April 25, 2025

#141: From FOMO to Pro: How I Use AI in the Clinic (And How You Can Too). With Dr Hubert Hiemstra

#141: From FOMO to Pro: How I Use AI in the Clinic (And How You Can Too). With Dr Hubert Hiemstra
Apple Podcasts podcast player icon Spotify podcast player icon RSS Feed podcast player icon
Apple Podcasts podcast player icon Spotify podcast player icon RSS Feed podcast player icon
Season 1

In this solo episode Hugh tackles a bold idea: if you're not currently using AI in your clinical veterinary work, you may be falling behind. And if/when artificial intelligence gets to the point where it reliably has access to all the most up to date information for managing your patients, will you be negligent if you DON’T use it? 

Hubert shares how artificial intelligence is quietly but powerfully reshaping veterinary practice. He paints a vivid picture of an AI-powered clinic, shows you practical tools you can start using immediately, and challenges you to rethink the future of your role as a vet.

Drawing on two years of practical experience using AI tools in podcasting and clinical settings, Hubert provides real-world demonstrations of how AI can be integrated into everyday vet work. He discusses the current reliability of AI in clinical decision-making, shares evidence from human medicine studies, and highlights where AI shines — and where it still falls short. Practical tips cover using ChatGPT, Perplexity, and custom AI assistants to speed up decision-making, improve case management, and even build clinic-specific knowledge tools.
The episode also delves into the philosophical implications of AI's integration into the profession, forecasting a shift in veterinarians’ roles towards becoming curators of context, emotional connectors, and ethical decision-makers. Hubert closes by reflecting on how AI could drive new business models in veterinary care and how vets can position themselves to thrive by focusing on human skills, technical expertise, and continuous learning.

What you'll discover:

  • A demo showing how to use ChatGPT for clinical reasoning, how to build your own custom AI assistants, and how to set up a personal knowledge library with free tools like Google's NotebookLM.
  • How to spot and handle AI hallucinations — and why your human skills like empathy and ethical judgment are still your unique superpowers.
  • The business opportunities of AI for your practice, and why clinics that embrace AI may thrive while others struggle.
  • How AI could reshape veterinary education and open doors to a more sustainable, fulfilling career.

 

Hubert doesn’t just speculate — he shows you how AI can work in clinics today, and how you can lead instead of lagging behind.

Tune in if you're ready to future-proof your practice, sharpen your skills, and unlock a better way to be a vet.

 

Try out our AI Blood Gas Assistant

Join us at Vets on Tour Wānaka, NZ, for the most fun CE event you’ll attend this year. 
Join our community of Vet Vault Nerds to lift your clinical game and get your groove back with our up-to-date, easy-to-consume clinical episodes at ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠vvn.supercast.com⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠.

Get help with your tricky cases in our ⁠⁠⁠⁠⁠⁠⁠⁠Specialist Support Space.⁠⁠⁠⁠⁠⁠⁠⁠

V⁠⁠isit ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠thevetvault.com⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠for show notes and resources related to this episode.

⁠Subscribe to our weekly newsletter⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠ ⁠for Hubert's favourite clinical and non-clinical learnings from the week.

 

Topics and time stamps

03:08 What could the future of clinical practice look like with AI?

12:48 Can I trust it? What does the data say?

18:22 Where does Ai get its information from?

25:20 How I use it: Using 'general' models for clinical decision making

33:29 Building your own Ai tools

34:06 Google Notebook LM

40:58 MyGPT - your own assistant

54:49 What might the role of the future veterinarian look like?

 

How AI Will Reshape the Future of Veterinary Clinics

AI is poised to significantly reshape the future of veterinary clinics in numerous ways, potentially impacting everything from client interaction to the daily workflow of veterinary professionals. Hubert believes AI is the biggest new thing to happen possibly in our lifetime and could shape veterinary science in the next decade, starting now.
Here's a breakdown of how AI could reshape veterinary clinics, according to the source:
Enhanced Client Experience and Communication:
  • Imagine a client arriving at a comfortable clinic lounge and being greeted by a front-of-house team member who offers them a coffee and a tablet. Alternatively, a scanner could recognise the pet's microchip, bringing up their records and prompting an AI-driven chat on the client's phone or the provided tablet.
  • A polite AI avatar, potentially even with the vet's face and voice, could interact with the client. This AI could verify client details, similar to how a human client care representative (CCR) would, but without the risk of forgetting.
  • The AI could collect a perfect and comprehensive history, remembering to ask about diet, toxins, deworming status, tick preventatives, and other important details that human vets might sometimes overlook. It could also access and utilise information already stored in the practice management software (PIMS), such as vaccination and medical history.
  • This AI could summarise the collected history, highlighting relevant past events for the vet or triage nurse/tech to review before the appointment. It could even incorporate knowledge of the patient's entire history from every clinic they've visited if integrated with a larger database.
  • Furthermore, the AI could access the most up-to-date clinical knowledge, including recent studies, providing the vet with key pointers and considerations before the consultation even begins.
  • After the consultation, and AI assistant would generate notes, discharge forms, referral letters and any other necessary documents.
  • Discharge sheets could include a 24/7 Ai service follow-up questions. For serious concerns, AI could connect the owner with the clinic team or schedule an appointment.
Examples Of Improved Diagnostic and Treatment Processes:
  • AI could analyse patient records and test results, highlighting points of concern and creating to-do lists for the veterinary team.
  • With access to current clinical knowledge, AI could provide insights and suggestions for diagnosis and treatment plans, even considering subtle abnormalities in previous results.
  • AI can assist with calculations, such as converting units of measure and determining drug dosages, acting as a double-check for critical medications.
  • AI can aid in interpreting complex test results, suggesting next steps.
  • AI can summarise long patient histories, making it easier and faster to identify relevant information, past treatments, and responses. This is particularly useful in emergency situations with extensive prior records.
Efficiency and Workflow Optimisation:
  • By automating tasks like history taking and note generation, AI can free up valuable time for veterinarians and other staff.
  • AI-powered tools could streamline administrative tasks and improve overall practice efficiency.
  • Building custom AI assistants using platforms like Google Notebook LM or MyGPT could allow clinics to create task-specific tools using their own protocols, standard operating procedures, and preferred resources. This could be used for onboarding new staff, accessing equipment manuals, or understanding practice policies.
The Evolving Role of the Veterinarian:
  • While AI can handle information gathering and processing, the veterinarian's role will likely shift towards being a curator of information, providing context, and making ethical decisions.
  • Empathy, reassurance, and trust-building will remain uniquely human skills and become even more critical. Veterinarians with strong human skills and comfort in using AI will excel.
  • Veterinarians will still need a strong foundational knowledge to ask the right questions of AI and to critically evaluate its suggestions.
  • Manual skills will likely remain important, at least for the near future.
  • AI could lead to increased efficiency and productivity, potentially impacting the business model of veterinary practices, with possibilities for increased wages, profits, or more affordable care. However, there's also a possibility of losing revenue from simple cases as pet owners might initially consult AI.
  • AI will make research more accessible, potentially accelerating progress in veterinary science.
While the seamless integration described might require combining various AI tools currently, veterinary software companies are likely actively developing such comprehensive solutions.
The key is to approach AI as a thinking partner and to always verify information, especially when using general AI models, by asking for references and checking the source.
Overall, AI presents a significant opportunity to enhance the capabilities of veterinary clinics, improve patient care, and reshape the roles of veterinary professionals. However, it also necessitates a focus on uniquely human skills and a thoughtful consideration of the ethical and practical implications of its widespread adoption.

