Dr. Petra Harms, veterinarian and founder of VetMaite, joins Dr. Andy Roark to tackle one of the most overlooked questions in veterinary AI: how do we know these tools actually work? AI scribes, chatbots and diagnostic aids are flooding into veterinary clinics, but who is testing their accuracy, safeguarding client data, or setting ethical standards? Petra is a big fan of AI (she’s not calling for chisels and stone tablets) but argues that without proper quality control, veterinarians and clients could pay the price. In this thought-provoking episode, you’ll hear why “data is the new gold,” what every practice owner should ask before buying an AI tool, and how to embrace technology responsibly without sacrificing patient care or privacy. If you’re excited about AI but also a little uneasy, this is the conversation you didn’t know you needed. Let’s get into this episode!
You can also listen to this episode on Apple Podcasts, Amazon Music, Spotify, YouTube or wherever you get your podcasts!
LINKS
Dr. Petra Harms on LinkedIn: https://www.linkedin.com/in/petra-kos-harms/
VetMaite: https://www.vetmaite.com/about
Dr. Know-It-All Card Game: https://drandyroark.com/product/dr-know-it-all-card-game/
Dr. Andy Roark Charming the Angry Client Team Training Course: https://drandyroark.com/charming-the-angry-client/
Dr. Andy Roark Swag: drandyroark.com/shop: http:/
All Links: http://linktr.ee/DrAndyRoark
ABOUT OUR GUEST
Dr. Harms works at the intersection of emerging artificial intelligence technology and practical small animal veterinary practice. She has spent more than two decades working in veterinary clinics, 15 of those as a veterinarian. She has practiced medicine through seismic shifts in the medical field, from the change from paper to electronic medical records and during the ongoing evolution of EMR’s.
She has practiced with fifteen veterinary medical clinics, taken part in three major technology transitions, one major facility transition, and is versed in five EMR systems. She has worked closely with a teams of highly trained emergency clinicians and specialists, and has also worked as solo practitioner in charge of day-to-day clinic operation.
In the past year, Dr Harms has focused her education on learning about Artificial Intelligence and its significance to healthcare systems, obtaining certificates of specialization from Stanford and Harvard with more on the way.
EPISODE TRANSCRIPT
Dr. Andy Roark: Welcome everybody to the Cone of Shame Veterinary podcast. I am your host, Dr. Andy Roark. I got a great one for you. I’m here today with Dr. Petra Harms and we are talking about AI governance guys. As Petra says, what is the quality control that we’re looking at for AI? What testing has been done on any of these AI products that we’re now using in our clinics?
We’re taking them in. We’re putting them to work, but do we know their accuracy? Do we, is there a set a standard of accuracy that we are okay with? Nothing is being enforced. People ask me, they say, Andy, you like using an AI scribe. Which ones are the best in terms of medical accuracy or reporting?
And I, I have no idea. It’s wide open and said, I found a lot of products that I really like and have really enjoyed using. I think Petra’s questions are really good. Also, I will tell you this, and I think this is important. Petra loves AI. She is not someone who’s saying, we shouldn’t do this.
This shouldn’t be the way forward. Let’s all go back to pencil and paper. Better yet, let’s carve our records into stone with a chisel. She’s not that person. She is an AI fan who is really thinking critically. How do we want to implement AI into our practices? How do we do this the right way the first time so that we uphold our medical standards so that we uphold our patient privacy and client privacy and just, anyway, a lot of really fascinating thoughts that I think this is the time to have them into discuss them.
And so anyway, I think you guys will really enjoy this. I really did. It’s a whole nother way of looking at AI enthusiasm and start to say, eh, let’s maybe pump the brakes a little bit here and start thinking about what success looks like and what quality control looks like. And so anyway, let’s get into this episode.
Kelsey Beth Carpenter: This is your show. We’re glad you’re here. We want to help you in your veterinary career. Welcome to the Cone of Shame with Dr. Andy Roark.
Dr. Andy Roark: Welcome to the podcast, Dr. Petra Harms. How are you?
Petra: I am doing very well, Andy. Thank you.
