AI 101 for MedTech Sales

Actionable advice for leaders who want their reps to be more efficient, and for reps who want to make every day a productive day.

60min

You’ll leave the webinar with:

  • Definitions of key terms/concepts such as LLM, agent, MCP, and harness
  • An explanation of how generic AI and MedTech AI differ from one another
  • Tips for prompting AI
  • Generic AI use cases (with prompts you can use)
  • MedTech AI use cases (with prompts you can use)
  • 3 action items/things to do right away

Speakers

Lee Smith

Co-founder and Head of Commercial Development, AcuityMD

Lee Smith is Co-founder and Head of Commercial Development at AcuityMD, where he leads the team helping medical device companies rethink how they identify, prioritize, and win new business. Before co-founding AcuityMD, he held marketing leadership roles at Medtronic, where he managed orthopedic implant and total hip product lines, and at Zimmer Biomet, where he led product management for the Knee Creations division.

Carl Pieri

Senior Product Manager, AcuityMD

Carl Pieri is a Senior Product Manager at AcuityMD, a key member of the team building the AI platform for MedTech. Before joining AcuityMD, he held product management roles at HubSpot, where he led strategy for the company’s AI-powered long-form content platform, and Accurx, where he grew the company’s patient triage product from a recent market entrant to the #1 online consultation provider in the United Kingdom.

View webinar transcript:

AI 101 for MedTech Sales
This transcript was automatically generated and lightly edited for clarity. Some errors may remain.

Welcome, everybody. Thank you for being here. Thank you for taking the time out of your Tuesday to join us live for today's webinar. My name is Conor. I'm on the marketing team here at AcuityMD, and I will be your MC host for the next 40, 45 minutes or so. And I am joined today by two individuals, two of my colleagues who have a ton of insight into how you all and your teams can benefit from making better use of AI. We have Lee Smith, co-founder of AcuityMD and our VP of customer experience, and we have Carl Pieri, one of our senior product managers here at AcuityMD. Lee brings a stellar MedTech commercial background. Carl brings a fantastic AI background. And together, these two guys have a ton of expertise on what it is we'll be covering today. So Lee, Carl, appreciate it.

And speaking of which, I do, before handing the mic to Lee and Carl, I do just want to quickly level set on what it is we're covering today and why we're covering this particular topic. I think we're all here because we're all excited about the opportunity that's in front of MedTech sales teams, and that opportunity, of course, is AI, which has the potential really to dramatically improve your team's productivity and performance. One customer of ours that's been really early on the AI adoption curve, they have tripled their reps' research capacity, meaning each rep on their team is now getting 3X more physician interactions per week compared to what they were doing before adopting AI. And not only are they getting 3X the interactions, they're actually better prepared for each of these interactions because AI is doing the heavy lifting of all the research on the physician. So this is the promise of AI, right? This is what I think we're all really excited about. This is what we all, this is where I think we all want to get.

At the same time though, AI can be a bit daunting, right? For most of us, it came completely out of nowhere. It seems to be advancing and changing at a nonstop pace. There's a lot of debate about what it can or can't do, what it's good at, maybe what it's not so good at. And I think when you put all those things together, it can make it really hard to figure out just how to get started and how to start building that initial momentum.

So that's why we're here because, we're simultaneously we're all excited about the promise of AI for MedTech sales, and we all just want to start making that positive forward progress. So let's start making some positive forward progress together today. Carl is going to kick things off for us by defining some key terms and concepts that will set the foundation for the rest of our presentation, and then Lee will define generic AI and MedTech AI and help us understand how those two things are different. Building off of that, Carl will then do some demonstrations. He'll demonstrate a couple of generic AI use cases that you all can take with you after the webinar and try out for yourselves, pass along to your teams to try. Similarly, then Lee will take over. He'll demonstrate a few MedTech AI use cases, and as we'll see, the use cases are really quite different. But same idea. These are real use cases that are applicable to all of you that you can take from Lee and try out yourselves, pass along to your team.

And then we'll wrap up with some universal tips for prompting AI, working with AI, and we'll give you all a few action items to go execute after the webinar to start building that positive momentum.

Oh, and just a couple of logistical notes. You all, at the bottom of your Zoom screen, you all should have access to the Q&A function. Please use that Q&A function if you have any questions for Lee and Carl. We'll save some time to answer as many of those as we can at the end of the webinar.