 

I'm going to start this one with a possibly controversial opinion.I think that if you're not using some sort of AI help in your clinical work as a vet that you might be a little bit negligent.That's a bold statement, I know, but bear with me.
I'll defend that statement as we go along in this one.But for now, I am Hubert Einstra and you're listening to The Vet Folk, where I love, love, love finding, learning and sharing anything new that can help you in your career as a vet.And in this episode, I am sharing some thoughts and practical tips on the biggest new thing to happen possibly in our lifetime.
Artificial intelligence and how it might shape what virtually science will look like in the next decade.Actually, never mind the next decade like tomorrow.And our guest for this one is nobody.It's just me.Now let me be clear, I am no AI expert.
I can explain AI tools to you in the same way that I can show you how to get a perfect slice of toast for my toaster, in that I can show you which buttons to push and which nerves to adjust.But don't ask me about the magic happening inside that box once you push that lever down.
What I am is curious, a little bit nervous, but mostly excited.I've been experimenting with AI tools in my podcasting and my clinical vet work for the past two years or so, and I find myself often telling friends and colleagues about the new tools that I'm using and how I'm using it.
So I thought I'd share some of that with my friends and colleagues on a wider scale.That's you.So here's what we're going to be talking about.I'm going to start with a vision of what I think a very realistic and a not too distant feature of work as a vet with AI tools integrated into your workflow could look like.
Then I'm going to get practical by showing you how I use a few of my favorite AI tools right now.Simple, easy stuff that you can use immediately.And by show, I mean that I'm actually going to show you on a video.This episode is going to be as video if you listen on Spotify or watch on Spotify, and I'll be screen recording so you can see exactly what I'm doing.
If you're audio only, I think you're still going to get the concept, but you'll probably get a little bit more from video.I'll give you a heads up when I switch my screen recorder on and then I'm going to wrap up with a little bit of philosophical, philosophical musing of what this could actually mean for us as vets, and I mean existentially mean for us as vets, which I think is a huge question.
Now before we jump in, just a quick reminder of our New Zealand Snow conference.From the 10th to the 15th of August we're going to be hosting some top notch education, definitely some live podcasting, but also long gaps for exploring, skiing, fishing or just chilling in one of the world's most spectacular locations.
That's Wanaka in New Zealand.We can only take 80 people in total and we've currently sold about 55 of the tickets, so tickets are going to run out.Don't dilly dally on this one.Go to vets on tour.That's all one word Vets on tour.com for your tickets.Oh, also just a quick thank you to our two key sponsors for this event, which is Index and Hillspet.
Nutrition drinks are on you guys.OK, let's get into our episode for today.I want to paint a picture for you.Imagine your client arrives in the waiting room.
It's a comfortable lounge area and a smiling front of house team member hands them a coffee and a tablet like an iPad.Not like a blue pill, red pill kind of tablet.Or better yet, your clinic is connected to Verdi, so a scanner at the door recognizes the pets microchip, which brings up the pets record on your practice management system, which triggers a message to your client's phone to start a chat or to the iPad that you've just given them.
On the screen on the tablet or on the phone is a perfectly polite AI avatar of the best client care person that you've ever met.Or it could even be your face and your voice.So the AI says, hi, Mr. Smith, I see you coming in today because Jax has been vomiting.
Before we start, can we just check that we have all of your correct details and it goes over phone numbers, addresses, everything.It never forgets to do this, unlike most human CCRS eventually do.Then it goes on.Now can you tell me a bit more about what is happening to Jax?
When did the vomiting start?The AI then proceeds to collect a perfect history, never forgetting to ask about the diet or toxins or deworming status or tick preventatives or any of the other stuff that I sometimes actually often forget to ask.Or maybe it doesn't even have to ask those questions because it is built into a practice management software and it has access to every record of every patient.
So it already knows that Jax is not up to date with his dewormers.He's fully vaccinated.He's had three skin conditions and two episodes of vomiting in the last five years.So when Mr. Smith walks in, it just double checks whether anything has changed since the appointment was booked.Then it takes all of that information and summarizes it into a beautiful current history, highlighting anything from the past histories that might be relevant for today's visit.
You.As the vet get ready for your appointment or maybe you have a nurse or a tech who does triage for you.Whoever it is, they go to the patient file and they're waiting for them.Is that summarized history with just some key points tonight.So AI says to you, hey, Hubert, Jax has been vomiting for three days, no diet change, no known toxins, is a bit lethargic, etcetera, etcetera, etcetera.
And because it is knowledge of that patient records in your pimps or even better the entire patient history from every vet it's ever been to because it is integrated with fatty, it adds things like.Note that Jax is 9 months overdue on his Parbo vaccinations and I can't find any information on deworming.
I've also seen three previous visits for non specific vomiting for Jax, the most recent 1-3 months ago.But maybe it even goes beyond that because your AI also has knowledge of the most up to So attached to your history is a summary of some with some key pointers.
Hey doctor, you Jax is vomiting quite a bit.I see that at his last visit his sodium and potassium ratio was on the edge of normal and notable.Also a lack of deworming.So I'd consider checking his electrolytes to make sure there's not a sneaky at a sodium patient and maybe starting with some fecal tests.
Let me know what you find on the clinical exam.So you go in, you start your consultation, you have just saved yourself 5 to 10 minutes of history taking and you can go straight to connecting with his client, doing all the vitally important trust building exercises and not getting bogged down and standing there with your notebook or your computer and saying, well, when did the vomiting start?
How long has been vomiting?Let me read through the history to see what happened last time.Time saved, more efficient, more connection.And maybe during the console, do you even have the AI open somewhere in the background?The client knows it and you know it, and you and the client and the AI all interact together.
Let's give our AIA name a picture.It integrated throughout the hospital like Iron Man's AI Jarvis.So let's call ours Jarvis as well.So you examine Jax, you ask a few more questions, you chat to the owner and you chat to Jarvis.You say, hey Jarvis, Jax is quite painful mid abdomen and his membranes are a little bit tacky.
I think we should do an ultrasound scan and run some bloods.Let me start with those vehicles, though.Jarvis, can you talk to Mr. Smith about the practicalities and the cost of blood, gas and a scan while I go and get those vehicles running?Mr. Smith?I'll be back in 10.In the meantime, you can ask Jarvis any questions about what we've discussed so far.
I'll be back to clarify anything that's not clear and then we can confirm a plan.So you go and run the fecal test and you collect blood and you run the blood and of course all of your equipment is connected to your perms.So as those test results comes in, Jarvis adds it to the record and highlights any points of concern and creates a lovely list of DDS for you.
You double check the results and anything of concern on the clinical examination.You spend more time connecting with the owner, you reassure them and you go back out and you do the scan and you cut that sock out that JAXS Ethan.And of course, by the end of all of this, your notes are already done for you, plus the discharge form for the owner.
And if you wanted to refer anything, there's a referral letter ready for you to check over.Tick the box and hit send.Maybe on that discharge sheet that Jarvis made for the owner, there's a QR code that the owner can scan to chat to Jarvis 24/7 to get their commonly asked questions answered for them.
So how long should the suture stay in?I can't remember what the vet said about when I can give meloxicam, when can I eat?What should we feed him?Because you know, when you're talking to them in your console room, they're not even hearing half of what you're saying.Of course, for anything serious, Jarvis will say, let me connect you to our team or would you like me to make an appointment for you for tomorrow?
Does all of that sound like sci-fi, like 20 years in the future?And how does it make you feel this?What I just described is not that far fetched.In fact, most of what I'm describing is possible right now.If you wanted to achieve something like this, you'd have to Frankenstein together a bunch of different AI tools and it might not be super smooth, but it is 100% possible.
