PH360^™ AI Webinar

Welcome

Juliana McMillan-Wilhoit 0:15

Welcome, everyone. We're so thankful and excited that you are here. And you should see a poll that's going to pop up here on the screen.

Brett Emo 0:15

Yeah. Yes.

Juliana McMillan-Wilhoit 0:26

And. for you, but you can also enter it in the chat as well.

Jefferson McMillan-Wilhoit 1:02

All right, everyone. Good morning, afternoon, evening, depending on where you are. Thanks so much for joining us today. My name is Jefferson McMillan Wilhoit, and I am the CEO of Flourish and Thrive Labs. And we are just super excited to have you here today to talk about something that's really important to us and really important topublic health as a profession. Just a real quick housekeeping item.

There will be things happening in the chat. So if you're not used to having chat open on Teams, make sure that you do have that open. We're going to be having a lot of interaction today. And so the chat's going to be where you can do that. So we are all here to talk about AI. And I think all of us have probably heard about the promise of AI in public health.

And I think that we've all been talking about it and you've heard other people talk about it. SoInstead of sitting here talking about it for 20 minutes, I'm just going to jump in and show you in real time how you can use AI in your workflow today. So we're going to start out with something that I'm sure all of us have experienced before. You know, 4 P. m.

Friday afternoon, you're the new person in your, you're the new person at your organization. And you get a call from a community member that you aren't really super confident in answering because you're new. Right? You haven't really gotten, you haven't gotten the hang of everything yet. So... My background is in epidemiology. So that's what I'm going to talk.

That's the we're going to relive a bit of my trauma from an early professional encounter today. And we're going to see how AI could have really helped me in my in that in that situation. So I got a call from a clinician at one of our local pediatric practices reporting a child who had a fever and a generalized rash, right? I look behind me, my boss's office was right over behind me, and I look and my boss is gone.

for the day. I have no one to transfer this call to, and now I need to deal with it. I'm 3 days in. I don't know what I'm doing, but I need to deal with it. So my training immediately kicks in, and I think that I immediately start thinking measles. But I don't know exactly what I need to do. do to confirm that suspicion.

So before, when I initially experienced this, what I would have done is I would have gone online, I would have Googled, I would have found the CSTE case definition, and I would have looked at my disease manual, and I would have done all of those things, and it would have taken me about 20 to 25 minutes while I have this clinician sitting on the phone with me.

And now I'm getting flustered and now I'm stressed and I'm likely to miss something. However, now with AI, I would just open up Public Health 360, we're going to refer to as PH360, which is a safe and secure AI platform built for public health professionals like you and I. and I get my answer in seconds. So what you just saw is that my agency had built an agent for me for this situation in PH360.

They had built it in our no code agent builder called Oak. And I accessed that through our chat interface called. Maple. I don't need to be a prompt engineer in order to use this. I'm sure that's something you've heard when you've been talking about how to use AI, that you need to be a good prompt engineer. I don't need to be in order to use Maple.

I just need to talk to Maple as if they were a public health colleague. And you can see that all I did was I said, I just got a call for a suspected measles case. Provider said the person has fever and generalized rash. What else do I need to ask for? provider for. It took about 30 seconds for that to come up, longer than, it's shorter than it was for me to even talk through the situation, right?

And I can now be on the phone with the clinician and I can start asking this information. What would have taken me 20 minutes to find took 30 seconds for Maple to pull up. And it cites the CST case definition for me so that I know that it's coming from a reliable source. So I review what Maple has given me, and I determine what to, I determine what.

I determine whether or not this is accurate before I ask the information of the clinician. And so I'm going to look through and it's actually done a really good job. It's grounded me in all of my local jurisdictional information. And it even gives me sort of a next step if I wanted to take it that way.

So while I'm on the phone, I ask all of my questions, I've reviewed everything that I need to, and I get off the phone and what I notice is that in my inbox, the provider has sent me a positive IGM result for the suspected case. Now, I don't know without going and looking at the CSTEdefinition or my local manual, I don't know if that necessarily meets my case definition or not.

I'd have to go look and I'd have to interpret that. But with Maple, I can just say, this is what I have. Does it meet the actual case definition? And what it tells me is pretty is pretty insightful that it's that the case definition earlier that would have been sufficient, but now it's not.

So that's the difference between like what my training would have said versus what my locality has adopted or what the new CSTE guidelines are. So now I know that I'm not quite there yet, right? We're not quite at a measles confirmation. So I need to continue my work, but I really need to bring my supervisor into this. I need to notify them and let them know what we've got going on, because again, I'm new.

I don't necessarily have the authority to do what I need to do, but I don't want to go to my supervisor empty-handed. So what I'm going to ask Maple to do is draft an action plan for me that I can review and that I can use as a starting point to send the plan to my supervisor. Maple has all of the information in it that I've already given to it.