And last logistical note, this is being recorded, and you all tomorrow will receive an email with the recording of today's session, a copy of this slide deck, and a companion e-book resource that I will share more about at the end. But for now, that is enough housekeeping and logistics. I am going to pass things over to Carl, and I'm actually going to stop sharing my screen in a moment, because Carl is going to share his screen and pull up a couple of AI tools to serve as visual references as he defines these six terms and concepts. So Carl is going to walk us through what an LLM is, what a prompt is, hallucination, MCP, which stands for model context protocol. He'll define an agent for us, as well as an AI harness. And like I said, this is being recorded. You all will get the recording and the deck, so don't worry about taking furious notes as Carl is defining these things.

You all will have plenty of material to refer back to. So with that, I'm going to stop sharing my screen. I'm going to let Carl share his screen and define those terms for us. Go ahead, Carl. Awesome. Thank you so much, Conor. All right. So I'm going to use Claude and AcuityAI to help kind of hand, to hammer these points home. So we're going to go through a few different things. So the first definition I want to cover is LLM or large language model. It's basically a model that can understand human language and generate responses based off of it. The helpful way to think about of it is sort of like a brain that powers a tool like Claude or AcuityAI. So each of these are actually different models that exist within Claude that you can use so that each of these is their own LLM. So that's what an LLM is. Next is a prompt. So a prompt is what you add to it, so anything you post here. So you might say, I'm going to pull an example in, "Help me write an email to a doctor for an upcoming meeting." That would be the kind of example of a prompt that a MedTech sales rep might post.

And the next thing to keep in mind is context. So context is really like the background information that you provide to AI that allow it to give you a really good output. So what I've shown here doesn't have much context. The AI won't know much about the meeting or the doctor or what I'm trying to cover in that email. So I can add with more context something a bit more like this, which asks for an email but also says, "Last week I gave him a demo of my brand-new bone screw. He told me that he wanted to reach out setting up some time with the other doctors at his practice. They want a demo of the product, and they want to hear about other doctors who found value." So all of this second kind of bullet here is the context that I've provided to the LLM to help it out.

Now, one thing to be wary of when using AI is hallucinations, which you've probably heard about. It's basically when AI confidently gives you an answer, which is actually fabricated information. So an example might be, I'm a rep, and I ask Claude to tell me, "Hey, where does this specific physician practice?" And then it might give me the example or give me a name of an incorrect site. So a site that doctor doesn't actually practice, but it might present it as if it's accurate. So that is an example of a hallucination occurring in AI.

Just a couple more terms to get through. The next one is MCP, or Model Context Protocol. It's a fancy word for basically just saying that your AI can connect to other systems and other tools. So I'll show some examples here. I've got a whole bunch of connectors set up. So I can use tools like Google Drives or Google Sheets or Google Docs. So this way, my AI can actually go look at those systems, pull that information, pull it into our conversation, as well as make changes in those systems. So it just gives your AI a bit more power. It can do more things when you have these MCPs set up. Now, I'm going to switch to AcuityAI to show these last two definitions. So the first one is this term agent. So you might be confused about what's agent versus AI. Agent is basically a more advanced version of AI that can take actions towards a goal on its own. So you may be used to when you use ChatGPT, you ask a question, it gives you an answer, and you go back and forth. An agent, you can give it a more complex task. It can break down that work, and it can go away and take the time to complete it and come back to you. So an example of a response that I might give to an agent, or sorry, a prompt I might give to an agent would be this one here. We'll give it a second to think.

And we should see the agent's brain sort of working in a second as it keeps going. So right now the agent thinking. It's using different tools. Tools are basically similar to how a human uses tools. It's what an AI can use to try to get you the right answer. So it's looking at top providers. It's looking at locations. It's completed its search. It's looking at top providers some more within my territory. So it's really taking step-by-step actions, which is the key distinction of an agent versus just a pure chat AI. Now, we'll let that keep working. As it does, I'll define the final definition, which is harness. So harness is basically the set of tools and skills that makes it possible for an agent to complete tasks correctly, efficiently, and safely.

So in the example of AcuityAI has a bunch of different tools. We can see in this one request it used 12 of them, which have a bunch of context built into it and allow the AI to work more productively. So a harness is basically an advancement on AI that allows it to do more work, access more information, have more context, whereas a chat-based AI that lacks harness, all it can do is kind of answer questions one-off.