But I would bet, I'd really, really hope that the guys who make veterinary software are working hard on something like this.If I owned a PIMS company, this would be where most of my resources would be spent right now.And that's for how that makes you feel.I am mostly just excited about all the stuff that I'm not going to have to do and about how much better I'm going to get at my job.
With a podcast.I obviously spend a lot of time in conversation with people from all over the profession.So that's from owners of the largest corporates to the world's leading specialists, all the way down to the freshest, least confident vet.And there's not one problem that I think can't at least in part be addressed by a good AI tool, from better patient outcomes to better communication to more profitable vet businesses leading to better wages and less staff burnout.
It's definitely not going to magically fix everything, but smart people who care are going to use it to amplify and expedite the things that we do care about.So let's get a little bit more practical for this episode.I'm going to focus on the practical clinical veterinary work.Like how can we use AI to help us make better clinical decisions for our patients?
Can we, can we trust it?And if so, how are we going to do this?And I want to go back to that statement I made at the start about if you're not using AI, you are being negligent.I'm going to give you an example or we're going to give you a metaphor for why I say that.Imagine you're working as a vet in a clinic and you have a textbook that you love using.
It's a textbook that you bought 50 years ago when you started working as a veterinarian, and you've got all your notes scrolled in the side and you know on which page roughly you can find which condition.So that is what you use for your source of reference.Then your boss goes and buys the same textbook but an updated version, but it doesn't have your notes and it feels a little bit new and a little bit strange.
So you choose to stick with your old textbook, even though a lot of stuff has changed since 15 years ago.Until the newest version of your textbook has come out, you stubbornly decide that you're uncomfortable with a new textbook.You're going to read that old textbook.Another way to look at this would be your boss is super forward thinking.
They've got a big team of GP vets but they want more information and more support for the vets on the floor.So they go hire a medicine specialist or an ECC specialist or somebody who really is up to date with this stuff.And almost guaranteed there's a lot more than you, the vet working on the floor.
But you in your ego or ignorance or discomfort because you're too shy to ask or for whatever reason, decide that you're going to plot along doing things the same way that you're doing it.Never asking the input of that specialist who's just sitting there ready to help you.You wouldn't do that, would you?
You're going to ask for help.And this I think is where we're heading.We're going to have the most up to date information and a very smart brain in your pocket or on your desktop and not using it.Like if you make the wrong decision, if we get to check for a DD, you work at a wrong dose.You've missed some update on the clinical condition.
Up until now, those things are all, let's say normal human error that won't get you into trouble.So there's definitely not negligent.It's just you're a human.But if a tool like AI becomes trusted and part of everyday general practice, I think we might get to a situation where not using it could be seen as you are negligent Right now.
You have the fastest, smartest search engine and research tool at your disposal that anyone has ever had access to.But it's not just search.It has built in knowledge.It knows stuff.We're going to talk about the different kinds of knowledge that AI has and how we can curate it.But for now, let's just say this.
For most of us, it's definitely smarter than you.Now, you might be thinking, yeah, but how reliable is it?Right now?The stuff you're talking about is true, but right now can I trust it?I've heard that it hallucinates and makes up stuff.
Great question, and absolutely it does.A couple of days ago I asked Jackie Petit to help me find a specific study that I'd heard about on from one of the podcast guests, and it started giving me really relevant answers with a study name, a date for the study, and a summary of the findings, including details like percentages of the different groups in the study.
Very very super believable stuff.But I wanted a link for the study so that I can put it in the show notes.So I kept pushing for details till eventually GBT went sorry there was a misunderstanding.Wait, I'm going to read you the response that I got said apologies for any confusion earlier.
Upon further review, I realized that the studies I previously mentioned do not exist as described.Very polite.I wish it would just say, my bad, I was messing with you, I totally made this shit.But there are ways to navigate this, and I'll share a few of the things that I've learned about how to use AI effectively for clinical decision making in a minute.
But before we do that, let's look at how reliable the models are now by looking at some data.I spend a lot of time digging into the studies.That aim to prove how reliable or unreliable AI is when used for clinical decision making.The data is mostly from human Med and there is a deluge of studies coming out on the topic.
Now One thing that I will say is that a big problem with reading the data on this is that by the time a study is published on AI, it's kind of outdated because the AI model that was used in the study has inevitably been replaced by a newer model that is significantly better than the previous one.
Most of the newest studies out there looked at Jet GPT 3.5 or 4, and if you've used four point O or 4.5, you'll know that it's a different beast again.Same goes for the latest iterations.Of Claude of Croc of.Gemini, all of them.There's a common saying in the AI space that says AI today is as bad as it's ever going to be, so it's hard to draw solid conclusions, but this all that we have, so here goes.
I was going to give you a detailed rundown of a bunch of studies, but it turned out to be a long.And boring part of the.Podcast.So I've chopped it out and summarized.I will start with one little study that I really liked.There was a study that evaluated clinical decision making by asking three study groups to make decisions on the same 6 challenging medical cases.
The first group was doctors working by themselves who were 74% accurate in diagnosing these cases.Group 2 consisted of doctors using an AI to help them and they did a little bit better but not much, with the diagnostic accuracy of 76%.
But for group.Three, they took the humans out of the equation and gave the cases to AI alone, and this time we had 90% accuracy.That is a little bit humbling.If you read the analysis, the conclusion state that the doctors using the AI didn't.
Really.Trust the AI and they kind of let their own knowledge override the AI suggestions.So basically it was a problem of ego.But beyond that study, looking at all the studies that use AI in different scenarios to try and get an indication of how good it is compared to humans, The takeaways time and time again is that a good quality general AI like JAT GPT is about as good and often a little bit better than a generalist physician.
Answers from specialist are still generally better than AI, but when specialist evaluate answers from both GPS and AI, they prefer the AI answer about 65% of the time.It was OK at triage and making treatment plans when provided with initial non specific histories, but it gets better and better the further down the diagnostic pathway we go.
Performing best when it came to making final diagnosis when all of the information of a case is presented to it.There was another study that showed that a specifically trained medical AI scored in the mid 80s on the final medical exams in Germany and GBT scored 0.1% higher than that.
Now 85% isn't 100% but my question is what did you score in your final exams?I found one study that compared Med palm 2.Which is a large.Language model fine-tuned on medical data with GBT 4.They found that better prompting of GBT 4 led to better answers compared to more training of the medical model.
In other words, better questions led to better answers more so than giving it more specific knowledge and more information.So what I am taking away from all of this, number one, it's far from perfect in 2025.You still need to understand your cases and have your own knowledge and experience to give you context #2.
But it's pretty.Bloody good #3 it's not a textbook, it is a thinking partner.Like having a very smart colleague with you #4 and like with a very smart colleague, the better the quality of your questions are, the better the answers will be that you get.This is why I think we will not be replaced yet.
Your clients are going to use it.They're already using it to get some very relevant information, but they don't know which questions to ask.We'll circle back to what I think our roles are going to look like in future, but I think.Provider of context.And information filter are definitely high up on our list of tasks that we still need to do.