Dr. Andy Roark: I am so glad to have you here. You are one of those hidden gems in my life that I came across. I met you through LinkedIn, I think it was after. So I wrote an article about Three Things that I Think Are Most Interesting in Vet Medicine Today, and I wrote about mid-level practitioners. I wrote about what comes behind AI scribes, and we’ll talk about that a little more in a second. And then I wrote about a distributive model of education. So, vet schools farming out students to, to practices to get their education. And I think that was what I had written.
and you reached out through LinkedIn and said, Hey, I, I’d love to talk to you a little bit more about the AI scribes and what goes into that and what the future looks like there. And that’s when I became aware of you. And I’ve been reading your stuff since then. I’ve seen a lot of the conversations and things you have on LinkedIn. you are a veterinarian, you are a practicing veterinarian. You are in Vancouver Island, and you also are the founder of VetMaite, which is an AI literacy hub as you describe it. You and I have talked a couple times. You have, I just really like your ideas. It’s fascinating to me to have someone whose interest is like a self-proclaimed passion for AI governance and ethical AI use.
And I was like, what does that even mean? So I have really enjoyed diving into what that means. But lemme go ahead and start at a high level and why don’t you go ahead and lay it out. Petra, when you say that you are interested in ethical AI use or AI governance, explain what is that, first of all, and what problem does that solve?
Petra: Oh my goodness, that’s such a loaded question. Okay,
Dr. Andy Roark: but it’s really important. Like it’s, it sets the groundwork for everything else.
Petra: Oh, absolutely. So, I think everybody’s aware right now that artificial intelligence is making inroads in veterinary medicine, human medicine, the world. It’s happening everywhere. And I first became interested in it, just a couple years ago actually. And, and as a practicing veterinarian, it, it was a little bit on my periphery.
And the more I looked into it, these tools that were available, the more I realized that there’s so much enthusiasm for what artificial intelligence can do for veterinary medicine, for practitioners. But I also recognize that there’s actually very little by way of a responsible framework or a logical way of integrating this technology into practice safely. And I’m of the vintage, I suppose that, that I remember when regular practice, that was paper and pencil practice transitioned to electronic medical records.
And, I do remember the, I suppose the growing pains that happened when, when the last technological evolution happened in Vet Med with EHRs. So I saw this, this new opportunity coming up with artificial intelligence influencing veterinary medicine and I kind of figured that we can’t afford to make the same mistake twice. We can’t afford to take this opportunity to improve veterinary practice for veterinarians and squander it. And essentially, by implementing it without, without due consideration, without, making sure that the veterinarian is at the center of the focus, without, without making sure that the veterinarian’s concerns are heard.
And, and as veterinarians we’re, we’re scientists. We like to see data and I’ve, I wasn’t seeing a lot of data coming out of, out of this, this AI integration push, right? There’s, there’s very little by way of, performance metrics. There’s very little by way of test results that an AI technology provider can give to a buyer to say, this is how well this product performs.
And because artificial intelligence, because of what it is and how it’s developed, contains baked in, flaws that, that we still have to, as an industry, wrap our brains around, such as bias and hallucinations. Those flaws can cause errors and, and those errors can come back and bite us in the butt as veterinarians, they can, they can harm our patients.
So the idea of responsible and ethical integration of artificial intelligence into veterinary medicine comes from a recognition that, that their, right now it’s, it’s a very loosely in fact, unregulated industry. A recognition that, lacks approach to quality control is, potentially harmful to, to our veterinary, users but also from an approach of enthusiasm. I love the idea of using these tools to help veterinarians practice more efficiently. Get rid of a lot of the busy work that we flush down the toilet every day, in order to get our jobs done. It comes from a two-pronged approach.
I want this industry to succeed. I want the technology to succeed and responsible and ethical and integration is in my mind the path towards that happening rather than trying to fit things in ad hoc, without, without some kind of structure.
Dr. Andy Roark: Yeah, I think the reason that you’ve been so interesting for me is that you, here you are, and you’re talking very much about AI regulation, and governance and quality control and things like that. And it is so obvious in the things that you write that you are a big fan of AI.
You are not the person who, I don’t like this. I hate it. Kids get off my lawn. Take that thing with you. It makes too much noise like that. That’s not you at all. You’re like, I’m, I really, like this. And guys, we want to do it well.
Petra: Yes.
Dr. Andy Roark: Talk to me a little bit. I think, so let me lay down where I was coming from. So when I wrote the piece about the most interesting things in vet medicine, to me, the most interesting thing regarding AI is it’s not the use of the AI scribes necessarily. I’ve been really impressed with them. To your point, I think this is a great way to reduce workload, and, just the administrative burden. And, it’s just, I can sing the praise of an AI scribe. Again, I know, I also know that it’s early on. I have limited expectations in what it can do as far as I, everything needs to be checked, and I think it’ll always be that way, but I really am a big fan of these products.