So if you're going to kind of bring it full circle, LLM is sort of like the brain, and then the harness is the thing that surrounds that brain, and allows it to actually take actions towards a goal. So there's some basic definitions. I'll stop sharing and hand it back over. Thank you, Carl, and like I said, we will share the recording and the deck and also the companion resource that I mentioned. It will have a full accounting of all those terms and their definitions. So no worries if that was a bit of a blur. But again, we just wanted to set the foundation for the rest of the conversation that we're going to have today. So I am going to go back to our slide deck. Hopefully, this goes the way that I want it to.

Lee and Carl, are you back on our slide deck now? Yep. Awesome. Thank you again, Carl. So Lee, we're going to hand it over to you, and as Carl was defining those terms and concepts for us, he first showed us Claude, which is an example of a generic AI tool, and then he showed us AcuityAI, which is an example of a MedTech AI tool. So could you help everyone understand what is generic AI, what is MedTech AI, and what's the most important thing to understand about how they're different from each other? Yeah, absolutely. Important question to parse through as we're talking about the application of AI, especially in a very nuanced space like MedTech. And so I think just to kind of lay out the definitions pretty directly, generic AI is really built to be useful to everyone, regardless of industry, right? So it's also known as horizontal AI, and I think most folks at this point are pretty familiar with general purpose AI tools like Claude or ChatGPT, at least the out-of-the-box versions. So that's how you can think about generic AI.

MedTech AI is really built to be useful for people who work in MedTech. And so Carl just defined an AI harness as a set of tools and skills that really allows an agent to complete tasks and do it correctly, do it efficiently, do it safely. And so an analogy that you can think of is actually picture a physical harness on a rock climber. Rock climbing with a harness and rock climbing without a harness are obviously two very different things. And so what separates MedTech AI from generic AI is the harness. And what I mean by that is if you put Claude, let's say, on top of a spreadsheet with some market data, that's really not MedTech AI. And you may say, "Why is that?" And the answer is because Claude doesn't speak MedTech. It doesn't know our language. It doesn't understand the relationships between networks and sites of care and providers and the procedures that they're doing. It can't see the pathways that bring patients from provider A to provider B. And so if you ask Claude to create a strategy, let's say, for getting in front of a certain physician, it really can't do a great job. It doesn't have all the right tools, all the right skills, or in other words, it doesn't have the right harness.

And so MedTech AI does have the right harness, and that's really what separates it from generic AI or using kind of general purpose AI tools for a very specific application like MedTech Perfect. Thank you, Lee.

Makes a ton of sense. But just to really drive home everything you just outlined in your expansion on the concept of a harness, we have a couple prompts here on the screen. These are two hypothetical prompts that you could imagine a MedTech rep or somebody here in the audience. Two prompts you could imagine one of you all giving to AI to hopefully get a helpful output. So Lee, could you just help us understand how these two prompts, how they contrast and how they help sort of illustrate the key difference that you just walked through? Yeah, absolutely. These are good basic examples to go off of that you may actually use on a daily basis. And so let's just read them quickly. So prompt one is a basic one. "Use these notes from my conversation with Dr. Jones to draft a follow-up email." Basic prompt.

And then the second, a little bit more strategic. "Rank my top five ACL surgery prospects, exclude hospital-based procedures from the volume analysis, and exclude any competitor KOLs, and give me an engagement strategy for each of them to help me get meetings booked." And so there's a lot of layers to that second one. And so I think the takeaway here is generic AI is going to be able to handle prompt one, but it's not going to be able to handle prompt two. It's not going to have the skills and the context to be able to connect all of those different variables and layers to actually generate a valuable and accurate answer that the end user can trust. And so, the MedTech AI is of course going to be able to support the first one, probably do it with maybe slightly more refined language that's in your terms, respects the information about your organization and the products that you sell.