Something that I want to double click on is the difference between those general AI models that have been trained on all of the knowledge in the world.This is the subject specific models with a limited knowledge based on your choosing.I've made quite a few AI tools for myself for work and also for the podcast.
For example, I have an AI that uses all of the show notes from our clinical podcasts as its knowledge base, as its brain really, and nothing else.So I can ask it questions that relate to things that we covered in the podcasts and it will pull information from all of the relevant notes.
It's super useful and mostly really reliable.And I've also asked it to tell me that if I'm asking you a question about something that we haven't covered it in the podcast that it should say to me, sorry, I don't seem to have information on that topic.So I have a pretty good idea that it's not trying to BS me with his answers.
Another one that I'm finding really useful that I made is an AI assistant that limits its answers to all of the open source small animal consensus statements from the last decade or so.Again, when I ask it a clinical question like do you have any information on immune mediated thrombocytopenia, It is only going to look for the most recent consensus statement to answer all of my questions about what's the most up to I know where it's getting the information from.
By the way, the two AI's that I've just mentioned are available to all of our clinical subscribers at no additional charge.So you can check it out at vvn.supercast.com.That's VVN for vetvaultnetwork.supercast.com.So that's in addition to our clinical podcast.So I like these two AI's because I know exactly where it's getting its information from versus as I said before, when I ask an open model like chant GBT.
I worry that it might provide me with an answer from the naturopaths vet plug or a textbook chapter from 1992 that says you should use metroniders.All for hemorrhagic diarrhoea.You shouldn't, by the way.Which made me wonder, how does those general AI models choose which information to use?
So I decided to ask an expert.I asked ChatGPT when I ask you a veterinary clinical question, how do you decide which information to prioritize to give me as my answer?Here's what it said.That's a great question, and one that is especially important in fields like veterinary medicine, where accuracy matters significantly.
Here's how I approach prioritizing information to ensure accuracy and minimize potential harm #1 relying on credible sources from training.My initial training includes A diverse data set consisting of textbook peer reviewed journals, educational content from reputable veterinary universities, and trusted clinical guidelines.
Information from these high quality, verified resources forms the foundation of my responses.Sounds pretty good #2 prioritizing consensus and expert guidelines.When providing clinical information, I give priority to recommendations and guidelines from recognized veterinary associations and authoritative bodies such as the British Veterinary Association, the Royal College of Veterinary Surgeons, American Veterinary Medical Association, and the World Small Animal Veterinary Association.
Information consistent with widely accepted veterinary standards, consensus statements, or peer reviewed guidelines is prioritized over unverified resources #3 assessing context and clinical relevance.I carefully consider the context of the question, EG, species, breed, predispositions, geographical factors, current clinical practice trends.
Responses are tailored specifically to your described situations to ensure relevance, practicality, and appropriateness for clinical decision making #4 the recognition and disclosure of limitations.If a question is highly specialized, contentious, or beyond my training data and capabilities, I clearly indicate uncertainty and recommend consulting additional resources or a veterinary specialist #5 continues improvement and caution with sensitive information.
The model undergoes regular training updates to incorporate new, verified information from trusted veterinary resources.However, this is not always real time, so for rapidly evolving topics or recent developments, I strongly recommend checking current veterinary literature or specialist guidance.I maintain caution when providing information that carries high risk, especially where errors could lead to serious consequences for patient safety or clinical outcomes.
Number six, web browsing for up to That's kind of what you want to hear, right?
If it is all true, I mean, I could have just made it up because it knows that's what I wanted to hear.As a veterinarian, I've been looking for an AI expert, a human AI expert to interview on the podcast to see if this answer is actually as stated or is it just chat with you telling me what I want to hear.
If you're listening to this and you know different to this or you know somebody who works in AI programming or training or anything like that, please do get in touch and I can pick your brains on it.I did also ask it if I'd ask it for a web search.How does it prioritize its information then?
And it's very much the same sort of information.Prioritize credible veterinary organizations and institutions.And it lists a bunch of them, university and specialist centers.Is it second to source?Then regulatory agencies like the Veterinary Medicines Directorate, the European Medicines Agency, and the FDA, then peer reviewed literature and databases, and then reputable veterinary resources and trusted platforms.
It does not list the Vet Vault yet.I'm going to have to work on that.Sites that clearly disclose their authorship, which are experts with credentials, and their methodology or editorial standards are prioritized.And then sources that are neutral, objective, and without significant conflicts of interest or commercial biases are prioritized.
What I avoid Websites with unclear authorship, poor transparency or obvious biases, information from blogs or personal pages, and less authored by recognized veterinary specialists, and size primarily focused on marketing or selling products without robust scientific backing.
Again, perfect answer.If that's true, and if it can achieve what it says it's doing, that's pretty impressive.And we wrap it up with what it says are its limitations.It says despite these careful processes, there remain limitations.My knowledge has a cut of date which is October 2023 in my case that was asking chat TBT 4 O now.
While I strive to prioritize accuracy and reliability, occasionally inaccuracies or outdated information may still appear, particularly regarding rapidly evolving fields.This is why your professional judgement and verification from current veterinary sources remain crucial.Repeat that.
This is why your professional judgement and verification from current veterinary resources remain crucial so your university degree is not time wasted.And my clinical podcasts are still useful.You do still need to know some stuff for now.Now, as promised, I'm going to show you a few of my favorite ways of using AI in practice right now in the first half of 2025.
First, I'm going to tell you what's working for me when I use open models like Chanchi Petit, which is what I use most often, but I am going to mention another one as well.I like ChatGPT as a thinking partner rather than just a source of information.It certainly has the capacity to provide information, but I don't 100% trust it when it comes to potentially consequential decisions.
In other words, I love it to help clarify my thinking and to help me come up with things that I didn't think about and to explain things to me that I don't really understand, which is really good at.But I'm not asking it to give me the dose for IV insulin or to give me the protocol for treating ADKA patient, or at least not without checking its references, which I will come to in a moment.
Let me give you a couple of examples of how I've used it recently and hopefully that'll demonstrate my thinking around how I think we can use it to get a trustworthy answer or at least speed up how we find information.So I had a complicated collapse case recently that I couldn't quite get to the bottom of.
It had clinical features and test results that kind of fitted with several DDS but not really with anyone particular diagnosis.For example, it had a stocking gold that a Halo and crappy mucous membrane colour.But it's blood pressure at was first normal and then it went high and it had a slow heart rate.
So I wanted it to be anaphylaxis, but the blood pressure didn't quite make sense.So I wanted a recap of what else could cause gold bladder Halo.The deed is for that, and I asked JGBT to give me the differentials for that as well as explained for the first time ever the pathogenesis behind the gold bladder Halo, and it was really good at giving me that information.
Then we talked a little bit about the high blood pressure and low blood pressure and anaphylaxis and unusual forms of anaphylaxis and neurological shock and all sorts of other things that didn't even enter my head.I didn't get to a final diagnosis.I still wanted to chat to a specialist, but for that moment in the clinic when I had to make decisions, it really actually did help my thinking.
The key there is that I didn't just give it the case history and then asked it to come up with DDS for me.It's generally not that great at answering vague case questions.I mean, it'll give you an answer, but I don't know if you can trust that answer.But if I say to it, here's what I've got, here's what I found, here's what I'm thinking, but I don't get this bit.
