the corporate side of things, and I did the same. So I joined the Navy as a way to pay for med school. And when I started working in the health care system, I was like, this is just so fundamentally broken. And so when I got out of the Navy, I was just looking to optimize my career honestly. And I ended up doing a degree in it. I spent some time at the NCI NIH working on precision medicine for the associate director and closely with engineering teams. I ended up developing a really cool, unique skill set. I was scouted to go to IBM's Watson Health through a merger and acquisition with Francisco Partners, which is a big PE firm. And I was working on, like, data and client success. So totally unrelated to being a healthcare provider, still in healthcare. And I didn't really think twice about being a healthcare provider again. But my mom got really sick. She got diagnosed with a very rare, aggressive form of brain cancer called astrocytoma. And being her patient advocate was very difficult. And seeing the complexities of navigating the healthcare system on the patient side put me in a position that I had not previously been in. And I think for me, that totally impacted how I look at healthcare, realizing that it's broken, realizing that there's an opportunity to fix it. And so it was like an all in all obsessive thing of this happened to my mom. This is happening to millions of patients daily, annually, weekly. How do we fix this? And so it was like, this is my life's calling. I need to work on something that I think is going to fix the healthcare provider shortage and pay them better and make sure that they have that patient FaceTime that is so critical to a patient's care journey. And Leanne honestly came to do like a wellness check on me. And she was like, hey, are you okay? You're saying you're going to quit your job and you're going to work on this thing? This is pretty crazy. And I was like, yes, and you should do it with me. And she was like, all right. And we both really understood that firsthand provider problem of burning out. And also from the patient journey side, Leanne lost her father to colon cancer when we were getting out of high school. So for both of us, this journey is deeply personal on the healthcare provider side, on the patient side. And the impact that we want to bring to the ecosystem is a shared value between us both.

Well, beautiful story. So before going to that part, because we will get into it and where you guys are building and what you're doing and everything. But before going into that, you described yourself as a first generation founder. What does that identity mean to you?

Yeah, so we're first time founders. I think we're the first in our families to do something at scale like this. I'm not a first generation entrepreneur. My stepdad is actually he was an entrepreneur while I was growing up. We're very close. I call him my father. And he was in real estate and he built one quite a large real estate business when I was growing up. And then he was wiped out in the 2008 financial crisis and ended up taking a getting a new job that was much more stable and not as likely to be impacted by economic downturn, which is why I chose healthcare to begin with. I think we are first time founders and first generation founders because we have lived the experiences that we're trying to make a huge impact on and this is like our call to do so. We didn't come from the ecosystem of startups, founders, VC capital. This is like we are very much the people that we want to solve this problem for. And I think that's what drives us and that resiliency. And I, I've thought actually about the future. If I'll end up having kids that will be founders, I hope that they are happier and more content than to choose this path.

I understand what you mean. It's tough. It's tough. Okay, so talking about ShiftRx, it's AI workforce management and scheduling software for pharmacies. So you talk about as well about disrupting medical workforce crisis. How did in your perspective, how AI help with that?

Yeah, so our hypothesis here started with there are a lot of healthcare providers that are burning out. That is not unique to pharmacy. But lots of healthcare providers are leaving their professions and they're leaving the clinical side entirely. They're not going to be healthcare providers anymore. That is bad. We can look at that statistic and objectively say that is not a good thing. So Leanne and I said how can we ensure that healthcare providers stay in their clinical roles and how can we make the ecosystem better to incentivize new people to become healthcare providers? And our hypothesis around that is the only way to do that is to raise their pay, give them better resources and better support, better working conditions. And if you're incentivizing new people to become healthcare providers, then you will start to fix that problem as well and incentivizing the people that are currently healthcare providers.

To stay.

So how do you do that? How do you raise the pay for healthcare providers? It's a really easy thing to say, raise the pay. Done. Well, a lot of these healthcare facilities, pharmacies, they're strapped when it comes to profit margins, right? So healthcare systems are typically operating on a 2 to 3% profit margin. And so you need to find money in the budget elsewhere in order to increase the pay for healthcare providers. And so our idea here is how can we do sweeping legacy vendor consolidation? Because they're utilizing tons of different vendors when it comes to workforce management platforms, they're largely doing things like scheduling very manually. So we're starting with this niche and this wedge of pharmacy because we know that they're doing scheduling manually, we know that they have staffing issues, we know that they have turnover. It was not nearly as competitor crowded. And so we said we'll start here, we'll automate something for them, which is scheduling. And this gives them time back in their day. Because right now the people that are doing scheduling within the pharmacies is typically your lead pharmacy technician, your pharmacy manager, your pharmacy director. These are high, highly important roles that you can't have turnover in and you want to retain those people. And if we can give them that time back, then it's time that's spent with patients in clinic training new roles and they shouldn't be spending all this time on admin overload load anyways. So we're automating some of these pieces and we're also trying to squeeze out some of these legacy vendors. So the hope here is that we will have automated a lot of this back office administrative work while also doing replacing some of these legacy vendors that exist and giving them a really frictionless experience. And we want to be the thing that powers all the back office for workforce operations for health care. And in doing so, we control the pricing dynamic and we can make it much more affordable than what these legacy vendors are charging. And so you can put that money back in the budget to pay healthcare providers more.