And so it's going to use all of that information plus all the things that my agency has built to give me a good action plan. And then I can take that action plan, I can copy it into an email or a Word document or whatever, and I can review it, I can add things into it that I might need to. You'll notice that Maple actually does give spaces and prompts for where you should put things in.

And I can send this off to my supervisor. And that's it, right? I now have gotten all the way from, I had a call from a community member, all the way to telling my supervisor what my next steps are. And that would have taken me hours of work before AI existed and before PH360 existed. But I got it done in just minutes. And here's the thing, it never declared anything, right?

Never declared that the case was or was not measles, never did anything without my permission. And Mabel showed me the evidence as we walked through the process together. I stayed in control as the public health professional, and I used my public health knowledge to walk, to get the information I needed quickly to make that decision. And that, everyone, is the promise of AI in public health.

The judgment stays with us. AI gets us there a whole lot faster. And we have the ability to then do other things that we may not have otherwise had time to do. So that's why I'm going to start out today saying something that may feel a little jarring. And that is that public health must use AI. It is not an if. It is not a should. It is a must. And it's not eventually, we have to do it now. And here's why.

I started my public health career in 2011. So I'm sure that there are people on this call who have started long before I did, right? And I'm sure all of you can remember that even before 2011, we were talking about all the structural things we needed to fix to advance the health and well-being of our communities. Right?

We need, we talked about more funding and we talked about better infrastructure and we talked about all of these things. And guess what? We're still talking about more funding and better infrastructure. And those same structural fixes haven't been put in place yet. And so I'm not saying that we should abandon that fight. Right?

I will be the first person to be advocating for more funding for public health, more resources for public health, and more infrastructure for public health. It's a just cause, and it's a noble one, and we should keep pushing for that. But we and our communities can't afford to wait another decade when a tool exists that can help us carry out our mission.

Like we had always wanted, and we don't need our politicians to acton our behalf. We don't need necessarily a whole bunch more funding. We have the tools at our disposal.

So if you have ever been told to look into AI by some of your leadership, and you haven't quite had a path yet, or you've heard about the promise of AI but neverseen it actually work in practice like you just did in a public health workflow, this hour is going to be for you. What I just showed you in PH360 is not the future. It's available today.

Over the next hour, we are going to show you how to use PH360 — different parts of the PH360 platform in your workflow to get from data to decisions quicker and more reliably using a system that you can trust. Now. But I can't trust the guy. If that's the thought you just had, you're in very good company. It's probably the first thing that we hear in every conversation we have. And honestly, you should be skeptical.

In public health, if we're wrong, there are huge consequences. And we, the humans, are the ones that have to answer to it, not the AI. On its own, no AI should be trusted. It's trained to sound confident, even when it's wrong. That's part of it's the way that we develop large language models. But here's how I think about it. Trust was never something that AI came with out of the box.

It's something you have to build around it. You already know this from your own work, and I already know this from my work. You trust a system when it does what it does what you need, when you need it, and nothing more, and it does it right every single time, right?

So we built Public Health 360 to earn the trust in the same way, in three layers, with most tools, where most of the other tools you may experience only give you maybe one. of these layers. So the first layer is purpose-built. It's built for us. It's built for public health. We took the time to figure out how to make these, how to make the models work for us, how to have the right data available to us as we need it.

The second thing is guardrails. PH360 checks itself in both directions, right? Every time. That's what that little extra moment that you saw in the demo that I just showed you was. It's a feature. What it's doing is it's looking to make sure that it's accurate. And the third thing is the knowledge bases. All public health is local, we all know that.

So the agent answers from your local documents, your SOPs, your manuals, your case definitions, your style guides, everything that is. You. And you get to update them yourself. You don't need us. You don't need the vendor to go in and update it. You get to update that for yourself. The last thing, so those are the three things that create trust, but really the foundation is you, right?

Nothing comes back or a yes or no verdict. Cite the source, shows you the reasoning, walks you through, you get to make the decision. It doesn't do anything else with, it doesn't do anything without you. And that's the difference between PH360 and other consumer AI systems. So with something like ChatGPT or Claude or Gemini, the burden to figure out whether an answer is safe to act on is on you, right?

On top of everything else that you're carrying, all the other burden that you have. But we've already done a lot of that underlying work for public health. PH360 already has public health best practices built in. It's HIPAA aligned. We sign a BAA so that you can put your PHI and PII into it. And most importantly, it lives inside of your workflow, so that you don't have to relearn a whole new thing.

PH360 will tell you when it's not confident. and it lets you decide how to use that information. That's where it really pays off. And let me just give you one example. A county that we worked with cut their food code citation lookups from about 30 minutes to under 5, simply because for the first time they had a tool that they could trust, right? And it was already inside of their workflow.