Can you help me think it through?What are alternative explanations for this?What am I not thinking of?That's when I'm finding it really, really useful.I also love it for.Can you explain this concept to me?Explain it like I am a primary school kid.I also really am loving AI to help me with calculations.
So things that I can do, but I don't do that often.So it usually means I've got to find a textbook or Google a formula and through the fairly complicated maths of figuring it out versus just saying, isn't my patient, wait, here's what I want to achieve, can you work it out for me?
So it's great for things like converting units of measure like minimals, millimoles to milligram per kilogram or dosages that uses percentage instead of milligram per mil and filling in all sorts of fun fancy equations.Again, if it was for a drug that could kill my patient, like potassium or insulin or something like that, I'm doing it myself and then I'm using AI to double check me, which is actually really also very nice.
If I've done it and I instead of doing it three times like I normally would, I do it once and then I give it to Chad GBT and if it agrees with me, then I'm like sweet, I'm probably right.But for the other stuff that's not that critical, it really speeds up my day.It's also proving really useful for helping me interpret complicated test results quickly.
For example, the other day I was reviewing a case where the history included a low dose Dex test but without the interpretation, and I was trying to figure out if my patient could potentially be an adrenal dependent cushionoid.So I just gave the results to JHBT and I asked it to interpret the test and tell me if I could draw any conclusions about pituitary versus adrenal dependent Cushing's.
And it gave me a perfect answer, long explanation of why I couldn't really tell, including what my next steps would be.Something else I've been doing more and more is to help me summarize long history and find relevant information in that history.You know, when you take over a case and there's like four years worth of history and a lot of it relevant and most of it irrelevant to what you're trying to figure out, and it takes you 30 minutes to skim through it and to find those old test results.
And how did the skin respond to the steroid treatment?And have they done this and this and this?And even then, it's really easy to miss some critical detail somewhere.But for the last week or two, I am downloading those histories and giving it to chat, to summarize for me and look for critical information.And then I can actually ask it questions about the patients and it will like that, give me the answers.
So last week I gave it a very long history for a skin case and you know how long those can get.And then I could ask it questions like when was the last time it had bread?Did it respond to the bread?How often is this dog coming in for events like this?Have we done a diet trial?Have we discussed referral?
Has it had allergy testing?All those sort of things.Really, really useful to getting me ready for the consultation.Specifically great for emergency where you will sometimes get those long histories for the dog's.Been in hospital for three days at surgery, had 1000 drugs, and you can pop it in into ChatGPT or whatever your AIF choices and just go.
When was the last time I'd had cereelia or methadone?Can you check for any adverse reactions?Anything in the history that I should be aware of?Fantastic.And then finally, information finding.I know I said at the start that I'm wary of using it to source information, but there are ways to do this that makes the output more trustworthy.
Here's a few tips of what I'm finding really works for me.In case you didn't know, there is a button in JTPT that will send it specifically on an Internet search instead of just using its own knowledge.I'm trying to remember later to point it out when I share my screen.But really just in that bar there where you type in your question, you'll see there's a little plus sign and then there's a little globe, a little circular thing that says search the classic web symbol icon.
You click that and it's going to search the web instead of just using its own brain.So it's basically like getting AI to do a Google search for you.What I like about it is that it will give you the links where it's getting its information from.So you can go and double check it.Is it from a journal article or is it some random blog post?
The key thing here is to always ask for references, even if you're not searching the Internet.So even if you're using its own knowledge and you want to double check something, you can always say can you please show me your reference for this information?It's a great way to call BS on something and it makes sure that it's not just hallucinating and making stuff up.
Like the story I told you at the start of this conversation when it just made-up an article until I specifically asked it.Can you please show me a reference for the article?And that's when it went.I actually don't have a reference for the article. 2 Quick Pro tips for using AI is a search engine.You can't ask it to limit its answers to specific kinds of information like scientific literature or ublications from academic institutes, which will give you less information but more trustworthy info.
And again, make sure you say to it, show me your references.Otherwise it might just make up a whole new study for you.Again, the other tip is to use a tool called perplexity.There's a web-based app.You don't have to download anything and it is free or that you can use a certain number of searches for free per day.
So it's perplexity.PERPLEXITY, and that is an AI site that is trained just to search.I think it still uses ChatGPT as well as Claude and another couple of AIS, but somehow they've trained it to be very good at deep searches, whereas GPT can be a little bit lazy.
And then it can sometimes just consult four or five resources where Perplexity will often look at something like 20 resources.Perplexity is also better trained at being fussy about where it gets information from.Plus you can select different search modes.So instead of having to prompt it, say just look for academic and academic papers or something like that, you can select a mode that says search the entire web or just academic papers or social discussions like Reddit or things like that.
And like a GPT search, it also provides you with detailed links for every answer that it finds for you.So you can go and double check the ones that you really want to.Basically, Google is dead.Although having said that, most Google searches now is also backed by AI.Then there is the other way of using artificial intelligence and that is to put it in an assistant of some sort that uses just a very narrow defined of knowledge that you give it.
So in other words, you can make your own AI assistant of some sorts.Now I'm going to share screen to show you how I do it.There's a couple of ways of achieving it.I use this all the time.I told you earlier about the consensus statement part as well as the what I call the key to the vet fault where I have the show notes that people can search.
And that's basically how I use it.And there's a couple of ways of achieving it.So give me a second.I want to pause this and we're going to screen share and we will go through it.All right, so the first way of making your own AI tool.So if you've not come across Google's Notebook LM, then you're missing out.
I want to think, talk about a couple of ways that I do use.I have used and think that you could use Google Notebook LM.So a little bit of background, Google Notebook LM, as the name suggests, it is a Google product to use.It is free, but you do need a Google account.So a Gmail e-mail address really is what it comes down to.
You can use it as an individual user and there is also a paid version that gives you a lot more data.Basically, you can ask a lot more questions and use bigger sources.But even on the free version, I can create notebooks and I can share my notebooks with up to 50 other Gmail users or Google users.
So what is Google Notebook LM?This is your home screen.So basically, in simple terms, you can create notebooks of a particular theme and in your notebook you can provide resources that all fits with that theme.So for example, that could be a clinical topic.
You could even make it a patient.If you have a really complicated case, you can store all of its notes in this particular folder.And then you can chat to it, ask it questions, and ask it to draw on all of the sources of information that you provided to come up with sensible answers for your specific case.
So let me show you create a new notebook and upload sources here.As you can see, you can add stuff from your Google Drive, so Google Docs or Google Slides.You can provide a web link, which is really awesome.The really cool thing is you can provide a YouTube link.So if there's a cool podcast episode or a webinar or something that has content that you wanted to draw on, you can actually give it that YouTube link.
You don't have to even provide the transcript.It will search.I don't know if it makes its own transcript or searches the audio again.Those two I can tell you how to push the buttons.Chuck that in there or copy text.Or you can add PDFC as well supported files, PDFSTXTXD files, markdown and audio.
