The Forward Thinking Dental Podcast
SEO METADATA
SEO TITLE:
Practical AI for Dentistry: Save Time, Achieve More, with Dr Daz Kasperek | PDP260
META DESCRIPTION:
A beginner's guide to AI for dentists: the RCT prompt, personalising ChatGPT, AI notes, CPD admin, choosing a model, patient data and a free smile tool.
URL SLUG:
practical-ai-for-dentistry-save-time-achieve-more
PRIMARY TOPIC:
A practical beginner's guide to using AI in dentistry
SECONDARY SEARCH TOPICS:
- How to write better AI prompts for dentists
- Personalising ChatGPT with custom instructions
- AI clinical notes for dentists
- Using AI agents to automate CPD records
- ChatGPT vs Gemini vs Claude vs Grok for dentists
- Patient confidentiality and GDPR when using AI
- Free AI smile simulator for dentists
- NotebookLM for dental education and journal clubs
PRACTICAL AI FOR DENTISTRY: SAVE TIME, ACHIEVE MORE
THE PROTRUSIVE DENTAL PODCAST, EPISODE PDP260
HOST: DR JAZ GULATI
GUEST: DR DAZ KASPEREK
EPISODE SUMMARY
On this episode of the Protrusive Dental Podcast, Dr Jaz Gulati is joined by Dr Daz Kasperek, an Academic Clinical Fellow doing a doctorate in endodontics at the University of Liverpool, whose research looks at AI in endodontics. Assuming zero prior knowledge, they cover what a prompt is, why you should personalise your AI with information about yourself, and Daz's "RCT" prompt framework: role, context and task, with context being the most important.
They discuss AI as a diagnostic second opinion, the risk of hallucinated references, AI agents for admin such as compiling CPD records, projects and custom GPTs for repeatable tasks, why speaking beats typing, AI notes, and how ChatGPT, Gemini, Claude and Grok compare. They also cover patient confidentiality and consent, NotebookLM for learning, and what AI won't replace. Jaz's Protrusive Dental Pearl is the free Dental Disrupt Smile Simulator, created by Dr Jason Lipscomb, which he demonstrates live.
IN SHORT
AI can save dentists significant time on notes, letters, admin and learning, provided you personalise it, give it plenty of context, check its output and never share patient-identifiable information. Daz's advice is to try the main models, pick one to stick with, and remember that you remain responsible for everything that goes out under your name.
KEY TAKEAWAYS
- Jaz's pearl: try the free Dental Disrupt Smile Simulator custom GPT by Dr Jason Lipscomb.
- Personalise your AI with information about yourself, and use your own account rather than sharing one.
- Daz's RCT prompt framework: role, context and task, with context the most important.
- AI can hallucinate references, so always check its output; you remain responsible for it.
- Projects and custom GPTs save repeating the same instructions for repeatable tasks.
- Speaking to AI rather than typing is faster and gives it richer context.
- Never put patient-identifiable information into a chatbot, and get consent before uploading patient photos.
- Daz believes AI won't replace the human connection, empathy and artistry of dentistry.
QUESTIONS ANSWERED IN THIS EPISODE
1. What is a prompt, and how do I write a good one?
2. Why should I personalise my AI account?
3. Can AI help with clinical diagnosis?
4. How can AI agents save time on admin like CPD records?
5. What are projects and custom GPTs?
6. Which AI model should dentists use?
7. Is it safe to put patient details into AI?
8. What won't AI replace in dentistry?
IN THIS EPISODE
1. Introduction: making AI in dentistry tangible
2. Protrusive Dental Pearl: a free AI smile simulator
3. Meet Dr Daz Kasperek: endodontics and AI at Liverpool
4. Where are we with AI in dentistry? Diagnosis and second opinions
5. Hallucinations, prompts and personalising your AI
6. The RCT prompt: role, context, task
7. AI agents, admin and automating your CPD records
8. Projects, custom GPTs and why you should talk, not type
9. AI notes, and the "what if I forget to say it?" objection
10. Live demo: the Dental Disrupt Smile Simulator
11. Which AI should you use? ChatGPT, Gemini, Claude and Grok
12. Patient data, confidentiality and consent
13. What will AI replace? Admin, education and NotebookLM
14. AR glasses, implant robots, and what AI won't replace
15. Daz's research, how to find him, and outro
CHAPTER 1: INTRODUCTION: MAKING AI IN DENTISTRY TANGIBLE
JAZ:
AI is everywhere. It's a massive buzzword and has been for a few years now. But can you believe that, at the time of recording, 86% of the entire world have had no interaction with AI, and only about 0.3% of the population are actually paying for AI? I have a monthly AI subscription, and we'll talk about that as well.
So I'm one of the 0.3% of the population. The reason I'm mentioning it is that we've barely scratched the surface. This episode is about making AI in dentistry tangible. My guest, Dr Daz Kasperek, and I assume you have zero knowledge of AI. We cover what a prompt is, what an AI model is, and which one you should use.
Is it safe to put patient details into AI? And more importantly, how can we actually start to get our time back, be more efficient, enjoy our lives more and have an easier life through the power of AI? Over the next few weeks, we'll talk about more advanced ways to use AI. But this episode is very foundational, because many colleagues say, "I'm a busy clinician. I come home to my family. I haven't had time to properly give AI a go."
Even though I've been banging on about AI notes for two years now, most of the people listening to or watching this are still not using AI notes, because they just don't know how to get started. Well, let this episode be your guiding light.
Let this episode pave the way for you to get into AI, because trust me, it's worth getting into. AI is super powerful. It can do bad things, but if we use it ethically, it's going to save us time and reduce our stress, and we talk about all the different mechanisms in this podcast. So this is for every single dental professional who hasn't really given AI a shot yet. You'll understand it a lot more after this episode. And for those of you who are already deep into AI, you'll hear some cool things Daz is doing with AI right now.
CHAPTER 2: PROTRUSIVE DENTAL PEARL: A FREE AI SMILE SIMULATOR
JAZ:
Every PDP episode, we give you a Protrusive Dental Pearl, and today it's a free AI tool that does something really cool. You can feed it an image of a smile, like a standard smile photo we take for documentation.
The AI does its magic and gives you the same image back, but with a smile makeover. It's like one of those instant smile makeovers. I remember five years ago, I used to send images like these to a dental lab, and they'd use some software to give you a few smiles and email them back.
Even those were a bit sketchy, but we used to pay 50 pounds for that service. It was something you could show a patient: "Hey, this is kind of what your smile could look like." Now we don't need to pay anyone, and I'd certainly say you don't need to buy any expensive software that does this.
There are smile makeover simulation programs out there, and I'm telling you, you don't need them. This is free, and I'm going to share it with you. I want to let you know that I didn't invent this. Even Dr Jason Lipscomb, who presented and curated it, didn't invent the technology, but he did the hard work of making it into a model, a project that can be embedded into ChatGPT.
So I don't want to take any credit for this. Credit goes to Dr Jason Lipscomb, who's doing and sharing so many cool things with AI in dentistry. I'm going to find his Facebook group and give it a shout-out. The Facebook group is Dental Disrupt AI and Automation, currently with 550 members. He posts daily uses of AI in dentistry and beyond.