So the question was never whether you could trust AI. It's whether the systems around it can earn your trust. And that's what we've built, and that's what we want to show you today. And one of the big pieces of that trust is that you are able to update when information changes, because we know that in public health, information changes all the time around us. So I'm going to hand it over to Brett.

to talk us through how in PH360 we update that information to make sure that you always have the right knowledge at the right time.

Demo — A suspected measles call: case investigation with Maple

Why public health must use AI (and what PH360 is)

Demo — Updating an agent's knowledge base with Oak

Brett Emo 16:33

Thanks, Jefferson. Hi, everyone. My name is Brett Emo. Happy to be here speaking with you today. So what Jefferson alluded to is we're going to talk about the agent that he was using within Maple. So that was the communicable disease agents trained on specific information.

In this case, one of the documents is the CSTCase definitions, and as you know, these definitions can be updated, so there's new conditions added, information updated, new testing. It happens periodically, but it does happen, and when that happens, we need to go back and update the agent, and so what we want to show you is the simple process.

through the no-code agent builder to go about updating guidance, new guidance as it arrives. So Jefferson is going to, see, it's logged back in. Jefferson is going to.

Jefferson McMillan-Wilhoit 17:21

Right, that's the security, the inherent security, right? I was, I didn't do something for a few minutes, so it logged me out.

Brett Emo 17:24

Yes, you get to see. Yeah, so you get to see security in action and multi-factor authentication happening in real time. So, here he has accessed the agents, so the back end of the agents, the no-code agent below that I mentioned, and it already exists, so there's some information already here, description what it's called, of course, and as you advance along, you see... A few other bits of information.

There's some instructions how it actually works. Like I said, there's no code here. This is just describing what it's supposed to do. And finally, so he's navigated now to a list of the documents that. that are the service knowledge base for the agent. And so in this case, there is a number of documents here, if you scroll down. So this is position statements from CSTE.

There's the case definitions itself, there's field guides, there's a number of documents here that are forming this agent. We just need to update one of those documents, and that's the CSTE case definition. So youLocate the file. And you, yeah, you just get rid of it that easy. Now, we're gonna update it, so we have the new file ready.

You drag it and drop it right on top, and that's gonna process the document, so digest it and prepare it for use by the agent. There's something I wanna point out while it's here and it's loading: Jefferson is the owner of this agent. this knowledge base. He's the expert, we'll say. He's the owner of it. He shares it with me.

I don't have the ability to edit it, and that's a good thing, because that way you can kind of control the messaging and everything. So Jefferson actually being the owner of the agent at the knowledge base is the one who could update it. So that's an important feature as well. And so this sort of demonstrates the no-code agent builder. You've seen most of the screens already. It's that simple.

And so that's pretty much the extent of what's necessary to update information. You don't have to contact us or some other vendor to update information. This could all be done by you. Your staff will build the agents. Your staff will maintain the agents. Right, and so...

The thought I want to leave you with, I mean, we've just seen this demonstrated and how quick and easy it is, is what's that document you would use to build that burst agent? Right, so think about that. And then as I think about that, I'll transition over to Nebu, who is going to be talking about how AI fits into workflows. So Nebu, go for it.

Where AI fits in your workflow — and where it doesn't

Nebu Kolenchery 19:58

Thanks, Brett. Hi folks, my name is Nebu Kolenchery. It's a pleasure to be with you here today. So Jefferson opened this claim, this hour with a claim, a pretty bold one that says public health must use AI. And now you've watched a couple of demos on how they can do it within our system. SoLet's talk about what you just saw. Like, think about that measles case you just saw, right? An epi gets a call.

They need to figure out what they need to do with it. They need to consult a reference guide. In this case, it was the CSTE case definitions, and then make a decision on whether or not it meets the criteria, what else is needed. and then comes up with a product, in this case, an action plan to their supervisor and sends it. So there's this five parts to that workflow. There's these steps.

So there's an intake, a process, a reference, creating an output, and then communicating that back out. So if that kind of sounds familiar to you, it's because that's kind of how like. Every public health workflow works. That's a step in just about every task that we do in state and local health departments every day.

Your environmental health team does this when a complaint comes in about a restaurant or, you know, the intake is the call, the reference is the food code, the output is maybe the violation. inspection report with the violations and they communicate that back to the establishment of the restaurant owner. Your grants team might do the same thing. Let's say that a progress report is due.

The intake is that that request or the task to do the deadline, the report. The process is beginning to understand what you all have done so far. the reference and maybe the grant documentation or the guidelines or the SOP. And then you draft that grant report and communicate it back out. Even outside the scope of sort of core public health workflows, your HR team might do this.

Or if you're a supervisor, maybe someone asks you about comp time. after a week in outbreak response where they work 10 hours on a Saturday. It's the same thing, except the reference in this case is your HR manual. It's the same five steps with all across public health workflows. You know, different reference documents, maybe different people, but it's that same, this loop every time. SoWhere does AI fit in?