And so that's maybe a little value add there, but the big difference maker is obviously in prompt two, where it is going to be able to connect those dots, surface all of that context on which physicians actually fit that criteria and hopefully, ultimately get that rep a bit closer to hitting quota, and doing it in a very efficient way. So yeah, that's the breakdown there. Perfect. Thank you, Lee. So now we understand now the difference between generic AI and MedTech AI, and I know that there's a lot to digest here. Something I want to pull out for you all in the audience about this distinction is that both can be valuable to you all and to your teams. It's not that a generic AI tool like Claude has absolutely zero value for you all, for your teams. That's not what we're saying, right? It's just that, as Lee was just saying, with generic AI, it's really about these tactical use cases where you're automating some busy work and putting some time back in your day. Great. Whereas MedTech AI is reaching that strategic level as well, pointing reps to their best opportunities, helping them capitalize on them.

All that is to say that it is worth us taking some time here, just knowing that many of you probably have access to a generic AI tool and may not have access to a MedTech AI tool like AcuityAI. So it is worth us taking some time now to demonstrate a couple of recommended use cases of generic AI. So that's what Carl is going to do. Carl is going to pull up Claude again, and he's going to show you all how to use it to, one, anticipate objections from physicians heading into a conversation, and then two, also heading into a conversation, extract the so what or some key takeaways from clinical studies. So Carl is going to do these two demonstrations, and then just for the sake of not jumping back and forth too much, Carl will then hand it to Lee, and Lee is going to demonstrate a few MedTech AI use cases. And Lee's going to do these demonstrations in AcuityAI. So Lee will show us all how to use AcuityAI to find opportunities for new business, get a foot in the door with a particular prospect, and then find referring physicians to educate.

So let's start with Carl with these generic AI use cases. I'm going to stop sharing my screen. Carl, if you could pull up Claude and walk us through those. Awesome. All right. Okay, so now we're coming back to Claude, and I want to do the use case of, I'm prepping for a meeting with a cardiologist, and I want to share some product information with Claude and ask it to help me prep for that meeting. So I'm going to put my prompt in there. I'm going to add some additional context by dropping in this file, which is some product documentation. And I'm going to ask Claude to look through it and try to help me prep for the meeting. And you can see with the right context and the right prompt, Claude is able to do a good job of giving me some really good questions that I should prepare for that meeting. Similarly, let's say I'm still prepping for a meeting with a cardiologist, and I have a specific clinical study I also want to be tested on a little bit. So I'm going to use Claude to help me test me on that. So I've added the context of this clinical study as a PDF, and we'll give it a second to read through that, analyze it.

And then we can see here, Claude has done a good job simulating what a cardiologist might care about most within this study. Now I want to shift gears. We're actually going to switch to ChatGPT to show kind of different systems. Here's a prompt that we'll try where ChatGPT will start to hit the limits of what it can do as a generic AI. So I'm asking to find the top 10 surgeons in Massachusetts that are my customers, that are total knee arthroplasty surgeons. So it's looking for what it can.

And what you see is it does pull up a list, but it's giving me some certain volumes that are only based on Medicare. The data's about two years old. So let's shift gears. I'm going to stop sharing my screen and we're going to have Lee show what this can look like in a specific MedTech AI.

Great. Thank you, Carl, and we'll kind of pick that back up here because I think, again, on the surface, there is some value to a generic AI tool. However, as Carl kind of pointed out there, it is going to reach limitations as you start to get into the nuances of the data and the output that it's producing for your organization. So let's kind of pick that thread up in AcuityAI. And so I just plugged in the same prompt. Give me the top 10 total knee arthroplasty surgeons that are not yet customers in Massachusetts.

And again, that's going to reference the whole system. It's going to look at, obviously, the most recent volume. It's also going to look at any integrated data around whether or not they're customers. So we're going to be able to identify potential white space through this. And great, now we have volume that's up through the middle of this year, and I've got my top 10 list, which I can be confident about as a starting point if I'm evaluating the market and trying to find the true next best opportunities that include all that information. So now let's go a layer deeper and, let's say we want to get in with Dr. Lehman and want to understand, maybe if there's any connections that we have there, understand a little bit more about him if we're trying to open doors to that position. And so I'm going to just ask the agent to give me ideas to get a meeting with Dr. Lehman, check his peer network for existing knee customers or product users that we may already have across our customer base. Maybe there's a peer-to-peer introduction that we can leverage or a connection that we don't know about across the country.