Okay, how do you balance for my understanding, it's what the way that you're using AI is more for the automation or scheduling and older admin until you start as repetitive task. But how do you balance AI and the human side of the healthcare?

Yeah, so the AI is not meant to replace clinical roles, but this AI hopefully will start to replace some of these non, non clinical roles that exist in healthcare or non clinical roles that are just burdened on the actual healthcare provider. So like I mentioned, you have lead Pharmacy techs, pharmacy managers, pharmacy directors, doing these admin jobs that are not part of their job description really, and they shouldn't have to be doing those. That is where AI is very key. Right. So it can learn on their scheduling habits, it can manage people. From the perspective of Sally needs to take pto, she can text Holly, who's our AI bot and a scheduler. I'm on pto, Holly takes her off the schedule. That is time saved of an interaction with a lead pharmacy tech, pharmacy manager, pharmacy director, and basically we're building an AI that should be an extension of their team. So it's like they have their own AI healthcare administrator, they don't need to employ one. It saves everyone on the team time and the goal is to have it cost less than it would be to employ an actual human. But the AI is not meant to replace any of these clinical roles. It's meant to help them practice at the top of their license.

Okay, and what are the limits of what AI being able to do in the next five years in healthcare?

I don't think it's there yet to replace these clinical roles. I know there are big companies, Hippocratic AI, open evidence, that are working on small and large language models with lots of peer reviewed research and very curated data sets. But what we found is that like the generic big foundational models, so OpenAI's ChatGPT Maestro Anthropic Squad, these are not ready to be rolled out and just totally take over patient provider interactions without a human in the loop supervision. Because a lot of times the treatments that they're prescribing, the course of actions that they're recommending, the diagnoses that they're giving is not correct, it's not repeatable, it's certainly not scalable and so it cannot be trusted. Right now, we have no way in this industry to prevent an LLM from hallucinating. And anyone that tells you that they can do that is lying. We have zero way to prevent an alarm from hallucinating right now. So what we're doing with this back office automation operations. If an LLM hallucinates and puts someone in the wrong schedule or the wrong scheduling slot, there's still a human in the loop who can make that change. But that is much less serious than if you let an AI fully interact with a patient. For you example, floor, you are not feeling well, you think you're talking to a doctor online via chat and it's an LLM and it's telling you there's always the joke in health care that you have a cold, you start googling it and WebMD is telling you that you're now dying of stage four lung cancer. That can't happen. And so until we have a way to correct for that, it cannot be trusted fully on its own without a human in the loop. And I don't think that will happen in the next five years. But I would love to be proven wrong because I'm all about the advancement of technology for whatever, whatever good or bad comes from it.

Yeah, that's a very good answer. Actually, I was thinking about the latest news that I've done over here. I don't know if it's true or not because nobody. You don't never, you don't know exactly if it's true or not. But I read, I think it was yesterday, something about a kid of 13 years old that took his life basically because ChatGPT said that basically recommend he, that he should just suicide. So I found a bit, bit too extreme and bit dangerous as well without having some like a human supervising what the kid is talking with. AI. Right. Since sometimes this can be a wrong answer, I guess it was definitely. That was a wrong answer.

There are a lot of ethical concerns when it comes to AI. I don't know if our government or any government is going to regulate quickly enough. We're really going to be left at the discrepancy of the big foundational models right now. But what you're referencing is AI psychosis and it's a largely emerging trend that is very concerning where these AI models are meant to be very affirming, largely agreeable. Even if you give it custom prompt instructions to not be agreeable, they will be. And so yeah, it can send people into psychosis and there's people that are forming actual human mimic like relationships with their AI. And people don't realize that every time you're interacting with it, it's a tokenization process, you are prompting it regardless if you look at it that way or not. And so if it's used to talking to you in this routine and it's largely agreeable, and you keep saying I would be better off because here are all the things that are terrible with my life, eventually you can get it to a point where it agrees with you. And that certainly can't happen because that's ludicrous.

So talking a little bit of safety in within AI and I will, since we are talking also about healthcare, what are the boundaries principles that for example, in ShiftRx you guys are applying to make sure that human needs of both patients and providers remain front and center.