Well, it takes this and compresses the time that it takes in order for you to do this. And it maintains you at the center, you public health professional, and more importantly, it maintains your judgment. So your emotions, your head, your heart, that's that it maintains that in there. SoAt the same time, it can make every step faster. So it took the Jefferson, let's say, 30 seconds to look that case definition up.

If that, you know, I worked in communicable disease, that would probably take me closer to half an hour. And what are you going to do with that 29 minutes and 30 seconds? We'll do more public health work. The reason you got into this in the 1st place, which isto help people, I think.

But the problem with the current state is by the time you've hunted through the manual, you've searched, you know, 3 PDFs and cross-checked the case definition, drafted the feedback, you're exhausted, you're rushed, and the step that actually needs your full attention, which is, you know, the judgment, the interpretation, that professional decision. It gets whatever's left of your brain at the end of that day.

On the other hand, let's flip it and say, use PH360. The rest of the process takes minutes instead of hours. You show up to that hard part fresh. So you reserve the part of your brain that you need to think. You have the space to be careful. The decision still belongs to you, but now you're just making it with your best thinking. and not your Friday at 630 dreaming of happy hour thinking, right?

So that's why what Brett showed you exists. It's every department has its own version of this loop with its own documents and its own expertise. And that's what PH360 is specifically designed for. So public health must use AI inside its workflows and not next to them. And what Brett and Jefferson just showed you is that we've designed this, which with, we've designed P 360 with this in mind.

So as you just saw in the demos, it has two parts. Jefferson opened with Maple. They're named after trees because we like trees. No real reason other than that. So Maple is chat. It's the chat interface. You ask the question in plain language, and the public health reasoning engine works the problem, checks the references, and show you how it got there. You can also select agents that you create here.

So Jefferson selected the CSTE case definition agent for that measles demo. Most of your staff would use Maple in their daily work. And then what Brett was talking through was Oak. Oak is the agent builder. So when you have a workflow that you run over and over again, give Oak your documents and your steps, and it becomes an agent that you can hand off to your team.

So the supervisor theoretically would create an agent, test it, and then send it to their team. No code, no waiting for a vendor to set it up for you. We'll train you on how to build one and then hand it over to you in order to do it.

So we've got clients that have taken this and created agents for their state's communicable disease system, for food code, for emergency preparedness, navigating QI projects for, you know, tick-borne disease reference manual. Each one took their stock, their documents, their workflows, the stuff they already had, and turned it into a tool that their whole team can use.

So to recap, Maple is for the question right in front of you, and Oak is for to create agents for the workflows that you run every week. So watch for both as we go through the rest of the demos. And back to you, Jefferson, to share some more of our tool.

Demo — Confirmed measles case: CDC guidance, cited

Jefferson McMillan-Wilhoit 25:37

Yes, thanks, Nebu. So Nebu mentioned that we are constantly, as public health professionals, going through this loop of intake to communication, right? And we sort of always go through that. And so the one thing, so we've seen a lot about the intake, we've seen about what we do sort of in the middle there and how we can change information. But one of the things thatwe haven't seen yet is how do we do communication?

How does PH360 help us do communication? And in, we're going to pick up right where we left off with that measles case. So in this case, as the Epi and with my supervisor, I have, we've determined that that measles case is confirmed. So now we need to send out our health alert notification, right?

And we need to send it out toOur local clinicians, our hospitals, anyone who we have on our HAN list, and before AI, what I would have done is I would have gone and found our HAN template, and I probably would have looked at a couple of previous HANs, especially since I'm new. I would have looked at a few previous HANs, and then I would have started compiling all of my information.

using all of the case definitions and pieces that I needed to do. Now, what I'm going to do here, last time I showed you with an agent. So that was just, so that was using something that my agency had already put together. I want to show you how powerful Maple is just on its own. with its public health reasoning engine sitting underneath it.

And so what I'm going to do is I am going to give it my prompt and I'll talk you through as it starts to generate. So all I'm saying is I need to generate my Han for my clinicians for measles case in Emerald City. We're much like we are big fans of trees, we are also big fans of the Wizard of Oz. And here's why it gets interesting. I actually put protected health information in here.

And you guys are probably like, Jefferson, how could you put public protected health information in an AI system? Well, PH360 is built for that. We have agreements in place with all of our processors to make sure that we comply withall HIPAA guidelines so that we help you maintain your HIPAA compliance.

So I can put in things like I have Glenda who's age 6 with a date of birth with a confirmed case of measles and they don't have an MMR vaccine. That is context that I couldn't put into another AI. There would be no way.

that I would even dream of putting that into ChatGPT or Claude, but PH360 has mechanisms behind it to protect that information, to make sure that that information never makes its way to a model, and it's never trained on, the model is never trained on it, and none of us at F&T can ever see it, and nor can anyone else in the system. It's all protected inside of your, I like to call it your little home for AI.