So I'm going to plug that in, and now this is going to search across, again, some internal sales data, customer data that we may have. It's also going to look across our database. It's going to look at where he practices, where he operates on certain days of the week, which clinics he might be in. It's going to look across his peer networks, where he did fellowship, who he's published with. And so it's going to scan a ton of different data points, and then put together some recommendations. And so, right off the bat, we have an access strategy. We've got contact information. That might be a good starting point. We also have some best paths in. So existing knee users in his network. Great. We have two physicians here that looks like are customers and co-authored a paper with him. And then we also have three fellows who did fellowship with Dr. Lehman and are customers of ours, obviously, in this simulated environment. And then in addition to that, we've got the office staff at New England Orthopedics.

And so, some peer-to-peer, maybe friendly intros, also some, if we're going to do this organically, some contact information that might be valuable. And also timing on where he is. He operates at different facilities. And so there's going to be a few different avenues that we could go after to get into Dr. Lehman.

And then let's take this a step further and forgive me, I'm just going to copy my prompt here. But let's say we actually go through the process fast-forward, and we convert Dr. Lehman to a customer. He starts using our knee. Fantastic.

And he's a high-volume surgeon. We might want to find referring providers in the area to educate and hopefully, improve our position in this market. So I'm going to say, "We converted Dr. Lehman to a customer. Now find me the top referring physicians to go educate." Great.

And right off the bat, in his geography, we are going to get a list of physicians. We're seeing a lot of PCPs that have referred into providers in the area specifically for total knees. And so again, when we think about market development and investing around KOLs, we're obviously fast-forwarding through the sales process here, but this is the flow of how you can see AI applying to a number of different workflows throughout the go-to-market process, and the customer education process as well. So really rich output that's referencing a lot of information, internal and external data points, and putting it together in a way that's really logical for the end user. Great. Thank you, Lee. Thank you both for the demonstrations.

And again, folks, you'll get the recording. I'll also, when I go back to my slide deck, I'll make sure to recap the use cases we just walked through. Something I do want to call out, though, Lee, just before we move on. As you mentioned before, the primary difference between these tools is the use cases and the tactical versus strategic.

But something else I just want to call out that's sort of a secondary benefit to what you just walked through is that it's a bit easier from a user experience perspective. Carl showed us how he had to go find specific documents, attach them to the chat, be very prescriptive with Claude. Whereas you, Lee, you were basically just texting with AcuityAI, and it already speaks the language. It knows where to go look for certain information. So I just wanted to call that out as another difference in the two demonstrations there.

And I will, like I said, I'll resume my slide deck here. Lee and Carl, hopefully you can see that. Yep. Great. Again, just to recap, what Carl showed us was how to use Claude to anticipate objections from physicians by attaching prior documentation and asking Claude to basically step into the role of a cardiologist and come up with some objections. And then Carl also attached a clinical study, and asked Claude again, in the role of a cardiologist, to pull out the key takeaways, what's the most compelling. And then Lee, switching over to AcuityAI, showed us how to find new business opportunities, how to get the foot in the door with a particular doctor, and then how to go find the referring PCPs. So, again, the goal here was for us to give you all some use cases that you all can try for yourselves, whether you have Claude and ChatGPT, whether you have AcuityAI. We wanted to give you all some inspiration on some things you can start doing and trying out and socializing with your teams, and everyone can start building some more momentum, getting more comfortable with these AI use cases.

And Carl, like I just said, some folks are going to leave this webinar, they're going to go try out some things in Claude and ChatGPT, Copilot. Other folks are going to leave the webinar and go try some things in AcuityAI. But here you have a few sort of universal tips. No matter what tool you are using, these are three universal tips that will help the folks in the audience get more value and get to valuable outputs quicker. So if you could just walk through these for us all. Yeah, absolutely. So the first one is really being specific about the output you need. I think it really helps to think about AI as almost like a really eager assistant with not much experience. So you really need to be specific about what are you trying to get, what's the result that you're trying to work towards, so that the AI can take that task and get to you a really great result. Earlier, I used the example of asking AI, "Hey, write me an email."

It's not very specific. It's going to write a very undefined email that's not going to be super helpful. And then I added more context to say what I wanted to cover in the email, why I was writing the email, and I'm going to get a much better result out of that latter one.