So a massive shout-out to you, Dr Jason Lipscomb. Thank you for creating that resource. I'm just passing it on to you. Wherever you're watching or listening, just scroll down and you'll see it in the show notes. If you're on Protrusive Guidance, I'll put a clear link to it. If you're on Spotify, it'll be in the description.
If you're on Apple, scroll down to the description, and there'll be a link there. Let me explain how the link works. This is a custom GPT, which means that if you're logged into ChatGPT and click the link I'm going to give you, you kind of install this custom GPT in your ChatGPT. Whenever you're having a chat with ChatGPT, you'll see this tool in the side panel.
It's called Dental Disrupt Smile Simulator. And did I mention it's free? It's frickin' amazing. I've been really impressed. In fact, I've been doing an experiment, giving it the before photos of real smile makeovers I've carried out, and the simulated after image it gives you is pretty on point when you compare it to the real after.
So it's not too far-fetched. It's actually quite accurate. I've been really impressed. It harnesses the technology ChatGPT has, but Jason has done the hard work of prompting it and giving it instructions for smile design. So once again, the link is freely available. Just scroll down wherever you get your podcasts, and I'll make sure the link is there for you.
Hope you enjoy that tool. You're welcome. Now let's join the main episode, and I'll catch you in the outro.
CHAPTER 3: MEET DR DAZ KASPEREK: ENDODONTICS AND AI AT LIVERPOOL
JAZ:
Dr Daz Kasperek, welcome to the Protrusive Dental Podcast, my friend, a long-term Protruserati. I've been seeing you since the Facebook days, back when we had a Facebook group. You've made so many Protruserati happy with the cross-polarising filters you made for them.
It's so nice to see you, not in the flesh, but in front of me, discussing something that's changed my world, and I know it's changed yours. Daz, tell us about yourself, why you're now studying endo at Liverpool, and a little bit about your journey. I always believe in the journey, how you got here, and how you're now doing research into AI in endo.
DR DAZ KASPEREK:
Yeah, first of all, thank you so much, Jaz, for having me on. I've been a long-time fan of your podcast. I still remember, back in the day when you started, that first episode on Singapore, I think, where you did the intro with that American guy.
It was so cool. I remember listening to you. I graduated from Sheffield in 2018, and I think at some point you were a DCT there, so I might have seen you around when I was a student. Right now, I'm doing my doctorate in endodontics at the University of Liverpool.
JAZ:
Can I ask what that actually means? I'm familiar with a master's, and with an MClinDent. How is a doctorate different? Is it significantly different?
DR DAZ KASPEREK:
It's still a specialty training pathway, but there's a lot more research. A big part of our course is a research component-
JAZ:
Like PhD level?
DR DAZ KASPEREK:
Yes, precisely. I'm also an ACF, which stands for Academic Clinical Fellow, so I'm doing my endo specialty training and the doctorate concurrently. A big part of my specialty training is a big research project. Everyone doing the DDSc has a proper three-year research project that requires a full dissertation, a thesis defence, et cetera.
It's a very big commitment, but I'm honestly so happy at Liverpool. I have the best supervisors, like my professor, Professor Jarad. Shout-out to him. He's amazing.
JAZ:
Oh, legend. He taught my wife back in the day as well, when she was a Liverpool girl.
DR DAZ KASPEREK:
Oh, wow.
JAZ:
Amazing. He's very well known.
DR DAZ KASPEREK:
Yeah. Honestly, it's such a nice place to be and extremely supportive, so I'm really, really happy. A lot of my research is in AI, and also augmented reality. I'm looking at all the digital technologies and how we can integrate them into endo.
It's such an exciting space, because literally every week there are new things coming out. You can go on holiday for a week, and it takes you a few hours to catch up on what you've missed. So it's incredibly exciting to be-
JAZ:
How exciting that you're at the interface of endodontics and AI. I can hear the enthusiasm in your voice, and that makes me so happy. Look, AI has changed our world completely, right? And it's an interesting thing. I remember something from when I was 12 years old.
I was in Year 7, reading something in IT class. It was the first chapter in the book, back in the day when we used big old desktop machines, Dells and that kind of stuff. And it had a chapter about the disadvantages of the internet.
One of the disadvantages of the internet was that it would create a big divide in the world, with a percentage of the population that's information-rich and a percentage that's information-poor, and the information-rich would be massively advantaged.
It creates inequality. We're in the same space now. Most of the world has now caught up with internet access, especially now, thanks to what Tesla are doing with Starlink and so on. There's internet pretty much everywhere now. But now it's about who's using AI and what they do with it, and who isn't.
You can see that in the next few years, AI is getting so crazy, and the time it'll take for people who are behind to catch up... AI has allowed me to come home to my children, because it saves me so much time. In the podcasting world, we obviously launched AskJaz and made everything accessible.
Years ago, before AI took off, when we were about a hundred episodes in, someone messaged me saying, "Jaz, your podcast is so rich in information. You should create a resource, like an ebook, with all the stuff in it."
With AI, it was amazing to be able to just bring it all together so you can chat with every single piece of content ever. And we're barely scratching the surface, right? I just wanted to share that perspective and excitement. So where should we start? Where do you think we are now with AI in dentistry? I think that's a good overarching place to start.
CHAPTER 4: WHERE ARE WE WITH AI IN DENTISTRY? DIAGNOSIS AND SECOND OPINIONS
DR DAZ KASPEREK:
I think you absolutely nailed it. I think we're still in the infancy days. ChatGPT has been out for maybe just over three years and has become mainstream, and every week, like I said, things are improving, because all these big companies, Google, Anthropic, OpenAI, are constantly competing with each other.
So they're all pushing themselves to get better and better, and as consumers we benefit, because it's healthy competition. Everyone wants you to use their AI rather than the competition's.
That's why the models are becoming smarter, easier to work with and more intuitive. With every iteration of every new model, I see little nuances, which you might not have picked up if you haven't used it enough,
being ironed out behind the scenes. And with more and more people adopting it, it's just going to be mainstream. It's going to be like the internet, exactly like you said. It's just going to get better and better and more integrated.
Like you said, if you're not using AI, you'll be left behind, because everyone else has the advantage of basically having an expert in their pocket on anything, precisely tailored to them. That's the key thing. It's like having a tutor who's omnipresent, and you know they have expert knowledge. That's what's so amazing about it, and I think it's only getting started. There are specific dental AIs for things like caries diagnosis, and there's research on CBCT segmentation.
There's a lot being done in medicine, where AI helps clinicians diagnose patients. That's a very exciting space. These are all tailored models, trained on very specific data with very specific instructions, but one of the things-
JAZ:
I'll give you an example, Daz. About this time last year, I had a spontaneous pneumothorax. Literally, my left lung decided it wanted to burst. The doctor had no idea what was going on. She said, "Maybe we should send you for a CT scan. I don't think there's anything, but we'll send you." Actually, no, it was a chest x-ray. And when the chest x-ray popped up, within a second it said "pneumothorax", and a diagram came up highlighting the area of the pneumothorax.
It was amazing to see AI being used in medicine in real time, on very tailored, specific stuff. That's how it's revolutionised diagnosis. And where does dentistry start with AI? Well, it started with diagnosis, right? Because diagnosis comes once you've gathered the data, which we have to do as clinicians anyway.
It allows us to eliminate our blind spots and tap into models and parts of a brain that we just don't have. We're not intelligent enough on our own, but we can now give it to AI to give us a differential diagnosis, and then use human judgement.