So let's take a look at what it gave me. So it's interesting. It told me that June 4th, 2020 would actually be five years old because they would be turning 6 tomorrow. So it would be, I'm a little inconsistent. which is a nice thing, right? Gives me that little poke. And then HAN advisor, and it tells me don't include the patient's name, right? Which is great. And then it gives me my HAN advisory.

So here's my alert summary, my background, what clinicians need to know, exposure conditions, vaccine recommendations, reporting. It gives me some places tofill in some information, and then any additional resources. So in this case, it's citing its source, which is the CDC measles for healthcare providers.

And then you would put in any of your state health department website, local health department stuff, any of that good information. So what I can do is I can now just take this, copy it directly into my HAN template Word document, and then I can send out the HAN in the same way that I normally would.

I've reviewed it, I've checked it, I've madesure that it has all the information I need into it and instead of taking an hour and a half to find all this information, compile it, get my Han ready. I have a first draft in about 45 seconds and I just now I'm taking the time to review it, make sure it's accurate, send it out to everyone. So. That's really the power here, right?

Is that even for communication, I can get that answer really quickly and I can get to decisions very quickly. But this is the clinical side, which, you know, is important, but how well does PH360 do for the community, for sending out information to our community? Members, and for that, I'm going to hand it over to Juliana to walk us through.

Demo — Drafting a community social media post from case data

Juliana McMillan-Wilhoit 30:23

Hello, I'm Juliana and I am not a communicable disease person at all. I'm the marketing gal who cares really deeply about telling our story and just communicating with people. I really care about those relationships and making sure that you're continuing to build public trust and how to cut through the noise and make sure that our message isn't lost. So how could AI work as part of that workflow?

I know my community really well. I know it resonates. I know what doesn't resonate. I also know that there's best practices in terms of how to actually communicate information. So using Oak, I created an agent that is something you could also add to your PH360 instance that has the best guidance on writing social media posts.

in terms of what to do, what not to do, as well as the best practices for science and health communication. So with this, your posts then come out as being something that's engaging and communicative. So instead of me just sitting and staring at that little box on Facebook and being like, what in the world do I post? I now have something here. I have a draft post.

that's being able to communicate out to the wonders of Emerald City and tell them about our confirmed measles case. It includes things like, it knows that we should, you know, include the signs to look for. That wasn't part of my prompt, but because it knows public health, it knows both what thosewhat those signs to watch for are, as well as the importance of including them.

So this is where AI is working inside of my loop. I'm going to take this information, I may edit it, but I'm no longer just staring at that blank page, not being sure what to do, as well as being able to rely on those frameworks that we know are incredibly effective. So you're no longer starting from scratch.

And it's going to be very different than what you would get from a generic AI tool, because this already gets public health. So now we've informed the community, and now we have another step, which is that we want to communicate back to our funders. And so I'm going to pass it over to Brett. to talk about that.

Brett Emo 32:30

Everyone, I'm back. And now, you know, Juliana is the marketing guru and I'm the one who does the reporting. You know, they told me, you know, you have a PhD in public health, you must love writing reports. And I'm going to be honest, I don't really love writing reports. And so, This is an area where in my workflow, I think it could have a big impact.

And so what you see, Jefferson has put into the prompt, just a summary. And I'm not going to read it all, but it's basically a BLOB summary in poorly structured English about what we did, right, related to this case. And the idea is it's going back behind the system and looking at reporting standards that we've provided it. So let's say it's a grant funder and they need specific information.

It's the state and this is a report related to an outbreak. You know, whatever the case may be, it's able to go pull those standards, look at what we've provided, and from thatBLOB of information, it's able to generate a draft report. I will say it's a draft report. It is still my responsibility to go through and look it over, make sure all the information is complete and accurate.

But you could see how long that took versus how long it would have taken me to gather all the information together in a way that was meaningful and met the standards and the protocols to report. It also demonstrates the power of that loop that Nebu was talking about. The intake was a lot of information about the case.

You had that process about working through the case, really gathering the information that we just supplied it. The reference was all the materials sort of under the hood, so that's the reference materials.

The standards for grant writing, the standards for the reporting, there was that interpretation of the AI, and now I have a product that's near ready for communication, it just needs my confirmation, so you can really see how that works, that coordination that I would have to do, that back and forth email.

Now that it has saved me, and everything I need is there ready for my review, and so with that, gonna hand it back to Nebu, and he's gonna walk through how to talk about this with it.

Nebu Kolenchery 34:52

Thanks, Brett. So, there's a poll on the screen right now that's asking you, what is the one thing standing between you and using AI at work? And I would love to sort of see your responses to that. But as you're answering that, talking to IT is actually thebiggest approval, sorry, the biggest sort of barrier that we hear as part of when we have these conversations with health departments.