The thing I found most interesting was when I went to the VMX conference in WVC conference early this year, one of the most common conversation that I had, just again and again, was about people who were like, all so you’re using AI scribes and all of this information is being taken in.
What do we do with that information, Andy? The opportunities are limitless and there are conversations around, and again I think a, they all came from good intent, I think, but it was things like, oh man, we could listen to the vets or I, listen to the veterinarians in the exam room. And make suggestions on how they might better communicate with the pet owners, or we could have the technicians use this and then we could do technician training or also doctor training and we can build out communication modules based on these things.
And we could also talk to individual doctors because we would know that they’re not recommending this products or they’re not doing this, or not having this pro conversation in an effective way. And, it just goes on and on to, we could not only listen to the conversations you’re having, but then we could look at what actually shows up on the invoice and say, you talked about this, but if, but pet owners are not following through on this recommendation.
Let’s look at the way you’re talking about the specific recommendation and part of me goes, that’s great. Another part of me goes, oh my God, we’re about to get micromanaged in a way that’s gonna make everybody just quit. Because it would be terrible. Again, it’s the implementation. If there’s another group that I’ve talked with, and this was, they talked about it a while ago. You could actually use AI scribes to measure the wellness of the team in the room. You start listening for burnout language. You listen for if they’re, if the doctor’s having steadily shorter and shorter conversations, you say, what does that mean? Is this person, is this person, are they feeling depressed?
Are they starting to burn out? Are they losing enthusiasm? are, they disengaging from their job? Is good or bad? I think the devil really is in the details, but I was shaken a little bit by thinking, oh my gosh, the amount of information that is about to be collected and be able to be processed in a way it never has been before. I don’t know. To me, when you talked about, AI regulation or best practices, AI governance, I thought, I can see you now the benefits of having a, a focused approach. So as I lay all those things down I wanted to see is your sort of head at as I lay that stuff down?
Are you in agreement that these things, that these are the things that, these are things, some of the things that are interesting. Are you, do you think I’m way off base with having some of these concerns? Let’s first start to unpack it.
Petra: Oh, Andy, you’re spot on. What you’ve brought up really brings to for the idea of bringing veterinarians into kind of a position of knowledge and empowerment with, with artificial intelligence. Veterinarians with the use of AI and these tools are important.
They’re critical to improving our quality of life in growth. Quality of practice, but they’re also changing the role of the veterinarian. We’re not just caregivers anymore. We’re not just doctors anymore. We are generators of data and we are generators of data in more ways than just what we now choose to type up.
We are generators of data in the sense that our conversations may be recorded. And, the data in those conversations is incredibly useful, right? as you said, you can, you can have, you can mine that data for not just things that you did or did not recommend, but, but even your tone of voice, you can mine the data for example, the actual interaction between the client and yourself and what the client says and how that influences what they are willing to purchase. And. This isn’t something new, right? You think about Alexa. When you have Alexa in your house and you’re, you’re asking Alexa for, or certain questions, all of a sudden you, you have that tendency for certain advertisements to pop up that you don’t remember requesting, right?
Dr. Andy Roark: Yeah
Dr. Petra Harms: So the mining of audio data is a real thing. We are now becoming data generators when it comes to these. These scribes, And it’s important for veterinarians to be aware of, of the potential for that data to be used. Be aware that they should be empowered to ask what their data is and will be used for, by these, by these companies.
Is there any, is there any possibility that their data will be saved and then mined for information down the line and potentially sold? Because as, as they say in, in AI, data is definitely the new gold. And there, there also should be a, a framework where, where in general in veterinary medicine, there, the concept of a responsible AI company or AI provider is accountable to the veterinarian and to the user of the, of the technology for, for the ethical collection or lack of collection of that data, right? and so at the moment we are counting on the AI technology providers to do the right thing, quote unquote.
And don’t get me wrong, there’s many and I’ve met. many, amazing startup founders that have all of the right, all of the right, ideas. As businesses go, businesses get sold, information gets transferred, without any kind of official framework as to what is ethical and right, and what is focused on the benefit of the veterinarian and the patient and client by extent.
Then. we are, we’re setting ourselves up for, for big cracks in, in, in our due diligence, for ourselves and for our clients.