That's New South.Audio files are New South.Any audio file fire out.This is huge.So you throw all of your resources in here and then you can give your notebook a name.So I'll give you an example of something I made for my own use in clinic, nutritional support in hospitalized and critical patients.
So I hate working out feeding plans.It's boring, but it's important.And because it's boring, I don't do it as well as I should for my hospitalized patients.So what I did, I went on a Google hunt and I looked for resources, valid resources that I like, PDF files, textbook chapters.
And I checked all of this, added this all to my Google notebook.Now again, I could even add a webinar or an audio file or a YouTube link in there.If I went to a conference somewhere and they had a live stream that I've got access to, I can put that in here.It has these automated things that are pretty cool.
So you can make yourself a study guide.Just going to search all my sources.OK, so this is my study I've done.So you can make yourself quizzes and with an answer key.So if you're studying for something, I think it's a super useful tool.
Let's go back a briefing drop.Let's say I have a complicated patient, as I said before long history and I want just a summary for my patient.It is going to search that and it's going to make a nifty little summary for me.So here is my summary.So it's taking all those sources and turning it into a single document, a report of sorts.
And as you can see, it is quite detailed when I compare Google dot PGLM to ChatGPT, because I'll show you, you can do similar thing in ChatGPT.Google is really detailed and not a lot, not super chatty.It's just facts.And if that's what you want, it's great.
If you want somebody who adds a little bit of flair or chooses a personality or talks to you about things, chats to you about things, chatty BT is going to be the one that I'm going to show you how to make a similar sort of a thing, but facts only.And it's kind of suits our profession really.So now the cool thing that I really like about this is now that I've created this nutritional support and hospitalized patients, I can actually ask a case questions.
So I'm going to say to this guy actually pre prepared prompt.Let me quickly find my prompt here.So I've said work out a feeding plan for a 14 kilogram dog with sepsis anorexia for three days feeding by an energy tube feeding Hills AD.As you can see in my folders here, I have included in addition to feeding guidelines and all the RAR requirement things and all that, I have got the contents for Royal Canon Recovery as well as AD with chicken.
Those are the 2 main ones we use where I work.So if I want to work out how much to feed an animal, I can just tell it which food and it has all the information it needs on that food.So I'm going to go send it.It's going to take about 1020 seconds to work out my plan for me.
All right, there we go.Based on this was provided, here's a suggested feeding plan for your 4 concluding render.Then it's going to go through calculate the resting energy requirement and it shows you the formula that it's using so it thinks through the plan for you so you can see what it's thinking.And all these little numbers are references from the notes, which is amazing.
It shows you which documents in your folder it is drawing on to give you the information and you can go double check it if you want to figure it out for yourself.And then it is taking the Hills 80 information and working it out, chunking it into that formula there.So then we have a detailed plan that says day one provide 1/3 of the calculated daily calorie intake with two third water dilution and it's going to work it out for me, blah blah, blah.
Day 2, day three, plan, feeding method and frequency.So for a 14 kilogram dog, this would be maximum of 140 to 168 mils per feeding.So it's a completely detailed workout plan, much better than I've ever done in my entire career.It talks you through tube management and important considerations.
I wonder if I could summarize this because that's actually a little bit too much.Let's ask.So I'm saying, can you summarize this into a quick reference guide for my nursing team, just showing volumes and frequency for the next 3A's shoot.And it's done exactly that.
A very short thing that I could print out and stick on a cage.Isn't that amazing?So that is hills.Not Hills, Google's notebook.I use chatbot for a lot of things and I'll show you how and why in a second.But for this, I'm even thinking for practice protocols, you know, if you have protocols around finances or client complaints or anything like that.
Oh, the other thing that I didn't mention, kind of don't want to mention it because it's too cool and it's kind of put me out of business one day as a podcaster.But you can, you can generate a deep dive conversation or an audio guide.So you can turn all this into a basically to a podcast.
So you can make yourself a podcast on any topic you want and say, I want to learn a little bit more about sodium problems and dogs.Chuck it a bunch of information, say generate a deep dive conversation.And it's actually not bad at all.I don't know if you can change the voice now, but you can listen to it in your car and update yourself.
OK, the next one and the one that I use most often is ChatGPT.I'm sure there are other softwares out there, but I'm comfortable with ChatGPT.I've been using it for a while and I get great results with it.So that's what I'm going to show you.So I am on ChatGPT on the paid subscription that is about 20 U.S. dollars a month, and I'm going to show you why I am on the paid plan.
Why like that versus just the free plan?The free ChatGPT is still an amazing tool, but for what I want to achieve with it and for how I think you can use it in practice, I think there's definitely value in the paid plan.So the first and main reason why I prefer the paid plan is something called my GPT.
So let me show you what I mean.So if you log in to ChatGPT, usually you'll have just like a chat history there on the side.But if you look at mine over here, I have a list of fixed things there.So these are all assistants, AI assistants that I made for myself using a specific data set for the GPT or for the AI tool or assistant to use as its brain.
So very similar to the notebooks that I just show you in Google in that you can create a task specific tool that draws on information that you provided.So why do I use ChatGPT rather than Notebook LM?Why pay when there's a free version out there?There are a couple of reasons for that.One is just kind of convenience.
I'm really spending a lot of time doing other things on ChatGPT and my ChatGPT account knows me and my work.I've been using it for two years, so I don't have to keep telling it who I am and what I do.It remembers it.So it's kind of tuned to me and my style and my work.The next reason is multifunctionality.
With ChatGPT, I could send my queries to the Internet.It can write stuff for me.It can help me problem solve using its own brain versus just the stuff that I provided it, versus the notebook, which is a lot more specific and a lot more robotic almost on ChatGPT.
When I make a specific GPT, so in a specific assistant, I can also tell exactly how I wanted to respond in what kind of format and what the exact output should be every time that I interact with that specific tool that I've created.And then the last thing is the ChatGPT app.
It makes how we interact with it a lot more versatile.So I can type something on my Mac and then I can pick up that same chat on my phone on the app and I can talk to it like a person or I can show it a picture or it could read text for me.I can record onto it, or I can even ask it to make me a picture.
It's image creation in the last week or two has just skyrocketed to a next level or make me a graph or summarize.So I just find it's a lot more versatile versus notebook LM, which is very task specific.So let me show you how to make your own tools in GPT.
But before we start with making a GPT, I'm going to give you a pro tip that is around prompting.So prompting a GPT correctly will make a big difference to how it performs.So you need to tell it exactly what you do, exactly what the outcome should be, what you avoid, all those sort of things.
And doing it in a sensible structure, detailed way is actually quite time consuming.Not days and days, but you're going to spend a good half an hour or an hour or so creating a great detailed prompt.So here's my tip.Before you start building your new assistant or tool or GPT, get Chad PT to write the prompt for you.
This way you can just brain vomit all the things you want to achieve.And you can even do it in a conversation instead of typing it.So again, on the phone app, you can do it on your Mac as well.Hit this little speech button.So if you look at my screen, that over there will let me talk to JTPT instead of typing.
And I would just start saying, hey, I need help with creating a detailed prompt for a new GTPT that I'd like to create for myself.Here's what the tool needs to achieve for me.I'm going to provide it with this kind of information.I wanted to answer me only using that information or alternatively, I want it to be use my information, but feel free to add to the answer using your own knowledge or knowledge from the Internet.
You can pick what you want to do add to it to say here's an example of my ideal answer.Here's a template that I'd like you to use every time.Avoid doing this.Never say this.Keep your tone friendly or make it highly professional.Anything really that will help you formulate that prompt nicely.