AndYou know, it's not that people don't want AI. It's not that leadership says no. It's that somewhere between excitement and implementation. Someone kind of says that we need to run this through AI. And we agree with that. So they should. IT is not wrong to ask these questions. They're actually right. But they should be asking these questions because most AI tools don't have the answers ready.

We do, and we want to make sure that you know that while you're still on this call. So as you can see the slide on the screen, you can also, Juliana, I'll drop a link in the chat for you to look at, which is trust. fntlabs. com. And... So I'd love for you to do something for me. Please take a second, take a screenshot of the slide or grab that link.

And if you're open to it, please text or email it to your IT leadership now. You don't have to wait for after the webinar. Please do it right now. If you're okay with it, I'd love to wait 3 seconds for you to do that. I'm going to trust that everyone on this call just did that. So thank you for doing it. So here's when you do it. Here's what your IT team is going to find. Two things that they care about most.

First, we're HIPAA lined, and we'll sign a BAA with you. A BAA, for those of you who don't know, is a business associate agreement that says you can trust us with your data and specifically protected health information. It's downloadable at that link, and you feel free to send it to them. The second is that your data is your data. It stays in your tenant.

It is never used to train anyone else's model, and it won't leave your tenant within the tool. So that actually comes within the architecture. Here's some stuff they might not expect. full records retention. So every interaction that you or your staff have on PH360 is retained.

This is just in case you ever get a Sunshine Law request, FOIA, or open records demand that requires you to show what you've been using AI for, you're covered. We built it for that because we've sat in your seat and know that question may come in the future. The rest is on the slide, you know, SOC type 2, encryption in transit, US data residency.

All of this will make sure that, you know, your IT team will be very happy. They can actually verify all of this at that link and can reach us directly with questions. So what we've really learned from working with health departments is that this is a real blocker and we want to make sure thatWe've solved it. So if you wouldn't mind, would love for you to type sent in the chat when you've sent it.

And yes, that does come for the t-shirt. OK, so. Let's sort of bring this home. We've spoken over the last 40 minutes or so. We've shown you a few demos across various public health functions that includes EPI, communications, admin, grant writing, all running that loop that we talked about, that five-step loop that all of you done, all of you run every day. We've shown you that.

AI can compress some of those steps and save you time to give back to your community while maintaining your judgment, your interpretation, your professional public health brain as part of decision making, because that needs to stay with you. And my question for you is, what could this look like for your department? And to answer that, I'll leave you with three asks.

one for today and two easy ones for maybe tomorrow or later this week. First, I would love for you to book a call with me. You can do so at that QR code or send me an email. You, to that call, bring one workflow that's eating your team's time. It could be one of the ones you put in the polls or something else altogether thatyou were thinking about during the demos. I saw some in the chat.

Would love to would love to work through some of those. So. Bring that workflow and let's run it against PH360 together. Maybe we'll build an agent together and see whether or not PH360 is a good fit for you. So the QR code is on your screen right now. It goes straight to the booking page. It's A 20-minute meeting and Juliana just put, will be putting the link in the chat shortly. Two more follow-ups after that.

If you haven't sent that IT slide yet, please do. Don't worry if you missed it. We'll send it to you in a follow-up email. And it's the same screenshot, so that's the last task. Watch your inbox within 24 hours. The recording, along with a one-page summary, and that IT slide will be in your inbox. And please, we'd love it if you could forward that on to your decision makers at your organization.

Our hope is that you won't have to write a single thing to bring your team up to speed, and you can just hit forward and send. So here's what we believe. Public health professionals should not have to become a prompt engineer and don't need to be an AI safety expert to be able to take advantage of these tools. You should have a platform where you don't need to be either, and that platform is PH360.

At the top of the hour, we said public health must use AI, and that's true, but public health professionals also must stay focused on public health. PH360 is how you can do both. Every hour, every minute that you save on this administrative work using this tool, we hope is that hour or minute that you give back to your community, and that's the whole point. Thank you for listening.

Now, we had a ton of questions in the chat, and we want to hear from you. If you have any more, drop your questions in the chat, and Juliana will be taking the questions that were already sent, and we'll work through them. Thanks, everyone.

Built to pass your security review

How to get started

Q&A

Juliana McMillan-Wilhoit 40:58

Thank you so much, Nebu, and everyone for your engagement and for your questions. So we have gotten some questions in the chat, which I'm monitoring as well as on the Q&A tab. So please, if you have questions, please, please don't hesitate. Yeah, please, please ask them.

So one of the questions that we got was, Jefferson, early on in the presentation, you were talking about how PH360 works, and you talked about how the results are checked both ways, that bidirectional check. And so I'm wondering if you could talk about what that means and why that really matters.