So no wonder diagnosis has taken the lead. Tell us what you've seen as good case studies at the moment, and what's potentially not there yet but is something to be excited about in the future, when it comes to diagnosis and AI in dentistry.
DR DAZ KASPEREK:
I think, for diagnosis, it's exactly like you said: it's like having an expert there for a second opinion, right? Because ultimately, you're always going to be responsible for your clinical judgement.
AI doesn't have a GDC number, but you do. It's always about having a different perspective. Patients give us a lot of information from the history, from the clinical findings, signs and symptoms, and sometimes we get patients where we're not quite sure what's going on.
You order a different scan, a CBCT, and you may get them to see another specialist, and it's exactly for that purpose. You can even feed in information like the patient's signs and symptoms, anonymously. But that's a topic we also need to touch on.
You can describe the signs and symptoms and your clinical findings, and AI has been trained on all the knowledge of the internet: all the websites, books, et cetera. As humans, we don't have the capacity to have that much knowledge available to us.
It gives you the ability to really hone in, and maybe get the AI persona to tell you, "Okay, maybe this could be it," and then you're the final boss who connects it all together, and ultimately you make the decision.
I think that's really, really crucial for diagnosis, because there may be something the patient told you that you missed or completely dismissed, but which is very specific and mentioned somewhere else that you're not familiar with. So I think that's the key thing right now.
CHAPTER 5: HALLUCINATIONS, PROMPTS AND PERSONALISING YOUR AI
JAZ:
The danger for me with AI, and let's use ChatGPT as the use case, is this. Obviously, you don't want to feed in any patient data, and we're going to talk about that. But say there's a scenario where a patient has a sinus tract and so on: "What's the percentage chance we can save it?
"What's the prognosis?" You feed it some more data. It can even read an example radiograph and give you information. But my worry is that when I've discussed various other things with ChatGPT and said, "Give me the evidence for this, give me the evidence for that,"
it literally made up papers that don't exist, right? It hallucinates. That's called hallucination. That's one of the reasons I created AskJaz: if it doesn't know something, it says, "We don't know." It only uses information that's been verified, from the mouth of me or an expert who's been on,
rather than scraping from Reddit and hallucinating. That's one of the reasons I was so worried about it, because a lot depends on the prompt. Now, you and I know what a prompt means. But some colleagues don't. One messaged me on Protrusive Guidance and said, "Look, I'm 30 years qualified.
"I feel like a dinosaur. I don't even know where to start with AI." So for someone like that who's listening to this conversation, we need to say what a prompt is. A prompt is what you feed the AI to help it unpack the problem. Depending on the prompt you give, you'll get wildly different answers, right?
And that itself biases the AI in different ways. For example, patients will type in, "Hey, I think my dentist has messed up my diagnosis. I think I have this, this and this." And the tendency of the AI, because of the way you've-
DR DAZ KASPEREK:
They're priming it-
JAZ:
...presented the problem, is to look for reasons to support you. That's a big issue. So what advice do you have for dentists starting to use ChatGPT or any other model? And I'd love to hear which models you use daily to help you make better judgements.
DR DAZ KASPEREK:
One of the key issues I've found, when people come to me asking about AI or for troubleshooting, is that they're not customising their model.
When you sign up to ChatGPT, Google Gemini, Claude, et cetera, there's an option in the settings called personalisation, where you can put in information about yourself: what you do for a living, what you're interested in, where you live.
In my personalisation, I've told it my name, that I'm doing my endodontic training at this university under this professor, that I'm interested in AI and technology, and what my preferences are. So I tell it a lot about myself and my preferences.
So any time I'm having a conversation with AI, it already knows all this information about me. Say you're asking a dental question. The AI doesn't know: are you a patient? Are you a child? Are you a dentist? Are you a specialist? You could be anyone. So you get a very generic answer back.
That's the key thing people struggle with. They try it and say, "Oh, I tried this, it's rubbish, it doesn't really work," or "It gave me garbage." So the first key thing is to always log in and use your own account.
Don't share accounts with other people, because at the moment that doesn't really work. Have one account per person, and personalise it with your own instructions. Once you've done that, you'll see a dramatic improvement.
CHAPTER 6: THE RCT PROMPT: ROLE, CONTEXT, TASK
DR DAZ KASPEREK:
The second thing, after you've done that, is prompting. The basic thing I'd say is that I have a little acronym: RCT.
Root canal treatment, right? Obviously I love endo, so it had to be endo-related. R is the role. What do you want the role of the AI to be? Is it a fellow clinician, a fellow dentist, a specialist? Is it an admin assistant? An email writer?
C is the context. Context is key, and it's the biggest thing people miss out on, because AI gets better and better with worse and worse prompts as the models improve. But context is everything. To me, context is like the lab docket you send to your laboratory when you're making an indirect restoration or a crown, right?
If you just say to them, "Make me a crown," they don't know which tooth it's for, what material, what shade, what characteristics, what the occlusion is. You're going to get a generic crown back, and when you get it, you'll say, "Well, what is this? This is not what I wanted," because-
JAZ:
Wonderful comparison. I love that.
DR DAZ KASPEREK:
Thank you. When you interact with a chatbot, you always have a model answer in mind that you want back. When you have that expectation and get something else back, you think, "Well, this is rubbish." But context is key.
So always give AI a lot more information than you think it needs. The last thing is the task. What exactly does it need to do? Does it need to brainstorm a diagnosis? Write an email? Help me brainstorm ideas because I'm travelling to a conference somewhere?
So that's RCT. That's the main thing, with C being the most important, because context is everything. Even with the models getting smarter and smarter, nothing can replace context. In the future, I think AI will just be integrated into our lives, whether through wearable devices, glasses or anything like that.
So the context will be fed in automatically. It'll know about your day: what you've done, who you've met, who you've spoken to, what's on your phone. So when you say, "Write an email to Jamie," it'll already know who Jamie is and where they work.
At the moment, though, to get the most out of AI, you have to spend a little bit of time customising it and tailoring it to you. But once you've done that, the quality of the outputs you get will skyrocket.
JAZ:
You made a wonderful summary: role, context, task. I love that so much. That's a real gem already. Even the role part of prompt engineering, like "Act like a specialist," or "Act like Jackie Chan and tell me this," whatever. I love that as well. And one thing I use a lot is speaking to ChatGPT.
I love having a proper chat as I'm driving. It's like a phone call. I love that feature. And with AskJaz, I sometimes talk to myself: "In this episode, what was the gem shared by this dentist?" I love hearing the voice back, and you can do it in different languages, which is just crazy, right?
The AI can speak Chinese, and I can't. I love that as well.
CHAPTER 7: AI AGENTS, ADMIN AND AUTOMATING YOUR CPD RECORDS
JAZ:
But when it comes to tasks and administrative tasks, whether we want to hear this or not, I think we're getting close to a point in dentistry where we'll need fewer receptionists and fewer admin people, because we now have AI agents that can do so much.
You'll be able to expand on this, but AI agents can work 24/7. The AI agent will literally read the email from the patient, help them with an appointment, book it for them, send them the welcome email, check in with them a few days before, and send them the treatment plan.
All the things a human does at the moment. I think we're two years away from AI agents just doing this, and it's going to massively shape the workforce of the future, starting with dental admin staff. What do you think? Do you think I'm wrong? It's a scary thought for sure, but what do you think of my prediction?
