Subscribe via your favourite podcast platform: Spotify | Apple Podcasts | Google Podcasts | Anchor | Breaker | Pocket Casts | RadioPublic

In this week’s episode:

• AI is here to help, not replace. Nico debunks the myth that AI is coming for veterinary jobs – highlighting that while AI is great at repetitive, well-defined tasks, it can’t replicate human expertise, intuition, or empathy.
• The real value lies in the basics. From summarising 20-page medical histories to guiding users through complex software tasks, AI’s biggest wins come from solving simple but time-consuming pain points.
• Data quality is everything. Effective AI is built on clean, structured, and consistent data – something the veterinary sector struggles with due to time constraints and outdated recording practices.
• Don’t automate for automation’s sake. Drawing from his corporate background, Nico warns against flashy, unnecessary AI projects and champions impact-driven innovation with clear purpose and measurable benefits.

Nico Frisiani, co-founder of Lupa, reveals how AI in vet med is designed to enhance – not replace – clinical teams by streamlining operations and supporting, not substituting, expertise.

Additional Guest Spotlights

• Recommend Resource: This week, we hear from Libby Kemkaran-Thompson, who recommends Notes From A Friend by Tony Robbins – a short yet powerful read that distills the core principles of mindset and personal growth into a format you can absorb in just 20 minutes. Don’t be fooled by its size – this book has the potential to shift your perspective and help you identify what’s been holding you back.
• Next Episode Sneak Peak: Next episode we hear from Eric Goldman, founder Vetology AI. Here’s a peak at how they are doing things differently… “At Vetology, we call AI a screening result, not a diagnosis – because true clinical decisions require far more than data alone.”
  
  

Show Notes

• Out every other week on your favourite podcast platform.
• Presented by Jack Peploe: Veterinary IT Expert, Certified Ethical Hacker, CEO of Veterinary IT Services and dog Dad to the adorable Puffin.
• This weeks guest is Nicolo Frisiani. Nico is an AI Engineer by trade. He spent the last few years working as a consultant and building AI solutions for some of the largest companies in UK and in the world. About 1.5 years ago he decided to bring his AI knowhow to the veterinary market by launching a revolutionary Vet Operating System (Lupa OS) powered by AI, aimed at bringing the latest of technology innovation right into the hands of vets.
• Find out more about Lupa here!

Transcription

Jack Peploe:

Coming up on Modern Veterinary Practice,

Nico Frisiani:

It’s important to give a time to these misconceptions. As in I think there’s a lot of things that today people think AI could be able to do, that AI is still very, very far from doing. I don’t think that necessarily will hold true forever, but it’s definitely true today and most likely true for the foreseeable future. So what that means is there is a number of sophistications in the profession of the vet, but also of the receptionist of the nurse, of anyone on the staff, of as in the clinic, to be honest, that are too complicated still for an AI to ever be able to substitute.

Jack Peploe:

Welcome to the Modern Veterinary Practice Podcast. I’m your host and veterinary IT expert, Jack Peploe. In this episode, I’ll be welcoming Nico Frisiani to the podcast who will talk to us about the power and limitations of AI in veterinary practice. Nico, the co-founder of Lupa, brings a rich background in AI and data science, and together we’ll explore how AI can genuinely support not replace veterinary teams from misconceptions and data quality to practical applications in daily clinical operations. This episode offers a grounding yet visionary look into the future of veterinary technology,

Nico Frisiani:

Hi, I’m Nico for Lupa. I have a little bit of a random mixed background to be honest. I studied very technical stuff, electronic engineering and computer science, ai machine learning, and then there a bit of bouncing around. Throughout my professional career, I did more or less half the time in there as a data scientist, AI engineer, building models for big companies. And then the other side of it was more of a commercial consultant role, so trying to help big corporates go through mostly digital transformations from the project management and commercial side of things. Then about two years ago, I decided I was a bit bored of that and that the veterinary industry, which at that time I still had had very little contact with just one project in the former consulting firm seemed to be far behind in terms of technological innovation and especially in terms of how widespread AI applications were. And so I thought it could have been really interesting and really fun to try and bring AI innovation to the industry. And so quite basically almost two years ago, I quit my job together with two other friends and started Loop and so been a loop ever since.