Jefferson McMillan-Wilhoit 41:32

Mhm. Yeah, so bi-directional checks means that not only is the text that I as a professional putting into Maple and Oak being checked going in for things like does it contain PHI and do we need to mask that PH360? Does it contain any biased language? Does it contain anyThings where we would not want the system to respond to, so things that are out of scope for the system.

So it's checking my input, but, and most AI systems will do that, right? So if you go to Claude or Gemini or anywhere else, they're going to check your input. inputs. They're going to make sure that you're not asking it to do something you're not, it's not supposed to, you're going to ask, it's going to make sure that it doesn't have hate speech in it or anything like that.

What most AI systems or none other that I know of do is check the output and that's what bi-directional means. So not only are we putting the, are we running. what the user puts in through all of those guardrails, we are also putting it through the same guardrails coming back from the model. So what we're checking for is, does the model's response show any bias?

Does it, is it cited correctly and does it have enough confidence that we can present it to the user? Is it returning with any protected health information that it may have found somewhere else in the model that we didn't? give it. All of those things are being checked as they come out. And what PH360 will do is in some cases, it will completely block the answer and say, we can't, it can't give that to you.

And it will tell you why. It'll tell you which guardrail it hit. Or in other cases, it'll give you a confidence rating. It'll say, I'm about.30% confident in this answer and that gives you the ability to sort of figure out do you really want to use that information or not. And that's where that big trust component comes in because again, most AI systems are meant to sound incredibly confident.

Ours is meant to actually sometimes sound less confident because we want you to be able to make those decisions and just have the system. support you in getting to those decisions quicker. So that's what it means to be bidirectional.

Juliana McMillan-Wilhoit 43:51

Thank you so much, Jefferson, for that answer. Right. One analogy that we often use is to think about AI, like an intern who knows so much about the world, they have a PhD in everything, yet they don't know you. And so, as he said, right, checking those results. And that's one of the things where wewhere that bidirectional layer is working on adding that. So Jefferson, this is likely another question for you.

So we had a question from Rachel on where does PH360 pull its data from? Are the users inputting the data? Is it coming from Google?

Juliana McMillan-Wilhoit 44:33

Does it only pull from reputable sources?

Jefferson McMillan-Wilhoit 44:37

Yep. Those are all really good questions. So the answer to the is it pulling from Google is no. Actually, PH360 does not go out to the internet at all for anything. And here's why. We can't ensure it's accurate and we can't, and there's really no way, no guardrail we could put in place to check information that's coming in from the internet. from reliable source, from a reliable source.

So PH360 on its own is not enabled to go out to the internet. What it does do is it takes information that we have curated, that's our public health reasoning engine, as well as layers on information that the agency provides or a user provides through an agent. And that's why we have those two layers.

The layer, the reasoning engine has all of that underlying public health knowledge that's applicable across the broad scope of public health. And then agents help to narrow that information down into your local workflows. And so that's, and it's. PH360 is not trained on anything, right? We're using foundational models that have been trained on the wide internet.

What it's doing is it's taking, providing the context needed for those models to perform at their peak, and then using that information to check against to make sure that we're getting it right.

Juliana McMillan-Wilhoit 45:59

Wonderful. And if anyone has any follow-up questions to these questions, questions about questions, please again, you know, feel free to ask them here. So another question is, does PH360 help with summarizing reportable disease data or is thereDoes it play a role in assisting with outreach or contact tracing?

Nebu Kolenchery 46:27

I can take that, so...

Juliana McMillan-Wilhoit 46:28

Great.

Nebu Kolenchery 46:30

Yeah, great question. Thank you for asking it. And yeah, absolutely. And we're super excited to get those tools into the hands of public health professionals because they are MacGyvers and will do great things. So here's what we have seen. You canbuild an agent with your data and get descriptive statistics, if that would be helpful for you. You can also, which includes summarizing reportable disease data.

You can also use it for outreach and contact tracing, where, for example, if Jefferson, instead of saying, what, you know, is this measles, if he had said, can you help me create a

Nebu Kolenchery 47:09

case investigation form or a contact tracing form, it'll generate that for you that you can edit. And so that alone took me hours. For example, during MPOX, that stuff takes forever. Yet another thing that's very helpful is communication. So if you want to use it to direct communications to possible cases or contacts, assuming you use email or even text or phone, you can do that.

We've had people use it to come up with scripts for their first case investigation call to their. to a suspect case. My favorite use case is we were on a call recently where they said, hey, can you use this to train DIS on syphilis case investigation?

So we tried it and it was just like, we, the prompt, you know, we said something along the lines of pretend to be a combative syphilis case and help you, you know, help me be a trainer for case investigation. And it said, absolutely, my name is Marcus. I'm A 34 year old male. And then you kind of just, you know, it like pretends to be on the phone and says like, hi, you know, I'm Marcus. Who is this?