DR DAZ KASPEREK:
No, I agree with you. The agentic workflows, and the way the latest ChatGPT 5.2 models have been trained, are actually amazing at agentic tasks.
In ChatGPT, you've got agent mode, and you can also use the 5.2 thinking mode, which is the reasoning model. The recent reasoning model released by ChatGPT is really, really good at agentic tasks.
I think that's where the industry is going, because all these repetitive, boring administrative tasks take a lot of time. They're very time-consuming, and they can easily be replaced by agents. I'll give you an example of something I did recently with GPT-5.2 Thinking and agent mode.
I had a lot of CPD certificates, and there's a GDC template you have to fill in with all your certificates, where you went and how many hours. So I uploaded all the certificates and my Excel spreadsheet, and I said to it, "Read through all the certificates and copy all the information from them, the hours, the objectives, everything,
"and put it all together in the spreadsheet." And it basically just did it, and I could go and do something else. Previously, you'd either have to hire someone to do this or do it yourself and waste a lot of time. So with agentic workflows becoming more and more prevalent, I think you're absolutely right. And you're going to get more and more personalised advice, right?
For every patient. That's why AI is brilliant: you can make personalised consent forms, personalised recommendations or explanations. For example, say I have a patient with external cervical root resorption, and I need to explain that to them.
If they're a mechanic or an engineer, I can say, "Okay, imagine you've got a car door with a bit of rust coming from the bottom, and you have to file it down and fill it before it gets worse." It can come up with these amazing strategies
that I don't think I'd ever be able to think of, and it can make personalised medicine very, very accessible. I think it'll make the experience much better for patients and for dentists, because it'll be less effort for the dentist and much higher-quality output for the patient, because the patient will have something they can really, really understand. I always want to get better at communication,
but I'm never going to be the best communicator in the world. AI can close that gap. It can make you a hundred times better than you already are. That's how I see it: like an extension of yourself. It amplifies your skills and fills in the things you may not have the ability to do at the moment.
JAZ:
Daz, I'm going to add to that. This month is AI month on the podcast, and one of the guests we have this month, from the States, uses AI in exactly this way.
I don't know whether he's putting the audio or the transcript in, but he puts in the transcript of the entire conversation with the patient, anonymised. Then he gives it the outcome, like whether the patient went ahead with treatment or not, or what the patient said, and it kind of audits it.
"Okay, your sentiment score: I'll give you seven out of 10. Next time, when the patient said this, you could have said this, because the patient was trying to tell you this." It's amazing how you can use that data to improve your communication
when you think of it that way. But I'm going to pause for a moment, because we can't gloss over the wonderful tip you just gave us. Certificates are an annoying thing, right? CPD certificates. We're getting fewer and fewer paper ones, which is good, because the more PDFs we have, the easier it is. The way to do it is to put them all in a Google Drive folder.
You can give ChatGPT, or Claude or whatever, access to your Google Drive, because Claude's agent stuff is pretty up-and-coming and good as well. Then it can pull out all that information, and you can say, "Okay, make me a PDP summarising everything I've learned so far this year, and leave a space for me to decide what I want to do next year."
Take manual input: one of my friends, Alex, was telling me he logs into a website and types in the title of every single lecture he went to. That's BS. An AI agent can just fill that page in for you, right? If you still want to use whatever website he's using, it can fill it in for you.
We've come to a stage where the only limit is our imagination. So now, guys, think of all those Protrusive certificates, all the certificates you're getting. In fact, Daz, it would be so easy for me to make a custom GPT on ChatGPT where you just give it your Google Drive folder
and it makes it for you. Behind the scenes, I'd say, "When the dentist uploads their Google Drive folder, do this and give them this output," and that's their PDP. And it could be free, because it takes me minutes to make, and it just saves everyone time. That's the reality we're in now. The only limit is your imagination.
CHAPTER 8: PROJECTS, CUSTOM GPTS AND WHY YOU SHOULD TALK, NOT TYPE
DR DAZ KASPEREK:
Absolutely. A key thing is how you set up this personalisation, with custom GPTs, but also projects. Projects are a big thing. ChatGPT Projects, Gemini Gems, Claude Artifacts: they're all basically the same thing, little subsets.
Imagine you're building a little folder for a very specific task. Any time I'm doing a big task, I don't have to give the same context and background every time I start a new conversation in that space. For example, I've got one for emails.
Every time I dictate my emails, it knows exactly my email style. It sounds like me, with British English, and it finishes emails the way I do. I want my emails to be professional and polite, and it matches my style.
If I did my email dictation in a normal chat, I'd have to give it the same instructions again and again for every reply. With a project, you save time, because on top of the personalisation instructions you have in a normal chat, you've got project-specific instructions.
That's really, really good for repeatable tasks, for example, a consent form, or my endo treatment notes. They're always going to follow the same format, so I can quickly dictate all my information, put in my working lengths and anything else that's relevant.
Then, from the same project, once I've completed a full treatment, I can generate a full summary for the referring dentist, and I can make a social media post if I want to. It's all in the same project. Because it has all the context, I have to put in the minimum effort to get really high-quality output.
JAZ:
And it's way better than what you'd have done manually, and a hundred times faster.
DR DAZ KASPEREK:
Absolutely. The only caveat is that you have to spend a tiny bit of time setting it up at the beginning. But once it's set up, ChatGPT is really, really good at following the instructions.
You're just recycling the same thing, and it already knows what you're talking about. So I have to give it the minimum amount of information and prompting to get a very high-quality output, because it knows all these instructions in the background.
And one of the key things about how people interact with AI is that the majority type. I think that's a massive mistake. The key is voice transcription. At the moment, ChatGPT and Gemini are the two main ones with voice transcription enabled.
In my experience, ChatGPT's transcription is way better. You can talk for longer, the quality is higher, and you can probably talk four or five times faster than you can type, so you can put in so much more information, and it doesn't take as much effort.
You can just be mumbling around, and that's where you give the AI the context. First of all, it saves you so much time. People say to me, "By the time I've typed it into the chat, I could have written it myself." I say, well, that's the key mistake. If you voice-type it, it's so much faster, and you get much higher-quality output, because you can put so much more information into the chat to get the best response from the model.
JAZ:
A hundred per cent agree. I also want to add an ergonomic reason. Why are we looking down at our phones? We should just be speaking, right? There's no reason to be typing. I do so many voice notes anyway.
CHAPTER 9: AI NOTES, AND THE "WHAT IF I FORGET TO SAY IT?" OBJECTION
JAZ:
Daz, I don't know if you're still working in practice part-time while doing your DDSc-
DR DAZ KASPEREK:
I work in practice as well, yeah.
JAZ:
Are you using AI note software? I know you're using ChatGPT in a way, but are you using dedicated dental AI note software?
DR DAZ KASPEREK:
I use a customised version of ChatGPT. I've made my own AI voice notes software. So yes, I do my notes with AI, but I know there's Heidi.
JAZ:
There's Heidi, there's Digital TCO, there's Dentistry Dashboard. Even Claude, you know, the Claude AI thing? It's got a dental note-taking mode on there, right?
DR DAZ KASPEREK:
Oh, right.