Jack Peploe:

Amazing Nico. Well, look, it’s awesome to have you on the Modern Veterinary Practice Podcast. How are you today? Very good. Energized. Excellent. That’s just what we want. Well look, as you mentioned, you’ve had quite an impressive journey from building AI solutions for major companies to bring that expertise into the veterinary space. You went into the light side, gone from the dark side, which is excellent. Now today I’d love to explore how AI can genuinely enhances decision making, improve efficiency and transform the day-to-day operations of veterinary practices. So let’s get started. Now, AI is often portrayed as a magic bullet for efficiency and profitability. From your experience, what are some of the biggest misconceptions about what AI can and cannot do for the veterinary practice

Nico Frisiani:

That’s a very good question. So I think it’s important to give a time to these misconceptions, as in I think there’s a lot of things that today people think AI could be able to do, that AI is still very, very far from doing. I don’t think that necessarily will hold true forever, but it’s definitely true today and most likely true for the foreseeable future. So what that means is there is a number of sophistications in the profession of the vet, but also of the receptionist of the nurse, of anyone on the staff are better in the clinic to be honest, that are too complicated still for an AI to ever be able to substitute really.

And I think there is a bit of a misconception. People think thinking that AI is out there, accounting for their jobs and that AI is this extremely sophisticated virtual intelligence, a lot more intelligent than them, that hence can do their jobs better than them. In reality, I think people seem to forget that AI is for now still mostly just a bunch of math. It’s mostly statistics. It takes from data and it learns from data and it replicates and it is able to automate and streamline tasks that have been very well-defined and very deterministic tasks that have happened in the past. Now with the innovations in the last two years that has expanded to slightly less deterministic ones, though still AI doesn’t really invent much, is still just replicating from past data that it has being fed too.

So for example, there is things like writing notes for a vet that’s a task that is a relatively deterministic listener to consult and figuring out what to write. It has been done a million times, there has been a million different notes generated, and so you can feed that into an ai. I can learn that and automate that. However, there is so many aspects of every individual’s job in a clinic from the vet that has a lot of that more humane interaction or complicated diagnosis, interpretation or gut feeling that a vet might use to determine a really complicated case that they have in front of them. And that gut feeling is generated from years of experience on their side. It’s really hard to put into numbers and to really explain and so really hard to teach to our robot. And so there is a number of things that a vet does that an AI is really nowhere near being able to automate.

And so I think the big misconception is not thinking that AI is there to get your job because it really isn’t. AI is there to make your job easier. It’s there to help you take the easy task despite it being it is very intelligent but it’s very focused intelligence. It can only do that one thing. It does it really well. So if you take the note taking ai, it does extremely well. It does better than you probably, right? The problem is it only does that and you as a veteran, as a receptionist or as a nurse or as you do a lot of other stuff in your job. And so AI can help make sure that some of the tasks that to you might be a bit more boring even can be automated and streamlined and done better because you have these very highly focused, very intelligent virtual assistants that can help streamline stuff.

But it’s important not to forget that our jobs of most humans out there are very broad. That focused intelligence can only get so far and that to substitutes, and would need hundreds of extremely focused different AI bots that can try and figure out how to automate hundreds of things that a vet does. Our spectrum of the job is so broad that actually attempting to substitute it is a crazy, crazy call venture, right? I think that’s the thing that people really need to keep in mind that AI can help more than anything else. We are still at the phase where I think personally, I think that a scary doomsday scenario of AI is still very far.

Jack Peploe:

No, that’s good to know. And I mean you mentioned transcription there and I’d love to pick on that, but in the veterinary sector we hear a lot about AI automating tasks, transcription being one of those, but what are some of the less obvious ways AI is making a meaningful difference in day-to-day practise operations?

Nico Frisiani:

I think there are a few different applications that are probably worth highlighting the most after the transcription, which I guess I agree is the most obvious and by far the most widespread we see in Lupa. We see ai, we have a number of different little bots, pilots and things that appear. Technically it was Jerry, it always appears in the same face, but the reality behind the hood, there’s a lot of different things that are happening in a bunch of different places and we see it mostly being used in three other application outside of note taking. I would say though, I think there’s a lot that we can still do, but three core ones that are being used the most often, the first one is probably the simplest one, but medical records search.

So vets spent quite a bit of time coming through medical records, especially if a pet is a bit old and has had a long medical history and there is suddenly a 20 page PD thing. You just scroll through to find the right information and remember who the pet in front of you is. So right before the appointment you do usually scroll through that 20 page P to try and figure out what is happening and who’s in front of you. And so for that AI is perfect, right? Because that’s just a digestion interpretation of information and that’s what the AI is great at. So we just feed the medical record into the AI and the AI is really helpful. That’s telling you the highlights are give you the bullets and obviously you still that still up to you still you need to go through if you want to make sure that a very specific event has happened or how it has happened or the notes of the vet wrote.

But it can give you a bit of that 2, 3, 4 bullet points overview at the beginning of an appointment. So you don’t have to scroll through any pages. That’s quite used in our software. I would say the second most used is third most used after the first is definitely note taker. The third most used is actually to help you use the software. So I mean we obviously spent a lot of time making sure that the software is designed to be user-friendly, but it’s really common, everyone knows, right? Practice management software is a complicated piece of tech. It has a lot of features, it does a lot of stuff. It’s not that obvious that you can pick it up on day one and learn how to use it every day.