The second one is a good one that sometimes you're not sure exactly what output you need to work on, and you can really think of AI as more like a thought partner or like a buddy that you might loop in here. So rather than starting with exactly what you need, explain the situation that you're in, the problem that you're trying to solve, and then ask AI what can it do to help you through that situation. Because there might be tools, there might be ways that it can work that you're not aware of, and it can be a really helpful thought partner in that regard. And then lastly is really going one step further. So trying to tell the tool to actually write the prompt for you. So sometimes when I get stuck, I'll say, "Hey, this is what I'm trying to do, but could you let me know how I would write the prompt for you that would be most successful to get the output that I want out of it?"

And that's a really helpful tool that if you're feeling a little bit stuck and you're not getting exactly what you want out of AI, turn the question back on AI and you can get a much better output. Yeah. AI knows itself well, right? So you don't need to guess all the time what's the right thing to say to get the right output. Sometimes the easiest thing to do is just have the AI tell you exactly what you need to give it, and that's the most efficient way to get to what you're looking for. Thank you, Carl. And, Lee, before we get to some questions from the audience, you had a few action items. Again, I opened up by talking about how I think we're all just trying to make some positive forward progress, build some initial momentum. So you've got some tips for folks, some things they can go do today to start building that momentum.

Yeah, absolutely. It's really valuable to talk about good starting points with deploying AI into your organization. And so, a couple ideas here. Number one, just identify a recurring workflow that consumes a lot of time, right? It's always valuable to save folks time, especially in the field. And so that could be prep for sales calls, that could be building pipeline. It may vary a little bit depending on your organization. But, there are certainly workflows out there where there's an opportunity here. And then literally list out the individual steps in that workflow and try applying AI to one of those steps, right? It's really easy to do trial and error. It's one of the great things about AI. And so kind of seeing where the efficiencies are and where there might be a lot of value to gain, I think is a great starting point. So whether that's like we just went through, researching a physician's peer network or defining your top markets or even refining your ICP and getting into your top prospects, that may be a great place to start. And then I think most importantly, standardize the workflow, right? It can't be optional. It needs to be mandatory.

It needs to be built into your organization and built into the culture of how you're operating. And so I think that piece is really important. But at the same time, while it should be mandatory, you should also make it as easy as possible. That's also one of the benefits, comparing AI even to just kind of legacy SaaS. You're no longer filtering through dashboards and needing to rely on everyone in your organization to click all the same buttons and add all the right filters. All these types of things can be set up with quick actions where it's right on your home screen, you click one button. And so it is, in a lot of ways, easier to standardize now, and I think that's one of the most important things about getting started.

Perfect. Thank you both again. We're going to wrap up here with some questions. I do want to say, though, before I forget, like I said earlier, we do have a companion e-book. So pretty much everything that was covered over the course of today's presentation is covered in this e-book. And as Lee and Carl answer your questions, I'll actually drop the link to this PDF so that you all have it accessible before you go. But this is just another resource that you all can refer back to after today. In addition to the recording and the deck, like I said, all the content we've covered today is in this PDF e-book. So I'll make sure that you all get that in the chat. But before we go, we did get a few questions that I want to make sure to service for Lee and Carl. So I'm going to stop sharing. So Lee, we'll keep it with you for now. One question we got.

We've heard this before. Some folks worry about, even if they themselves are pro-AI, they're totally bought in on the benefits, sometimes that's not the case for everybody on the team. So any tips that you have for folks where maybe they have some team members who might be a bit more resistant to bringing AI into their workflow? Anything you can think of that can help there. Yeah, absolutely. I think like anything, two things. You need to show the value to them, and you need to show them how easy it is. And one way to do that we've seen be very effective, and it kind of translates into the MedTech space in a lot of ways, is a peer-to-peer, right?

Highlight someone, promote someone who has adopted it and has seen value in their day-to-day. And I think taking that to the next level, finding folks with different backgrounds, right? Even more tenured reps who know their geography really well, have been in the space a long time. There's more that they can learn. There's more that they can uncover. And sure, maybe it's not as basic as learning the top accounts in their geography, but if they can get an early signal on changes in their market or new physicians entering the territory, anything to get them a leg up to help their performance, it's not really a drag on them. It should be a value add. And so I think kind of showcasing that through success stories with folks that are early adopters is really a good way to do it.