JAZ:
Hopefully this month, if I've got my ducks in a row, I'll be posting the video where I compare them. My biggest review yet. I'm going to feed the same consultation into all these programs and see who makes the best notes, the fastest, and the best letters. But whatever software you're using, the amount of time it saves us is huge. One of our orthodontist colleagues, Stuart Yeaton, on our Protrusive Guidance community, loves it for writing back to referring dentists with templates. On Digital TCO, it's called Molar Mail.
Just yesterday, I was having a telephone consultation with a patient. I was following up with her, saving her a visit. I get patients coming from all over for TMD. She wanted a telephone consultation, so I put her on loudspeaker. All my patients know everything is being recorded.
They all consent to it. No one's ever said no. So I had her on loudspeaker, and the notes just wrote themselves. You just have to read and check them. It's little time-savers like that that mean we get so much more done. And for those dentists who are stressed by notes, not only are their notes lower in quality, they're also having to spend so much time on them.
One objection I received from a young dentist about AI notes was, "What if you forget to warn them about something?" Compared with templates: your template for an extraction, for example, has got everything in it. Whether or not you warned them about a sinus breach, it's in there, right?
So this dentist said to me, "Well, what if you forget to say it?" And I said, how is it fair that it's allowed to be in the template when you didn't say it, but not the other way round? It's a reminder for us to use a checklist, right? If we follow The Checklist Manifesto and always warn, it makes us better clinicians and better communicators, and patients are better off. Now, Digital TCO specifically has something called Super Dentist Speed mode, right?
It's literally what hospital consultants do. They say, "I saw Mrs Smith today. She had a mouth opening of 39 millimetres," and someone in the back end types up the letter. Within a few years, that's going to be gone. All those people, unfortunately, are going to lose their jobs.
It's a sad reality. That's going to be replaced by AI, if it hasn't been already. But for me, being able to be like Busta Rhymes for a few minutes and say things so fast that in 30 seconds I've written a whole letter, and it has captured everything beautifully...
It's such a massive time-saver.
CHAPTER 10: LIVE DEMO: THE DENTAL DISRUPT SMILE SIMULATOR
JAZ:
Now I want to show you something cool, and for those listening, I'll describe it, because you mentioned custom projects. There's a dentist called Jason Lipscomb in the US. He's a podcaster in the US and an awesome dentist. He's got free groups online, and he gives away so much. So a massive shout-out to him.
Let me show you this cool thing he made, and for those listening, I'll describe what I'm doing. I'm going to open up my ChatGPT. My team will blur out my stuff, so you don't get to see what I've been saying to ChatGPT. While it's loading...
Okay, I'm going to go to Dental Disrupt Smile Simulator. This is the free thing he made. Basically, Jason was saying, "Listen, stop paying all these companies to do your before-and-after visualisations."
There are so many programs out there saying, "Hey, upload a video, and we'll do it." They're basically using AI anyway. So he's saying, "Listen, this is free, and it took me a minute to make." You upload a smile, and it gives you the result.
So let's do it. Guys, I've got my ChatGPT loaded. I'm going to load a photo of my sister before she had ortho. I'm just going to find it. Here's my sister before she had ortho. I'm going to add the smile photo, and let's see how long it takes.
The clock is ticking. It's creating the image. And by the way, ChatGPT's image creation is just phenomenal, right? For someone like you who's been YouTubing, with the thumbnail thing, it's been a revelation for us to be able to save time and effort, and it's so much fun to work with Gemini and ChatGPT image creation.
So I've put my sister's image in, and it's still thinking and creating. It's creating it based on what Jason has trained this project on: when someone uploads an image, follow these smile design parameters to create a realistic image we can show a patient. Now it's starting to load. Have you used these things before, Daz?
DR DAZ KASPEREK:
I have, yeah. With a custom GPT, you upload your own custom data sets of things you want it to reference as its knowledge, and then you've got your own master instructions in the background. I've made my own endo one as well.
JAZ:
What does your endo one do? Does it assess complexity?
DR DAZ KASPEREK:
The endo one is basically a reference guide. If I want to ask it about a specific type of very complex case, it references all the sources I've uploaded. So it's more like a second opinion from another clinician who's a specialist.
JAZ:
Very cool. So here we are, the before and after image. How realistic does that look? It looks amazing, right? The funny thing is, those listening won't get to see this, but I'm now going to load up her actual before and after.
Then we can compare whether it was actually close or not. Let's put this here. Not bad, right? Not bad. Obviously the shading and so on is different, but what a wonderful, almost identical result we managed to achieve. The AI is very clever, very good.
Obviously, we have to use our judgement. We don't want to over-promise, but we knew that already from using these simulation programs. By next year, the video stuff will be free as well, right? It's so easy. There's probably a mode where I can upload a video of someone speaking, and it'll send a video back to me.
Grok can probably do it, to be honest with you. Grok AI. So which AI software should we be using, personally and then professionally? Give us some personal use cases you've found good, because even though we're dentists, we're humans too.
And we can actually benefit from these AI use cases.
CHAPTER 11: WHICH AI SHOULD YOU USE? CHATGPT, GEMINI, CLAUDE AND GROK
DR DAZ KASPEREK:
What I'd say, if anyone is starting with AI or wants to get more out of it, is to try the different models first. There's ChatGPT, which is the most famous and most popular one.
You've got Gemini from Google, Claude from Anthropic, and Grok from xAI. There are also others, like Meta's, and Chinese models like Qwen and DeepSeek. But the main three or four are OpenAI's ChatGPT, Gemini, Claude, and xAI's Grok.
Try different ones and see which personalities and types of response you like. Again, set your custom instructions and introduce yourself to the AI, so it knows exactly who it's dealing with and you get high-quality responses. Then see which personality you like.
Then just stick to that one, because especially now with ChatGPT and Gemini, there's a lot more emphasis on memory. As you use these models, they learn more and more about you. ChatGPT now has memories, not just the ones you tell it to remember. For example, in your dental practice there might be a nurse, a practice manager and an associate, each with a certain name.
You can introduce all these people to ChatGPT or Gemini, and they'll remember them. So every time you say, "Write an email to Jason, my associate," it knows exactly who that is. My AI knows the names of all my supervisors and professors.
So if I just say, "Write an email to Prof," it knows exactly which professor I'm talking about, and I don't have to give it that extra context. Again, it just makes things faster and saves time.
JAZ:
And it'll say it in your voice, because you've trained it: "This is how I write my emails. This is me." Then you get to check it. Look, it is a shortcut, but I don't think it's disingenuous, because you've literally trained it to sound like you, you approve it, and then you copy and paste it over. These are basically your words.
DR DAZ KASPEREK:
Yeah, exactly. Once you start learning about it, just pick one and stick to it, because it'll learn more and more about you. In terms of which AI is better at what, I'd say Gemini is definitely better if you're really in the Google ecosystem.
You've got integration with your Gmail and Drive, and it has the biggest context window of all the models at the moment: 1 million tokens. So Gemini is good if you're working with really big documents, and it's the only AI from the Western side at the moment that accepts video uploads as well.
So, for example, you can upload a practice marketing video you're making and say, "Have a look at this video we're making for our practice marketing. Score it and tell us what we can improve. Is it good? Is it bad?"
Gemini is really, really good at that. For image generation, Gemini has the Nano Banana model, Nano Banana Pro. In my opinion, it's faster than ChatGPT at the moment and better quality. They're very close, but Gemini is just a little bit faster.