Dr. Andy Roark: It’s funny as you say this as like we are generators of data think what would a pharma company pay to have to look for links between in these appointments your product ended up on the invoice like it got sold and we will search all of these appointments and find common or repeated phrases like what are the things that the veterinarian or the technician says In the appointment that turns into a sale of this pharma product versus the ones that don’t. And you go that people would pay for that they would absolutely love a breakdown of these are the things that happen when your product gets sold or gets purchased by the pet owner. And it’s just again I’m not trying to throw shade on pharma or anything but it’s just off the top of my head thinking about the value of that data. That’s not necessarily directly to the veterinarians I go to, oh that’s really interesting. Petra we are we’re looking at more and more corporate vet practices and multi-site vet clinics and things like that I I think about the value of that data that’s when you think about the individual exam room at that at scale Let’s say that you are one of the one of the practice groups that has a hundred vet hospitals. You have no way of being in the rooms and hearing what’s going on but you suddenly have this ability at scale to really look and start to break down performance in a way that’s well beyond what you used to be But I don’t know how the veterinarians are necessarily gonna feel about that You know what I mean It’s it I think it goes back again in my mind to how people use the data. But I got into vet medicine to support people and have relationships with pet owners and to be a real resource for them. I didn’t necessarily get into vet medicine to generate the maximum amount of money that I can generate. And I know that there are veterinarians that work for companies that totally get that. I know that there are veterinarians who work for companies that do not get that and they really are saying guys we need to 10 x our investment here. How do we do that? And so I guess my question for you is when you think about sort of regulation, Petra, are you mostly thinking at like a professional level of across the board these are the types of things that veterinary medicine should be thinking of and that would be things around the line of product accuracy things like that? Or is there some sort of deeper level of privacy that people should be able to expect even if they’re working for another company that is technically the purchaser of the scribe software?
Petra: I think what we should be looking for is really an approach from both directions. And I’ve used this, this, I suppose this comparison before, but, but AI isn’t the first time we’ve run into potentially dangerous tools that have been very useful. You look at,
Dr. Andy Roark: I think that’s a great way to that’s a great way to put at it is
Petra: Yeah.
Dr. Andy Roark: The internet is I remember concerns the internet is I don’t want people to know about me and I don’t want them to be able to email me and
Petra: Yep.
Dr. Andy Roark: Okay
Petra: Yeah, exactly. There’s the internet. and then let’s go down to something even more basic, right. Medications, drugs, and medications. We as veterinarians know that drugs and medications are incredibly helpful. They’re essential but we also know that they can carry the potential for harm.
And, and when you look at how, the veterinary industry regulates drugs and medications. There are multi-stage approaches to this regulation. Yes, there is top-down regulation in the sense that the veterinary medical industry has a strong kind of science-based push against things like inappropriate use of antibiotics, right?
Then you go down one step, you look at regulatory bodies and, and they enforce the safe, storage and documentation and prescription of controlled drugs. And then you also see though an individual responsibility that, that sits with the veterinarian and with the, with the actual practitioner to, to know the side effects of the medications that they are using and, and the, make, making sure that, that there are no, negative interactions with, with other drugs that, that the patient is.
On. So the allocation of responsibility, how you know, who you saddle with, the responsibility to do certain things with medications is tiered throughout the industry. And in my mind, the same thing needs to, needs to happen with artificial intelligence. We’re starting at, to get some, some movement from, from the upper levels.
In terms of the CVMA, coming up with a position statement, the AVMA has, has created an emergent tech technologies task force, to, to address kind of the overall overarching, messages about responsible use of AI.
And I think that’s going to start to trickle down through the levels of corporations hopefully paying attention and creating responsible use, frameworks for AI in their, in their companies.
And then, at the bottom, having, having the veterinarian know that they are ultimately responsible for any privacy loss that happens as a result of AI for any errors that happens as a result of using AI and having them empowered to make decisions based on that knowledge.
But that all, that all depends on how much information these veterinarians can get about artificial intelligence. What are their risks? What are the benefits? What are potential pitfalls? Do they know that if they use an AI tool for a diagnosis or a treatment or a, let’s say even medical recording, if there is an error, they are liable for that, for that error.
It’s not the company, it’s not the technology. It is their liability. And so what kind of tools can we put into their hands, in order to let them do their due diligence? And right now there are no tools and so
Dr. Andy Roark: More AI! That’s the answer.
Petra: Exactly. Yeah, exactly. Have the computers monitor the computers, but, so right now there are no tools, but, but that’s part of the push.
there should be performance benchmarks. There should be standards, by which, your average veterinarian can say, how does this tool measure up to these standards and can I trust it? There should be guidance information from these, these regulatory bodies and these associations to help veterinarians, evaluate the technology and say, what are my potential, what are my potential risks, right? And, how do I manage this?