I think usually true that people can pick it up on day one for the basic stuff, the usual flows that they go through every time. But with that one flow that they go through once a month because it’s a bit of a rare one and they probably forget how to do it between one time they do it and the next in that Jerry usually make helpful, it helps you figure out how to do things. So instead of having to wait on a customer support or someone else to explain to you how to do it or read through the user guide words you ask in the chat to Jerry and Jerry points, it highlights things. It click on this button and highlights in red and that’s button it highlights in red, it guides you through the actual flow and not necessarily an assistant that actually automate any part of your data. It’s not really the core topic of conversation here, but it is still an AI system. It still is really, really helpful. It makes sure that you don’t have to waste time on the user guides or either worse, they cap the phone with customer support. Cool. Nice. Yeah.

Jack Peploe:

So I mean going for looking about your background and outside of the veterinary world now you’ve obviously worked with someone like the largest companies outside of veterinary medicine. Are there lessons or approaches from other industries that you think veterinary practices could benefit when adopting ai?

Nico Frisiani:

Yeah, definitely. I was just thinking which ones, but two very important things that come to mind. I mean there’s probably a thousand learnings that I can sit here and think of, but two very important ones. The first one that we had, so I was working for a consulting company that gets hired by big corporations who go there and build stuff for them At a certain point, I worked for a long time for this huge car manufacturer, German car manufacturer. And I would say for me the big lesson there was don’t try and automate things that there is no need to automate.

We were building a whole recommendation engine to tell users what’s the next best car they should be buying? The lease is coming to an end and are you ready for electric? And if not, what is the car within the various options that the car manufacturer had that you should be buying to incentivize the lease upgrade and all that kind of stuff. And we built probably the most sophisticated algorithm I’ve ever spent my time building and 90, 95% of the gain came from the first 10% of work we did and then the remainder 90% of the work to get that extra 5% benefit was such a waste of time. We were trying to automate such niche use cases and making sure that we capture such really, really small niches of people that had the most unique habits and making sure that we have the right recommendation for the people that had the most unique habit of which there was a hundred users in all of Germany.

It was a complete waste of time. And I think there is a tendency, especially in big corporations that have a lot of money, though it is sometimes a tendency all over the place. So maybe this is more recommendation for the gigantic veterinary groups and I have a lot of cash, but there is a tendency in corporations to automate for the sake of automation so that I can see I have AI and I have AI everywhere. It is not just AI in the note taking and in the medical history search, I have AI in every click I do on the screen, but it’s probably useless and I think that’s something that people should keep in mind not to do AI just because it’s nice and fancy to say you have ai, but actually to prioritise things that make a difference and have an impact and save you time or create more revenue or whatever it is, but something that actually you

Jack Peploe:

Have a key purpose.

Nico Frisiani:

And there was a lot of the car manufacturer. One is the one that comes to mind, but there was a lot of projects that we did that had a lot of wasted time on trying to build overly sophisticated things for the sake of them being sophisticated. The second probably even more important learning is, and this is what in general for the entire profession, it’s the quality of data makes the world of difference in building ai. That’s almost the only thing that counts. What we used to say in my previous team, a job was that somewhere between 10 and 15% of the effort is the engineering algorithmic effort and the rest is just it. So we spent the vast majority of time cleaning data, working on data, ma manipulating data, and then a bit of time building the algorithms and algorithms. In many ways we are all using the same basic models and basic math formulas and basic, there’s nothing particularly complicated reinventing the wheel unless you are really working in a company that building foundational models, unless you’re there, you are likely just using someone else’s models and building on top of the track.

And so for that scenario where most of us live, the only thing that really makes a difference is data quality. And I have seen the difference between working in industry with good data quality in one with Met because I’ve worked with companies on both sides of the spectrum on the tech companies that tend to have really sophisticated databases with really clear stuff. And the more old school ones that tend to have really problematic data and the difference in the quality of our algorithms we could build was uncomparable really uncomparable, it’s a night and day, it’s AI tools from 15 years ago versus cha PT four oh that had a difference that you can build between good and bad data.

And I think the veterinary industry has particularly bad data. It’s usually kept really badly in the practice measure software. To be honest. I blame us, not necessarily Lupa, but I blame us as a general part of the industry. But also I think there is a bit of an old school way still of doing things in the veteran clinic themselves where vets have no time and no incentive in recording data the right way in recording data and deterministic clean way that you can then analyse in saying this is the diagnostic with the right code associated to it in a very deterministic list of diagnosis that I’m picking from. It’s all free text always. And not necessarily because investors know what they’re doing. If anything, it’s the opposite, right? They have no incentive and no time, they’re too busy. And that however creates, creates really the complicated data to work with on this side when you’re trying to build AI tools because we are spending most of our time cleaning the data up and trying to make sense of the deterministic diagnosis under deterministic notes and treatments and prescriptions that the vets have in their systems so that we can build a tool that actually helps them.