So for big context documents, photo generation and video, you can use Google's Veo, their flagship video generation model. I'll tell you a really interesting use case from the other day. A friend of mine who's a specialist is running an endo course for FDs or GDPs, I think.
He had a photo of an access cavity with a normal handpiece and bur over it, compared with an ultrasonic. He wanted to show that when you use the normal handpiece, you don't see much because the handpiece is in the way, whereas with an ultrasonic, you can see a lot more.
He had the two photos side by side and asked me, "Can you make a video out of it?" So I went to Gemini, used the Veo model, and got it to move the handpiece and move the ultrasonic. Your photos come to life, and it's so much nicer, because it makes it easy to communicate your idea. Gemini is also really good with maps, because it has all the Google Maps stuff.
I was buying a car and had to visit different dealerships. When you have to go to different places, you go on Google Maps: "Okay, if I go from this place to this place, it takes this long."
Then you do all these calculations: "Should I go to this one first, or that one? What's the most efficient way?" So I told Gemini all this, and it did all the calculations for which dealerships to go to first.
It said, "Okay, if you live in this area, at this time of day, if you leave at this time, this is the best route." So for Google Maps stuff and video, I'd say Gemini is better. But I use ChatGPT most of the time. And I'd say: don't just use one model.
Use one as your main model, but you can also ask other models for their opinion. If I'm doing something very big or very serious, or I need a lot of opinions, I'll literally open ChatGPT, Gemini, Claude and Grok and paste the exact same thing into all four.
It's like a council. I get the opinions of all the different AIs, and they have slightly different training and slightly different instructions, so with the exact same prompt they give me different opinions, and I think, "Oh, actually, that's a good idea."
That's really, really useful for a big task where you're using their reasoning models, the thinking models. So I'd definitely recommend that. I prefer ChatGPT, but some people might prefer others. My parents prefer Gemini, for example, and some of my friends prefer Gemini just because of its personality.
So it doesn't mean you have to use ChatGPT because everyone else does. Just try them and then stick to one.
JAZ:
It's a real rabbit hole you can go down. It's just a wonderful thing. There are so many dangers to it, and we'll talk about patient confidentiality in a moment. But for me, I'm rubbish at DIY. My dad was never into DIY, and I never got into it. So any time I have to do DIY or IKEA stuff, if I'm confused, which happens very soon into any DIY task, which is why I don't do implants...
I get out the voice mode on ChatGPT and switch on my video: "Wait, which screw does it need?" And it says, "Oh, the one over there labelled B." "Okay, cool, I've got you." Little things that help you navigate life. I don't need to pay someone or record someone.
I can just get ChatGPT to give me that opinion, and it's mind-blowingly amazing. For personal uses, your imagination is the limit. It's a wonderful thing.
CHAPTER 12: PATIENT DATA, CONFIDENTIALITY AND CONSENT
JAZ:
But for dentistry, where are we now? In the States, things have to be HIPAA-compliant.
In the UK and Europe, there's GDPR and that kind of stuff. Are there any rules of engagement on sharing patient details? Because there might be someone in the world right now saying, "Write me a letter to a patient. Here are the patient details," giving it the patient's date of birth, name, address, et cetera. Surely that's dangerous.
DR DAZ KASPEREK:
Yes, exactly. I'd certainly say that, at this point, you want any interaction you have with a chatbot to be completely anonymous. And when is it really needed? How much higher quality will the output be just from putting the patient's details in? I don't think it's going to make much of a difference.
JAZ:
But one thing you said earlier was about context, and the whole communication side. For example, if you want to enhance your communication and the patient is a mechanic, you could say, "Use my template for my endodontic consent form, and adapt it for a 59-year-old male who's a smoker and a mechanic."
That's not patient-identifiable, but it's good context to help you make a better letter. So I think that's okay. As long as it's not patient-identifiable, it should be fine.
DR DAZ KASPEREK:
Yeah, absolutely. Patients' dates of birth, names, addresses, things like that: at the moment, that's a big no. You can do that bit yourself, and it doesn't really take much time. I think the quality of the AI comes from the background information rather than patient-specific information. So any time I interact with any AI model, I never, ever put any patient information in.
It's quite interesting that we're talking about this, because some time ago there was an issue with the share button you get in your chats with ChatGPT. So-
JAZ:
The Grok one? You're coming on to the Grok issue, right?
DR DAZ KASPEREK:
I think it was ChatGPT. I'm not sure, but from what I remember, it created a public address for your chats, and those chats started being indexed in Google, so people were able to find others' private chats in Google. Obviously they've fixed it now, but you never know where an issue like that might happen again.
JAZ:
Or a big data leak could happen with DeepSeek, whatever it's called, in China, for example. Or Grok AI could be affected, and suddenly someone has all this data. So I think it's a very easy debate: don't overshare. Give what you need for the context, but don't give any patient-identifiable information.
Now here's a curveball, Daz. You just saw me, and everyone listening heard me, upload my sister's smile to ChatGPT. I didn't ask my sister. She's cool, but I didn't ask her. Jason Lipscomb made this wonderful thing. So at what point do we need to say to our patients, "To help you visualise this,
"I need to share your image and data with ChatGPT"? Where do we draw the line with a patient's image? Is it okay if it doesn't show the eyes? Do you know anything about this? Because I don't.
DR DAZ KASPEREK:
To be honest with you, I'm not sure. I think it's always better to err on the side of caution. As long as we follow the principles of patient confidentiality, I think that's the key. And in terms of just an-
JAZ:
The ideal way, Daz, if you were going to upload it to a model for an AI simulation, is to ask the patient: "Look, I can give you a really cool simulation. It's mind-blowing, and I'll tell you whether it's realistic or not. I just need your permission to share this image with AI, with ChatGPT or whatever, just with this model." As long as the patient gives you consent, that's fine. Where we can get into trouble is if we don't get that level of consent.
DR DAZ KASPEREK:
Absolutely. Everything in dentistry is to do with consent. You have to get consent for treatment, and I'm sure patients understand that we're doing this to help them.
We're not doing it for our own gain. We're only doing it to give them a better understanding. To us, it's evident and very easy to visualise what the outcome could be, but for them it may be a little trickier. And it's just another tool, right?
It's like an x-ray. It's another way of conveying the data in a very patient-specific way. And again, you're the judge at the end. You're the one who says, "Okay, this is actually quite realistic,"
or, "Oh, actually, just ignore it, this isn't going to happen." At the end of the day, you're ultimately always responsible for the information given to patients. You can't blame the AI: "Oh, it told me to do this and I did it," or "It wrote this consent form and I never read it,"
or "It wrote my notes and I never read them." You're always responsible for everything that goes out with your name on it. I think that's the key thing. And you're absolutely right: it's all about patient consent at the end of the day,
and making sure we and our patients have the same understanding of the best way to communicate with them.
CHAPTER 13: WHAT WILL AI REPLACE? ADMIN, EDUCATION AND NOTEBOOKLM
JAZ:
Let's get our crystal balls out. What will AI replace in dentistry in the foreseeable future? I think admin teams will reduce massively. I'm not saying they'll be fully replaced. You still need someone to oversee things, because AI can make mistakes, and someone needs to quality-control it. In my business, in education and Protrusive, we've used AI massively. I've got a team of eight or nine now, and I haven't let anyone go.