Yeah. And thank you. And then somebody else sort of on the flip side, there's the question of how do I prove that value to folks on my team? On the flip side, sometimes it's a matter of talking about the value to an executive, to a senior stakeholder, and sort of making the business case because folks are cost conscious. Folks know that AI usage costs money. So any tips you have for folks in the audience today who need to go make the business case to a budget holder to unlock some more AI usage?

Yeah, absolutely. I think first and foremost, you need to consider what your organization is prioritizing in terms of AI and what initiatives are actually underway. And I think for the most part, we've seen a big shift over the last six, eight months where C-suites and boards are really encouraging this industry and a lot of the organizations that we work with to adopt AI in some capacity, right? And I think you need to look at which workflows make the most sense, where there's a lot of efficiency to gain, where there's top line growth to add, or even bottom line growth as well, and look at the trade-offs, right? Token spend is a major concern. Trying to build your own solution costs a lot of money as well. And so there's a lot to evaluate when you're looking at the business case on building versus buying and what solutions exist in the market and what the trade-offs are. And we talked a lot about the harness concept today, but as you're thinking about initiatives that you're considering, I think you really need to weigh all those variables. Think about the token spend that you may expose yourself to by building your own solution and some of the limitations along those lines.

And essentially put all those pieces together to make the business case, because if you're not using AI in some capacity at this point, I think it's pretty clear that you're falling behind. So those would be my general tips. And obviously each situation is unique, but that's what I would say. Perfect. Thank you, Lee. And then Carl, we got a couple folks who had a question I think is for you. So some folks were hoping to return to the concept of the harness. You defined this for us earlier, but I think for some folks it's sort of a new concept. It's a little bit abstract. Would you mind just sort of reiterating what you shared earlier about what the harness is and why it's so important in this generic AI versus MedTech AI conversation?

Yeah, absolutely. It's something that took me a little while to wrap my head around, too. But basically, the harness is all the software that's wrapped around an AI model. I think a metaphor can be helpful here. So I would think about the model as the brain and the harness as the body. So the body gives the brain the eyes and the ears to take in information, like the hands to use tools, the memory of what's happened so far, the nerves to carry out instructions. And then the brain gives all of the power, all of the critical reasoning, the thinking behind that. So without a body, a brain can really only think and talk. And without a brain, a body's just a mannequin. So you really need the combination of the two to be more powerful.

And related to that, the harness has a big impact here, because if you think about a body that's perfectly built for one thing versus another thing, it's very different. Michael Phelps, perfectly built for swimming. He's a great person for swimming.

And that's similar to how AI harnesses work. You want an AI harness that's specifically built for the use case that you're thinking about, which in this case is MedTech. Got you. Thank you. We have a question that just came in the chat about a particular use case. Lee, maybe you'd be open to answer this one on the fly here about using Acuity to find new distributor or rep business partners. Any thoughts there for Tyler? Yeah. It's a good question. So to answer it directly, we don't have a broader database of independent distributors out there.

So it's something we have heard before. We've looked into a bit. I would say, though, as you are bringing on distributors, if you do go through an independent channel, one thing that we have seen quite frequently is using this during your interview process and actually vetting them out to say, hey, if they're going to bring in this account or these physicians, is what they're telling you about them actually true? And you can use it as a validation mechanism and also as a way to drive accountability with the distributors you do have and making sure they're hunting in the right places.

So that's what I would say. Happy to talk more about that one offline, Tyler, but great question. Got you. Thank you. And then we've got one more here that might also need to be taken offline. This is more of a technical question, so I'll pass this to you, Carl, in case you have any insight here from Mark. He was just wondering specifically with regards to the harness about how it's powered, whether there's an API there or a fine-tuned open model. I don't know if you can speak to that here. Yeah, I can speak to it. So AcuityAI runs on the leading third party LLMs. Excuse me. So we're always upgrading what we power things on. We're always testing to make sure that we're using the most effective models that will deliver the best output within AcuityAI.

Your data is always staying in our platform. We're not using your data to train those models. It's just about what we've built on top of that. So the harness that we've built on top of that is what makes the AcuityAI answers really helpful specifically for MedTech. Got you. Well, Lee, Carl, thank you both so much. That does it for the audience questions. Thank you all again for being here, for bringing great questions. Again, look out tomorrow for an email with the recording of today's session, a copy of the slide deck and that companion e-book, and we will see you all on the next one. Thanks, all. Thanks, everyone. Thanks, everybody.

‍