So again, brain vomit.Talk to Jackie BT, give it information overload.And then you say, can you turn what we've just discussed into a detailed prompt for me that encapsulates all of this in a structured way that will give me reliable outcomes every time.And boom, it's going to make you a lovely prompt in 30 seconds.
And just copy that and use that as your instructions for your new GPT assistant.So now we have our prompts.Here's how we build our own GPT.I'm going to say explore GPTS.So I'm logged in here on my paid subscription.I actually have a team subscription and I'll tell you why in a second.
So explore GPTS.You are going to say great.And there's two ways of creating either configure is where you manually do it yourself.So if you have that lovely prompt of yours, you're going to give it a name test bot and give it a one sentence description.
So let's say practice soaps and I'm going to say store and recall all of our standard operating procedures, blah, blah, blah.You get that sort of terrible.That's why I like to talk instead of type.
As you can see, I'm dyslexic when I type.Then this is where you're going to copy and paste the prompt that you produced earlier.So bang it in all there.Next is your knowledge base.So whatever you wanted to draw its information from, you're just going to say upload files.So you're going to click on that and drop and drag or download from your computer wherever it is, multiple files, long PDFs, and that is going to use that.
Then it's going to look at that information every time you call on that specific task.So for example, I built myself an e-mail assistant.So I have a bunch of emails that I write fairly frequently.Same sort of e-mail, but not the exact same e-mail.I used to use an e-mail template and then just fill it in.
Now I've given all those templates to a GPT and I all even have a name.So now instead of writing that e-mail, I will just go to my GPT and say, please write an e-mail to Jane.She's a new guest on the podcast.Please use my Thrive podcast episode new guest template.
And here are the topics we get.So I'll fill in all the extra details and boom, it spits it out for me and an e-mail that is almost good to go.So anything like that.This is where you're going to upload your files capabilities.Do you want it to be able to search the web If I want mine to just look at my references.
So my clinical tools that I built for myself, for example, the consensus statements assistant, I switch off website as I don't want it to go looking for its own information.Canvas I'm not going to get into.It's just the format that it gives you the answer.Canvas is a format where you can Co work.
It's almost like a Google doc that you work under the document with GPT.But if you just want simple answers you can switch that off.Do you want to be able to make pictures or not?Most of mine and do not.Code interpreter and data analysis.I always switch on because I really wanted to be able to look at that.
I think that's the right answer.That's what I always select and it works for me and that is really it.It is as simple as that.That's the one way of doing it.Or you can do create and it'll just talk you through it.That's what I did before.I actually find I'd get there faster if I just use the configure option, but using it that way is maybe a great way to start.
But if I know exactly what I want, I don't need it to hand hold me and namby pamby me through the whole process.And once you are done with it, you just say create and then it's going to ask you about sharing options.So who do you want it to be a tool just for you?Do you want it publicly available or just for people with the link?
And then really is as simple as that.You can always go back to it.So I'm going to go back.So if I say so let's say my e-mail assistant that I talked about earlier, I can click that down button and I could just click edit it and I can change my prompt there or I can add more files to it.
I can just upload more files there and change anything and just update it over there.Or you can share it as well.There's the share button.And is it available just for myself or for anybody else?So this one is obviously for me, my personal e-mail 1 And when you're done, you just say update and there it is ready to roll for the next round.
Quick note on the team subscriptions, why I like the team subscription.I have an assistant that works with me and several of these tools I want her to be able to use as well.So I make a tool and then I just make sure that she has access to it as well.That means we don't have to double up.So for example, in a practice scenario, if you have the practice GPT or the practice ChatGPT account, I think it's about treating you as per person per month as well.
And you don't necessarily have to sign up everybody.You just need to make sure that different people can log in from different places.With a team account, I can keep a lot of this private and my chats stay mine.So I can still use my team account as a normal ChatGPT.
It doesn't have to be just be in this, I can still ask it problems about my psychological problems and use it as a therapist.And I don't necessarily want to see the rest of my team see those chats, so those will remain private, but anything that I share with them they'll have access to as well.One other cool thing, I'm pretty sure this is part of the normal paid plan, not just the team subscription.
But either way, projects is a pretty cool tool as well.So it's kind of like GPT, but not so much for repetitive tasks.It's more for a finite or this is the way I use it, finite start and end date project that I need to use, but I need to keep a lot of information in the same little box so they can refer back to it later.
So you just make a new project.And so for example, for vets on tour for Monica, I've made a project over there.So these are all my different chats within that project and the same thing.I can open my project and I can add files to it.So speaker BIOS and information about the venue and flight tickets and all those sort of things.
And that will remember that.So I can just instead of having an open chat, I just go into my file and I'll say, hey, can you tell me what time I'm flying again?Or what did we say the pricing was?Or can you please create a structured program based on the, this is pretty exactly what I've been using it for.It's a really cool way.
And this is not shareable, unfortunately, actually be quite cool to share it with the rest of the team.But for now, this is a private thing just for yourself.So in terms of how I think you can use this, things have obviously there's the clinical tools, the ones that I've built for the team, I've built in a different platform still with open AI, but that I've built for my subscribers.
So like the key to the vet vault, which has all the show notes or the consensus statements, you won't see them here because they actually hosted somewhere else.That's a conversation for another day.But if I owned a vet practice, if I was working full time and vet practice or owned on things that I think about immediately, how this would be super useful for a team would be I mentioned soaps there.
So let's say standard operating procedures like how do we deal with this situation?What's our policy on refunds?How do we log into the server if we get logged out?What's our contact at the drug supplier?What's our backup supplier if we can't order Prednisolone from supplier X?
How does the CBC machine work?What are the error codes?All information onboarding procedures, where you order a name tag from, all that sort of stuff that you know, 10 people in the practice know little bits of that you can Chuck it all in one GPT that all the team has access to or maybe on one computer at work.
So instead of having to call the practice manager and say, hey, June, I don't know how to do this, who should I call?You can just go in there and say, I have a client, brought some medications back.What are our practice protocols for returns?But it's going to pop it up for you and tell you.And then you can say, please write a polite e-mail, Please draft a polite e-mail to Mrs. Jones to say that unfortunately we can't take back the Prednisolone we dispensed a year ago.
And here's why.That's what I think.I mean, I have this idea of even having, for example, on equipment in the practice, putting QR codes that link to this particular GPT.There are ways of doing it, but my mind explodes when I think about all the possibilities for this.
And once you start playing, you'll have the same experience if you're not using it already.And I think just the most useful thing, thing that I get stuck on or that I'm thinking more and more because you're going to be seeing ads for all these AI tools.I can do this.
I can take your notes, I can do your finances.There's all these amazing AI tools coming onto the market, but they're all cost and they're all subscriptions.And by the time you sign up for 10 of them for 20 bucks a month or 10 bucks a month or something, you're going to be paying a lot for all your tools.So why I like making my own things and I say make makes me sound really clever.
As you just saw, it's super simple where I can make my own things.I have one subscription and I can create my own tools that are task specific for me, for my needs.They might not be as whiz bang as some of the other ones, but it does the job for me and it is one single subscription.