I've just helped them use AI so we can achieve more and serve more dentists, with AskJaz and so on. We've all been part of that. Rather than getting rid of someone, which I wouldn't want to do... but clinics want to be profitable. I'm not saying "getting rid of", it's such a harsh term, but they won't need so many people.
If they have a team of six receptionists, they might only need two now, which is good and bad, because it's a very sad workplace reality. So my prediction is that admin tasks and admin teams will reduce. With AI notes, we're pretty much already there, and I think the vast majority will be using AI for writing letters, notes, communication, that kind of stuff. What else do you think it will replace? Then we'll come on to what AI won't replace in dentistry.
DR DAZ KASPEREK:
Again, I think education is quite big, and tools. You mentioned education, and as dentists, especially the Protruserati, we want to get better, we want to improve, we want to learn. We go to courses, we want to revise and study. I think personalised education is a big thing as well. Like you said, with your custom AI, you've got your own custom knowledge, and it's almost like giving a copy of yourself to mentor that dentist.
JAZ:
It's a digital mind. They call it your digital mind.
DR DAZ KASPEREK:
Absolutely. And with tools like NotebookLM, you can upload your PDFs and papers, and it can turn them into a podcast, a presentation or an infographic. It makes it so much easier to revise, consume knowledge and learn. So I think this is going to have a big effect on teaching and training,
including staff training, because different people learn in different ways. One person really likes to read, someone else likes to listen to a podcast or watch a video. With AI able to deliver all of these, plus personalised quizzes, tools like NotebookLM are going to be incredibly useful.
JAZ:
Think of us: imagine we were students again. Students today have no idea how much information they have access to, all these concepts. So many times I've had a chat with ChatGPT and said, "Look, explain to me like I'm 12 what this paper is trying to say."
Because some bruxism papers are really complex, with all the neurotransmitters and so on, and it's helped me out so many times: "Think of bruxism like a computer going into standby mode." "Okay, that makes sense now." So from an education perspective, even self-education, it helps so much.
NotebookLM is a very noteworthy mention for the students of today. We used to do flashcards, past questions and quizzes. You can now literally tell Gemini, "Okay, here are a few past papers from a few years ago.
"Create me a bank of 500 questions in a separate document, and create the ideal answers I can refer to after I've had a go." Wow.
DR DAZ KASPEREK:
Yeah, it's incredible. That's actually what I was talking about. I just went to a conference in Paris, the ESE, the European Society of Endodontology, and we were talking about NotebookLM. We actually use it in Liverpool for our journal club.
Every week we have to read papers, so we've introduced the podcast as part of the journal club. You just listen to the podcast, and you're already primed, so by the time you actually read the paper, you already have so much more information, and you're not falling asleep trying to read it. So I think-
JAZ:
Plus, in our busy lives, as a parent, I sometimes don't get time to read. I do audiobooks. So if you're part of a journal club and struggling to read because of kids or life or whatever, but you have a half-hour commute, wow.
NotebookLM will make a podcast summary of that paper. You can listen to it and be clued up. And a lot of people prefer to learn that way anyway.
DR DAZ KASPEREK:
Absolutely. You mentioned that some jobs may go, but I think it's just a transfer of skills. There were the same worries when the internet came out, that it was going to take everyone's jobs, but look how many jobs were created because of the internet.
I think exactly the same thing will happen with AI. The expertise will shift, and new roles will be created. Rather than doing boring admin and reception work, you'll be the prompt master, supervising the AIs, making sure the output is correct, and troubleshooting when the different AIs and tools talk to each other.
AI is amazing at coding, especially, and it can teach you how to code. It's the best teacher. So over time, I don't think jobs will necessarily be lost. There'll just be a transfer of skills, and new roles will become available because of the high demand for this kind of expertise.
And patients always appreciate personalised care and attention. You can have an AI chatbot personalised to the patient. If the patient talks to it, the AI can flag if something's wrong: "This patient was talking to me, and the socket isn't healing after the extraction." It could ping your practice and say, "Check on this patient, because they've been interacting with our chatbot and are reporting some complications."
If anything, I think it's going to improve patient care further and give a better overall experience to everyone involved.
CHAPTER 14: AR GLASSES, IMPLANT ROBOTS, AND WHAT AI WON'T REPLACE
JAZ:
Talking of experience, you also mentioned augmented reality earlier. One cool thing I'm doing at the moment is having patients wear XREAL glasses. The XREAL Pro 2 is what I've got. They wear the XREAL glasses, and I screen-mirror my loupe camera to the glasses while I do my check-up. So I'm doing my check-up live, and the patient is hearing and seeing. When I'm doing the mouth cancer check, they're sticking their tongue out and seeing their tongue through the glasses.
This has been a revelation over the last four or five months of experimenting with it. I'm really pioneering this, and I love it. Not a single patient has said it was terrible or that they hated it. They've been blown away. They say they never want to have an exam any other way again.
Normally, they don't get to see that level of insight. And the AI element is this: say the patient says, "Oh, you know what, can you send me a video of that?" That's totally cool. Previously, my team would spend hours editing my clinical videos for education. Now there's AI within the software.
Every time you look away, it removes those bits and keeps the relevant parts, and you can just package it up and send it to the patient: "Here's the video of your check-up." A bit like when you get your car serviced and they send you a video of your car: here's what's good.
This is the future of dentistry. It's got AI, it's got the glasses experience, and I think it's only going to evolve. What I want to know from you now is: what are the limits, and what will AI not replace? Where do you think AI can't replace dentistry, because it can't and won't do certain things?
DR DAZ KASPEREK:
I think you're absolutely right about the amount of information we can now transmit to patients effortlessly, to educate them and engage them with us. We can take the dreaded, boring task of going to the dentist and turn it into something some patients will be very interested in, and I think that's a really good way of building extra rapport with them.
I recently went to a conference in Venice on digital dentistry, and there was a company there from China that has made a robot that can do implants. They have hundreds of these robots deployed across China, and the robot can literally just place implants. I think they said they've already placed 30,000 or 40,000 implants with the robot, and they have robot implant cleaners, which is crazy.
JAZ:
Wow.
DR DAZ KASPEREK:
And they said the next thing coming is that the robot will do the crown preps and everything. It's incredible how much technology is advancing in such a short time. But I think what AI definitely won't replace is the human connection we have and the emotions behind it, because AI is just an algorithm of probabilities at the back of it.
With a borderline restorable case, it might say, "Well, from my training data, this shouldn't really be restorable." But with your experience as a clinician, you may know a way to predictably restore it. And it's about having empathy. AI can generate the most empathetic-sounding letter, but there's always that additional human touch that patients, and staff, will always appreciate.
It's the ability to have that additional layer of interaction with people that AI will never replace. And there's the artistry in dentistry. There's so much nuance in how creative we can be. AI can give you suggestions and recommendations, but we all know our skills.
We know what we're good at and what we're not so good at. We can use AI to amplify the things we're not very good at and give us additional edges to fill the gaps. But we'll always be the experts in patient communication and in the artistry behind it.
So I don't think the creativity and artistry can be replaced. But with Neuralink coming out, I think over time humans and AI will start to merge, and you'll have pure humans, and half-humans, half-AI.