So our platform does not interact with patients whatsoever. I don't really want to get into the patient side, not at least in the next five, 10 years, the foreseeable future. I think it's messy, it's complex. Right now, I don't see a way that AI is. You're able to prevent hallucinations. And so for me, it's I'm out. I don't want to touch that right now. On the back office side, we have a human in the loop all the time. And when it's creating schedules, it's making adjustments, it has rules in which it needs to escalate to a human. So let's say a provider's licensure is coming up to expire, and our Holly AI is pinging that provider saying, hey, floor, your HIPAA is about to be expired. You need to do it. Hey, floor. It's sending a reminder every day for three days. You're not responding. And it knows that your HIPAA needs to be renewed or you're going to be pulled off the schedule. And that's going to interrupt patient workflow at that point, just depending on the organization, because they can custom train these, but we have these certain standards of escalation points. It will contact the manager that is directly above that person and say, hey, I contacted Flor three times about renewing her hipaa. She's at risk of being pulled off the schedule. What are we going to do about it? I'm obviously limited in my capacity here. Please intervene. And so there are human escalation points along the way. Even when it comes time to publishing the schedule, there's a human in the loop that can take a look and make sure that it's good to go. We're obviously wanting to get it to a point where it can run the workforce, scheduling, payroll, onboarding, all of these things without any human interaction. But we will not do that until we are absolutely confident that it can make all those decisions without any mistakes.

Okay, so ShiftRx is using generative AI to build autonomous staffing agents. So do you have a real case study that you could share with us?

So we are in the process of doing case studies now. We have ShiftRx, which was the PRN marketplace. We have case studies from that. That is the staffing side of things. It is still running and operational with over 10,000 pharmacists and technicians and almost a thousand pharmacies across the U.S. we are now on to our next evolution. And Most of the team's focus is on Shift os, which is our AI operating system that's doing the scheduling, the credentialing, management, the licensure tracking. And we are just kicking off pilots with some very large enterprise customers which is super exciting and we will have case studies for that in the upcoming months. So we'll have to, we'll have to circle back. But right now we've got some outcomes that we are looking forward to and that is time saved. Among pharmacy teams, expected ROI and reduced medication errors, which goes back to that patient data set. Because right now in the US I think we're at over 2.5 million medication errors per year annually impacting patients. And so if we can make a dentin that'll be a great proof point for us.

Okay, thanks for your honesty. So what there is any it doesn't have to be famous, but someone who inspire you in the AI field on healthcare that is doing some great job that inspire you or help you to keep pushing through on this kind of challenging journey of entrepreneurship.

I don't know if I can think of anyone that's in AI healthcare immediately just because it's so new. But in healthcare in general, I think that there are a lot of principles and lessons to learn from Judy at epic. Some people really like her, some people do not. She has built a monopoly with Epic. They power like 70% or something of the healthcare systems in the US EHR, EMR records. She has built something that is truly incredible and she is still working even though she is in her 80s and they've had lawsuits and been broken up in certain ways by the government for having created some type of monopoly. But I think what they did there was truly innovative at the time. And I think that even though they are one of the legacy incumbents that I'm referring to that needs to be replaced it what they've been able to do from a largely bootstrapped operation, never having lost the customer. I would love to meet Judy and I have a ton of respect for her in the field, especially being a woman founder. And this is back in the 8, 70s and 80s so I can't imagine what what that uphill battle was like and she's still reigning queen of health care.

Great, thanks. I don't really know about her but it's good to. I will check it out. I found it. I found it interesting. I always like also talking about women founders. I found a bit it's as I don't know if you even have time to check my episodes of Most of my podcast episodes are the guests are men. But I know that this is not like a coincidence. In AI and tech, mostly it's mostly men work. So what is your thoughts about that? About the most of the founders are men and there's kind of women. We are not much in this field. What is your thoughts about that?

I think that there are fewer women in AI and tech, but the ones that I have met are incredible, truly incredible. There's Mira Moradi, was CTO of OpenAI and now she has started her own thing, Thinking Machines. Thinking Labs.

Machine.

Thinking Machine labs, I believe. And she's one of the very famous AI female founders. I think even though there are fewer women in this industry, it just makes it so that the ones of us that are want to show up 10x100x and set the precedent for other women who are thinking about entering this industry. I actually, I also think there are a lot more women that are in this AI tech industry than we know about because they're operating in other spaces that aren't just largely labeled AI. Like I have friends that are building consumer platforms, but they're largely AI powered and they're considered to be consumer founders, but really at their core they're AI founders. And so I think there is obviously the disparity between how many men and women are in tech in general. But I think we started to see those trends change and I think they'll continue to change as more women get interested in stem. And I think it's awesome. I'm super passionate about women being in stem and I always joke that I hope I have four girls and they're all in robotics.