So basically I'm being a cheapskate and I can customize things to fit my exact needs.Otherwise, there's a lot of redundancy in some of those things you're going to pay for.So basically my GPT for any repetitive task or for any Bank of information that is large and hard to search, GPT can do it for you much, much easier.
So I said at the beginning, I'm going to wrap up with a little bit of a thought experiment, a philosophical discussion about going forward, what our role as vets are going to be.So up until now, we were the guardians of information.We were The Walking textbooks and the libraries of experience and that is being taken over by a very large brain that knows everything about everything and can rapidly simulate information from many resources and come up with a sensible answer.
So what's our job going to be?Does veterinary science still need veterinarians?Think about it.Nurses and techs are increasingly already taking over the day-to-day small procedures like taking blood and making smears and doing X-rays.So a good AI can take a history and tell a nurse which test to do or what to look for, ask for the results, and then come up with the answers.
Maybe not today, but soon.So what am I going to do when I go to work?I'll give you my opinion, but I would love to know yours.So here are some of the things that I think about.The first thing is, will we still need to have knowledge?If AI knows everything, then do I need to know shit all right?
Do I have to spend so much time at university learning a bunch of facts?I think the answer is yes.Not so that we can be the source of information, but more so that we can be the curators to decide what's relevant.You need someone who can contextualize that fire hose effect that is currently just coming at us all the time and is only going to get worse.
You need to be able to shuffle the puzzle pieces around to make them fit in a picture that make sense in that scenario for that patient.And as I said before, to get good answers, you need to ask the right questions, and you can't do that without expertise.But what we don't need is to remember a lot of facts anymore, which I think is very relevant when it comes to how we select our future veterinarians.
Until now, you get into vet school based on your high test scores, which in turn is based on how good you are at memorizing and regurgitating information, right?But that skill is now rapidly becoming obsolete.But what an opportunity.What if we can select for curiosity, for problem solving skills, leadership skills, for lateral thinking and technical skills, surgery, making stuff, building things.
And we can even start selecting for things like resilience and empathy and generalized human skills like emotional intelligence.And Speaking of those human skills, that study that I mentioned right at the beginning, where AI outperformed the doctors in case management, well, in that experiment where the AI fell flat on its face, where it bombed out was when it came to empathy.
For now, empathy, reassurance and trust building remain uniquely human skills.So I think that one of our most, if not the most important role for veterinarians going forward will be to be more human, to act as that human machine interface.But here's the thing, as I alluded to a minute ago when we were talking about how we select feature vets, any of us are not that great and empathy, reassurance and trust building because we've been selecting for the humans that are a little bit more robot.
So for those of us currently in the profession, this might be the time to really focus your CE hours on improving your human skills, your emotional intelligence, your communication skills, and your leadership skills.Because the vets with those skills combined with a good level of comfort with utilizing AI will be the ones that shine.
And if not that, then definitely work on your manual skills like surgery or ultrasound.I think we've got a few more years before that's replaced.There's one more human skill that I think will remain ours that is combining ethical thinking and context.AI thinks algorithmically.
There's the question, here are the facts.That's the right way of doing it.If A&B leading to C, if Doug has spinal pain, then it needs an MRI and CFS analysis, CSF analysis.But does that make sense for that geriatric patient to have an MRI?And what do we do with the cost constraint owner?
Those type of questions, which we all know is like 90% of what we do, will still fall on us with our contextual thinking and our empathy to figure out.Here's something that I wonder about when it comes to AI taking over a lot of the work that we have to do for now, can we make more money?
We being everyone in the profession, if we can outsource half of our duties to AI at $20.00 a month, our efficiency and our productivity can massively scale.So one vet with two or three skilled techs and a good AI will be able to get the same amount of work done as a.
Team twice that size.And any practice owner knows that wages are the biggest expense in running a practice.So if we have our biggest expense, how does that change the business model?I reckon there are a few options here.A.We charge the same, revenue stays the same, but we have our team sizes and everyone's wage is double.
As an employed vet, that is obviously my pick.Alternatively, everything stays the same and the practice owner double s profit.Of course there's a third option where the cost of running a practice decreases and we decrease what we charged to make vet care more affordable.Or maybe it goes the opposite way as pet owners increasingly use AI to solve many of their pet related problems without our help, so they sit at home and ask GBT.
So we lose all of those consults that end up with a just keep an eye for a day or two.Or here's something for nausea, or some simple solution and something that doesn't end up in hospital with major diagnostics.So we end up losing a lot of our revenue.So we just see the ones who need diagnostics and who need procedures and we charge more for that to compensate for the loss of the others.
Basically, by the time that somebody comes to you, they've sort of self selected.It's almost like when you're a specialist and the Jeep is already seeing something and by the time they get to you, they know what they're up for.They know what they're going to have to do and they are sort of mentally ready to fork out the money or get their insurance involved.I don't know which way it's going to go, but it's definitely worth thinking about.
And I think a really potentially a great time to change our profession into the profession that we want to be a part of and then we want to see and the one that we can be proud of.And then there are some opportunities outside of clinical practice that might become more of a thing.First, we said that AI knows all the stuff and is up to date with what is new in the world of vetting, but it only knows stuff that we have figured out.
So research will be absolutely essential.Conveniently, many aspects of research will become or are already much easier, making it more accessible to those of us outside of traditional scientific institutions.Suddenly you can search and process large amounts of data in seconds, and you can have a very smart assistant who can help you with things like statistics and even writing papers for you.
But what we do need more of is curiosity to want to figure stuff out and the drive to do it.I personally think that the rate of progress on veteran science, like in medicine, will exponentially increase, ironically making the need for a way to keep up to date with all of this progress even more acute.
OK, I think that's me done.This episode has taken me three times as long to put together than I originally planned.And it's also three times as long because there's just so much to talk about.And every time I thought it was done, something new and shiny would pop up that would make me add something or take something out.
But we're going to pause it there again.The more I talk to people about this, the more I realized that we are all in the same boat and we all have a bit of FOMO or a lot of FOMO because we know that it's something big, but we can't keep up.So maybe an AI for Vets online community where we share our trials and tribulations and have a think tank for new ideas is a solution.
If you think it's a good idea, send me an e-mail at info@thevetvault.com that's info@thevetvault.com or fill out our contact form at the vetvault.com.Thanks for listening and we'll see you next time for a normal episode.Back to normal programming with me and a guest.Before you disappear, I wanted to tell you about my weekly newsletter.
I speak to so many interesting people and learn so many new things while making the clinical podcast, so I thought I'd grant a little summary each week of the stuff that stood out for me.We call it the Vet Vault 321 and it consists of three clinical pills.These are three things that I've taken away from making the clinical podcast episodes.
My light bulb moments.Two other things.These could be quotes, links, movies, books, a podcast highlight, maybe even from our own podcast.Anything that I've come across outside of clinical vetting that I think that you might find interesting.And then one thing to think about, which is usually something that I'm pondering this week and that I'd like you to ponder with me.
If you'd like to get these in your inbox each week, then follow the newsletter link in the show description wherever you're listening.It's free and I'd like to think it's useful.OK, we'll see you next time.