CHAPTER 15: DAZ'S RESEARCH, HOW TO FIND HIM, AND OUTRO
JAZ:
It's like all the sci-fi stuff coming to life, really: cyborgs and that kind of thing. How fascinating. Daz, tell us how we can reach out to you and learn more from you, and also, what are you actually studying with AI and endo? What's your thesis on?
DR DAZ KASPEREK:
One of the things I'm looking at is off-the-shelf models, ChatGPT and Gemini. I'm learning how they work, and I'm looking at endodontic complexity assessments, seeing how they perform in assessing endodontic complexity. We've got a lot of custom-trained models for analysing caries, CBCTs and all of that.
But I'm trying to see whether these models, with all their vast training data, can be tailored to be used by everyone. Because you'd need a separate subscription for everything else. And yes, a custom model is arguably better, but what I'm looking at is how we can use the technology that's in your pocket and my pocket to help us,
and how we can customise it to get the best outputs, without having to have custom AIs for everything. That's one of the main things I'm looking at.
JAZ:
I imagine it's scarily promising, how good it'll be at assessment based on images and radiographs. I'm sure it's scarily good at assessing complexity.
DR DAZ KASPEREK:
There are still some nuances. What I'm still working out is how the models think and perceive things, and then tailoring them to get a better output.
But that's my AI work. Every few weeks, I do an AI update for everyone I'm doing endo with, all my professors and specialists, and I'm doing some AI teaching sessions at the University of Liverpool. It's super exciting, and it's really changing the world.
JAZ:
What a time to be alive.
DR DAZ KASPEREK:
Absolutely. It's incredible how fast things are progressing and how much they're improving. Like you said, it's the future, and it's only going to get better. Honestly-
JAZ:
How can we send you some love and thanks, and follow what you get up to, man?
DR DAZ KASPEREK:
On Instagram, I'm endodaz. If you want to reach out to me, I'm more than happy to help. And if people are interested, I'll be more than happy to organise some things for people who want to learn more about AI.
That's my main hub. If anyone's interested, give me a shout at endodaz. Thank you.
JAZ:
Amazing. And thanks so much for being part of the community, and for all the Protruserati you've served in the past with the cross-polarising filters. You're a man of many talents, I have to say: endo as well, and AI.
Mate, thank you so much for coming on. Maybe in a couple of years we'll do an update and find that everything we said today was laughably irrelevant, because it all moved so fast.
DR DAZ KASPEREK:
Yeah, it's changing so fast, I wouldn't even be surprised. But thank you so much, Jaz, for having me on. It's a massive honour to be here, and I love what you're doing, and your team and everyone are amazing. I'm excited to see what happens in the future. Thank you again.
JAZ:
All the best with your studies, my friend.
DR DAZ KASPEREK:
Thank you. All the best, mate. Take care.
JAZ:
Well, there we have it, guys. Thank you so much for listening all the way to the end. You got through this episode with Daz, and wasn't he great, sharing all his enthusiasm, ideas and geekiness? I hope you're feeling a bit more clued up about AI, and hopefully a little bit excited.
At the time of recording, it's AI month on the app. And if you're not yet on Protrusive Guidance, it's not like Facebook. It's not toxic. We're all helping, sharing, chatting and posting, and of course there are hundreds of hours of CPD on there as well.
We're a PACE-approved education provider, and this episode has a quiz. Scroll down and answer the quiz to get your CE for this episode. Your CPD Queen, Mari, will email you a certificate, and every quarter she'll send you your folder with all your certificates inside. Every month I do a live webinar and a new consent form for you guys, and every so often we release a new mini-course.
We also now have our own Protrusive AI, AskJaz, which is an expert in clinical reasoning. To access it, you have to join the Infinity Plan, and the best place to find out more is protrusive.app. That's the website: protrusive.app. Once you've made a login, you'll be able to download the iOS and Android app, and hopefully you can make an account in time to get involved in these discussions.
Every day we're discussing new things about AI, or whatever the theme is that month. Last month was all about personal finances, and next month is all about preventing stress and burnout. So check out protrusive.app. And don't forget to try the Dental Disrupt Smile Simulator. The link is below.
If you like what we're doing, give us a like and subscribe wherever you're listening. Thank you once again. I'll catch you same time, same place next week. Bye for now.
FREQUENTLY ASKED QUESTIONS
QUESTION:
What is a prompt, and how do I write a good one?
ANSWER:
A prompt is what you give the AI to help it work on your problem. Dr Daz Kasperek uses the acronym RCT: role, meaning who you want the AI to act as; context, meaning all the background it needs; and task, meaning exactly what you want it to do. He says context matters most, like a detailed lab docket.
QUESTION:
Why should I personalise my AI account?
ANSWER:
Daz explains that, without personalisation, the AI doesn't know whether you're a patient, a child, a dentist or a specialist, so you get generic answers. Adding details about your job, interests and preferences in the personalisation settings, and using your own account rather than a shared one, dramatically improves the quality of the responses you get.
QUESTION:
Can AI help with clinical diagnosis?
ANSWER:
Daz sees AI as a useful second opinion: you can describe anonymised signs, symptoms and clinical findings to help build a differential diagnosis. He stresses that AI doesn't have a GDC number, so you remain responsible for your clinical judgement. Jaz warns that general chatbots can hallucinate and invent references that don't exist.
QUESTION:
How can AI agents save time on admin like CPD records?
ANSWER:
Daz uploaded his CPD certificates and the spreadsheet he needed to fill in, then asked ChatGPT's agent mode to copy across the hours, objectives and other details, and it completed the task for him. Jaz suggests keeping certificates in a Google Drive folder, so AI can read them and help draft your Personal Development Plan.
QUESTION:
What are projects and custom GPTs?
ANSWER:
Daz describes projects and custom GPTs as folders for specific, repeatable tasks, with their own instructions and reference material, so you don't have to repeat the same context in every conversation. He uses them for emails written in his style, endo treatment notes, referral letters and a reference guide for complex endodontic cases.
QUESTION:
Which AI model should dentists use?
ANSWER:
Daz recommends trying the main models, ChatGPT, Gemini, Claude and Grok, then sticking with the one whose responses you prefer, so it learns more about you. He uses ChatGPT most, finds Gemini strong for Google integration, large documents, video and maps, and for big decisions pastes the same prompt into several models.
QUESTION:
Is it safe to put patient details into AI?
ANSWER:
Both Jaz and Daz advise never putting patient-identifiable information, such as names, dates of birth or addresses, into a chatbot. Non-identifiable context, like age, smoking status or occupation, can still help personalise a letter. For patient photos, Jaz suggests asking the patient's permission before uploading their image to an AI tool.
QUESTION:
What won't AI replace in dentistry?
ANSWER:
Daz believes AI won't replace the human connection, empathy and artistry of dentistry. He explains that AI works on probabilities, so it might call a borderline tooth unrestorable when an experienced clinician knows how to restore it predictably. AI can draft an empathetic letter, but he says patients will always value the human touch.
ABOUT THE GUEST
Dr Daz Kasperek graduated from the University of Sheffield in 2018 and is an Academic Clinical Fellow undertaking a Doctorate in Dental Science in endodontics at the University of Liverpool. His research looks at AI and augmented reality in endodontics, including how off-the-shelf AI models handle endodontic complexity assessment. He also runs AI teaching sessions, and is known in the Protrusive community for his cross-polarising filters.
- Instagram: @endodaz