That's so fun. Yeah, yeah, I agree with you. I think more and more as well. I see there are a lot of women when they're in this kind of position as being a founder, you empower yourself, right? Empower like you. You see yourself like empower. And I think that as well being around men, it's another kind of power as well. It's just not just you, but I can be one as well. Right? Not just you. So this is a bit of different topic later going to it. But yeah. Okay, thanks for sharing your insights. I think it's very interesting. How do you see Shift RL X sorry ShiftRx in the next five years? What is the evolution that you see?

I want us to be the all encompassing healthcare company that is powering back office operations across verticals. Payroll, onboarding, credentialing, management, licenser tracking, scheduling. I really think that we have so much room to grow. We have a very passionate team about fixing healthcare and building AI healthcare infrastructure, as boring as that sounds. But we are all very much powered by this mission and this is truly my life's work and I think we've got a lot of momentum and health care is hard, it's complex, it's slow, it's fundamentally broken and that's what makes it so exciting for me to work on and I think about the patients that we impact like my mom likely and that that makes this so worth it.

In your opinion, how challenging is to get investment from VCs or Angie investors in the healthcare field?

I don't think it's challenging. I think you just have to be determined. Show up, do the work, pitch lots of investors. If it's not a fit, cut your losses, move on. Time is your most precious resource and it's the only non renewable resource in your life. You cannot make more of it. And so if someone understands the vision and it clicks, great. If they don't, just move on onto the next. We actually, yeah, we did not have a super hard time fundraising after doing our pre seed for our seed and some additional funding. And so we've been largely lucky with fantastic investors. But it definitely takes determination. Fundraising sucks and it is time consuming. So you have to make sure that you're using that like ruthless prioritization that you do every day as a founder to your fundraising process as well. You are selling a piece of your company, you're selling a piece of your success, you are selling your vision for the future and you need to be able to get swift conviction from people on that. And so if you're having a super hard time fundraising, it sucks to get feedback on that. But pitch friends, watch yourself pitch and say what? What impact can I make here? What difference is can I make? It's something that is not. I feel like you should ignore most of the fundraising advice. You see too. It's totally dependent on you as a founder and what the vision is that you're selling. It's not a one size fits all.

Okay. Because I also hear a lot about I have friends that are founders in healthcare as well. But they always kind of oh I like because for what they want to do, they need funding. It's not for example if you're doing something related with marketing or another kind of. Yeah field. Right. That is easy. Easier to make money before even you don't really need into fundraising or raise money at all. But healthcare. I always heard from friends of mine, I'm not in the healthcare field, but I have friends that they are and they struggle a lot to. They don't struggle a lot, but they have to have funding otherwise they cannot even kick off the startup. That's where. Can I ask?

Yeah, it's hard healthcare, like I said, it's complex and it's slow. And most healthcare companies do require funding before, but it's not impossible. If you go listen to Judy Acquired did a fantastic episode on Judy. They were almost bootstrapped. I think they had a couple hundred thousand dollars when building this. It's easier now and cheaper now to build software than ever. But you definitely need the right amount of capital and the right tools because healthcare is a long, slow process to sell into.

Okay, so we are almost at the end of the interview, so I've got a few questions for you. What kind of healthcare work are you working toward? Not just in terms of technology, but also in terms of how, sorry I gotta say to myself, the same thing. But in terms of how care feels to patients and providers too.

I'm so sorry, I did not understand the question.

Yeah. So basically, what kind of healthcare do you. Are you working towards not only in a technology way, but as well in the caring way with the patients and providers? It's a bit like a vision of how would you like healthcare going towards in the future?

Yeah. So I think the way that healthcare will work and should work is that the clinical provider's only job is to spend time with the patients. So you arrive, your entire ehr emr, your patient history is synced, you see a nurse, then you see a doctor. If you need treatment, then you go see a pharmacist. But it is lots of FaceTime and their entire day should be high quality facetime with the patient and not sitting and typing on a keyboard while you're sitting in the doctor's office after you've waited for 40 minutes and you only get 10 minutes to talk to them. I think that there's so much backlog and admin stuff when it comes to even like billing and insurance that AI is really going to make a huge impact. And the way that health care can look in the next 10 years is unrecognizable. And it can be very high quality care for patients. It can be a high quality life for healthcare providers with better pay and better resources and better support. But we can't be the only players in this industry. There, there has to be more of us that are largely working towards fixing this system that is plagued with inefficiencies.

Thanks Aun. Thank you very much for your time. Thank you very much. I appreciate a lot you were able to make it to the Humans on a podcast.

Thanks so much Laura. Great to see you.

Thanks for listening to the Humans and AI podcast. We are here to spotlight the people making AI more human one story at a time. See you in the next episode.