So you want to start building a real AI vet, a system that starts recommending diagnosis, prescriptions, treatments so that you really, all you care about as a vet is your own final voice and saying yes and no to things, but really you have client interaction at the maximum and your only interaction with the computer is saying, yes, you got this prescription right? You got this one wrong. To have that mega sophisticated AI VE system, which would really be the dream because it really frees up so much time of vet and they can focus on the things that matter you have. You need to have a system that is able to interpret notes and consults and everything and diagnose to be able to diagnose. They need to have fast data with clean diagnosis that they can study, that they can understand and that they can extrapolate into the future stuff. And having that without deterministic data is really complicated I think. So data quality would be, I think the biggest thing that the veterinary industry should be working on if they want to be ahead of the curve in five years with AI applications.

Jack Peploe:

Yeah, no, a hundred percent. Well, Nikka, I warned that this was going to happen. I knew we were going to go down the rabbit hole and our time is up, which is incredibly annoying, but we’re going to have to drag you back on, I don’t mind because obviously there’s so many other elements we’d love to talk about around going down the privacy route and the ethical dilemmas around AI because again, that’s a whole different segment, especially when we’re considering data. But look, thank you so much for sharing your insight today. I think you’ve sort really helped start the journey in helping demystify AI for a lot of our listeners, which is great, and provide a balanced view of both the opportunities and challenges. Now for anyone who wants to learn a little bit more or want to keep the conversation going, what’s the best way to reach out to you?

Nico Frisiani:

It really depends what they want to talk about. If they want to talk about ai, that’s probably directed to me. That would be the easiest way. And I usually either reply on LinkedIn or on my email, just nicolo@lupapets.com. If they want to talk about Lupa in general and they’re interested in Lupa, then probably the contact us form on our website is the easiest way we monitor that by, because we usually reply within 24 hours.

Jack Peploe:

Brilliant. Well, we’ll make sure that they are on the show notes. But look, it’s been an absolute pleasure having you on the podcast, Nico, and I really appreciate your thoughtful perspective on the role of AI in veterinary care. But thanks again for being here and obviously catch up soon.

Nico Frisiani:

Thank you, Jack.

Jack Peploe:

Every week we ask professionals and experts to suggest a best business resource for our listeners. This week’s recommendation is from Libby Kemkaran-Thompson.

Libby Kemkaran-Thompson:

Love to recommend people start with Tony Robbins notes from a friend, all roads lead to Robbins. And I went to a firewalk that he ran when I used to work in the city of London years ago in my old career before I retrained as a vet. And I walked into this arena and there was 13,000 people and I had no idea what was about to happen, and I stood there with my arms folded like this, what is this? And then that four days changed me so much and changed the way that I thought. And it’s all in this tiny little book notes from a friend that is literally, you can read it in 20 minutes. It’s a really fast flick book, but it gives you the central concepts of the biochemistry of success in one place in a very accessible format. You can be a complete geek and overdo it and go and do a degree in it like I did start there. That gives you everything in switching that growth mindset in a heartbeat. Like in a heartbeat, because we change our values easily. Very uneasily. We change our beliefs in a heartbeat, but our values stay the same. So if you can figure out what they are behind that belief layer, the values that you ground into can drive you towards success. But the beliefs, when you figure out what’s limiting you, when you figure out what’s in your way, what’s holding you back, you can change those. So yeah, notes from my friend Tony Robbins, my favourite.

Jack Peploe:

Coming up next time, we welcome Eric Goldman, president of virology to the podcast. With over 25 years in technology and a passion for purposeful innovation, Eric shares his journey from Fortune 500 consulting to transforming Veterinary Diagnostics. Through ai, we explore the real world applications of AI and veterinary care, address common misconceptions, and discuss why the future of diagnostics is a partnership between clinical expertise and smart technology.

Eric Goldman:

One of the things that we’re careful about in Vetology is we call AI a screening result. We don’t call it a diagnostic. The reality is that software doesn’t know anything about clinical signs, doesn’t know anything about prior visits, doesn’t know anything about blood work, and sometimes the vet techs will put in the wrong breed, they won’t update it, they put in the wrong weight. All those things factor into the diagnostic chain. And so the point that what we know is computers are consistent, right? So if you take a well-positioned, well-taken radiograph, complete orthogonal views, you should be able to get a consistent result. That result has to tie back to something that you’re seeing and something that,

Jack Peploe:

That’s it for this episode. All links and recommendations we talked about are in the show notes. Don’t forget to subscribe and share the podcast if you found it useful. In the meantime, thanks for listening and see you next time.