Dr. Andy Roark: What advice do you have for practice owners? Are you still in a place where you think yes I think these tools are valuable or are you at a place where you’d say I think they’re valuable and this is the due diligence I would do.
Petra: Yeah, absolutely. That would be option B, Dr. Roark. So yes, I think they’re incredibly valuable. I think they’re incredibly useful. I think they’re, they are godsend, right? I mean it’s, ironically it seems like we’re using the second wave of technology, which is AI to try to fix some of the challenges that came from the first wave of technology, which is the EHRs, right?
I mean it’s, we’re at first we had to generate too much data with the EHRs and now this is decreasing the data that we have to put in, right? And I definitely think that, as a practice owner, if I were to start with an AI tool, I would sit down, and ask those hard questions.
What are you doing with my data? Is my data being taken offsite or is it being scrubbed as soon as, as soon as it’s being transcribed? Is it being collected for training purposes? And if so what kind of security to store that store that data?
Is there a possibility that you will sell at some point and sell that data? We’ve, I don’t know how many of you have read about the 23 and me, situation where, where 23 and me kept the data of users safe. But, of course they are now, essentially. I think they’re bankrupt or going out of business.
That data is potentially up for sale at, to the highest bidder, right? And number one, asking those, hard questions about data privacy and not being afraid to push, push and, and really look for accountability. That’s number one. And that’s not to say that the accountability isn’t there.
The important thing is that you do your due diligence because again, any loss of privacy for your client, any mistakes or other badge consequences from the use of the AI rests with your license. The other thing is, as a practice owner that I would say, what is the frequency of your errors?
What’s your error rate? Do you have an error rate? Are you keeping track of those things? If you run into an error, what’s your error reporting system? Is it robust? Are you actively seeking to get information from your users about when errors happen?
And then what are you doing with that? Are working with your technology to make it less error prone, right? All of those things are, I think, part of our due diligence, that we have to do as consumers of this technology. And this doesn’t only, this isn’t limited only to the tools that we purchase.
It also extends to things like Chat GPT. Chat GPT doesn’t give you any assurances about privacy when you put information into it. So, the idea of, even though it’s super attractive, let’s put this client’s entire 52 page medical record into chat GPT and ask it to summarize for us.
You’ve essentially released that client’s private medical information unto the world or unto whoever it is that, that is, controlling the data from chat GPT. So this, these questions that you’re asking about, about, safety and liability, and, an ethical focus, they’re, not just limited to, to, companies that are aimed at, providing veterinary AI it, they extend to all AI related services used in your clinic.
Dr. Andy Roark: Yeah. Petra, thanks for talking through this with me. I appreciate it. There’s so much here and this is such, it’s such an interesting, fascinating time. I think things are changing fast and they’re only gonna continue to change faster and it always feels like we’re behind as far as what questions should we be asking here.
And so I think you’re doing great work. I really love to hear the way that you think, where can people find you online? If they want to continue read your thoughts. And I said I see you, you posting a lot and talking a lot, and I just, I always benefit from your insights. I love the fact that you’re paying a lot of attention to physicians in the human side of medicine and how they’re using AI on the human side, and I just, there’s, I think that’s a real service in the way you condense that stuff and put it out in bite size pieces. So yeah. where’s the best place for people to find you?
Petra: Oh, thank you Andy. So LinkedIn is probably the most reliable way to reach me. If you wanna reach out and connect, I’m very happy to chat. In terms of, the website that I have that has some basic, AI foundational literacy, lessons, that’s a, that’s vet made.com.
So those are two resources that are connected to me. Definitely LinkedIn is the one that I’m most responsive to. Otherwise my email inbox gets a little bit crowded. Yeah, absolutely. Happy to reach out, happy to chat. Also very interested, not just in North American practice, but worldwide, AI in veterinary medicine has potential benefits, not just in private care practice, but, but in agriculture, and wildlife practice.
Or wildlife medicine. This isn’t limited to North America, I think, we have a lot to learn from, from, different regions of the world, different countries and how they’re integrating their AI. And, and I’m very keen to, to, connect with people worldwide. So yes, looking forward to seeing you there. And, Andy, if, if there’s any, anything more you wanna talk about? I’m always here.
Dr. Andy Roark: Oh, thank you so much, Petra. Guys, thanks for being here, everybody. Take care of yourselves, gang.
And that’s it, that’s what I got for you guys. I hope you enjoyed it. I hope you took something away from it. Thanks to Petra for being here. Guys, be well, everybody. I’ll talk to you soon.