And you say, I'm Nebu from the, you know, so-and-so County Health Department. He said, why are you calling me? And you can kind of go back and forth with it and help you do that.

But really, it's that if you think about that, that five-step loop and how the agent builder works in Oak, as long as you give it sort of some reference documents and good instructions, you can use it really like the sky's the limit and your creativity is the limit. That was a much longer answer than you probably wanted, Sanjay. So, the short one is yes. Back to you, Juliana.

Juliana McMillan-Wilhoit 48:46

Yeah, okay, so we've had a couple of questions about this. And so Nebu, I'm going to point us, sort of punt this over to you, but about the questions about the environment and AI. And so

Brett Emo 48:55

Yep.

Juliana McMillan-Wilhoit 49:04

Just wondering if you want to can answer that.

Nebu Kolenchery 49:07

Yeah, absolutely. The energy, so there were two questions, one on the environmental and one about energy use. And both of these are very thoughtful ones. And you know, we're the right people to ask it. I'm glad you asked it. So here's how we think about it.

Yes, there is an environmental footprint to AI, but that's kind of true of every piece of technology infrastructure that we rely on, including the servers that you all are using to be on this Teams call. That being said, the industry is moving aggressively toward renewable energy and more efficient operations. And that, I think, is a conversation worth following.

But more than anything, what we want to push back on is this sort of binary, that the idea that public health is to choose between using AI and protecting communities. We think that's the opposite. The communities you serve have problems right now.

Like we said, our thesis at the beginning was that public health must use AI for staffing shortages, overwhelm teams, and the community who's expecting faster services, right? And the people who understand health and think about prevention for a living cannot sit on the sidelines, while every other industry, including our customers, adopts AI.

If we do so, we've lost our seat at the table on exactly the issues that we care about. And you know, we're the harm reduction people. And I think harm reduction. involves being in the room. And our everyday lived reality means that these tools can be particularly helpful. So that is our answer to that.

Juliana McMillan-Wilhoit 50:40

Thank you. Thank you so much for those questions. So I have some questions that we often get about this, about this platform and a couple of questions that were emailed to me beforehand. But if anyone has like something in particular that you'd like to see, please schedule that meeting with Nebu orlike ask it in the chat and we'd love to just like play around and show with the platform right now.

So Nebu, just like one question is just like, how is this priced? How does that, how does this work?

Nebu Kolenchery 51:11

Thanks. I'm surprising no one asked that. So we have population-based pricing. We firmly believe that costs for public health technology should be equitable. So what that means is we have counties that have 5,000 people using this, and we have counties that have much more than that. And if you have more people in your county and a larger budget, you pay slightly more. We are priced at 2 levels.

First is just Maple, so that includes Maple and all the agents that you create. And the second includes both Maple and Oak. So the rough sort of calculation should be that supervisors should be able to create agents and your team should be using those agents. So the Oak agent builder, I mean, included in the package costs more.

The pricing ranges from anywhere from $26 per user per month for a smaller county all the way up to a little over $100 a month per user per month per user per month in larger counties and states slightly more than that. We actually have a pricing calculator on our website where you can go and estimate the price for you.

Nebu Kolenchery 52:12

What you'd have to do is input your population and the number of users for each license type that you want, and you can sort of see how much it would cost you for estimate purposes.

Juliana McMillan-Wilhoit 52:22

Thanks so much, Nebu, for that. So I guess another question that I get a lot is, why does it matter that we use something that can have PHI or the TIPA compliant? Like, why is it worth an investment? AndUm, when I can just use Chat GP, Danny, you've already acknowledged that. We're underfunded and we're strapped.

Jefferson McMillan-Wilhoit 52:52

Yes, I guess I can take that one. So I think all of us in public health know that if you put garbage data in, you're going to get garbage data out. And that's true for AI too. The more context we can give the AI, the better off we're going to be. And share there are ways to de-identify data. information to get it into a form that you could put into AI, but you're going to lose all of that context.

So I was able, in that measles case, to be able to put in like the age and the birthday and vaccine history and all of those things. And I was able to really allow the AI to have as much context as possible. to give me the most accurate answer.

If I didn't give it that context, it's probably less than a 50% chance that it would have gotten that right, because I couldn't give it all of the information that I have to be able to make the same level of inferences.

And so having a platform where you can put that in is really important for context, but also just like the mental load that all of us are under in public health all the time is like, if I have to remember which systems I can put PHI in and which systems I can't put PH360 in, I'm more than likely going to screw up and I'm going to put PHI in a system that II'm not supposed to. And I'm not going to do it intentionally.

I'm going to do it because I'm working quickly, I'm under a time crunch, I am super stressed, and I just forget that I'm not supposed to put PHI in the system. And that's why with PH360, we needed to make it where you don't have to think. You put the information in that you have. It is safe and is protected, and it allows you to get to a decision clicker.

Juliana McMillan-Wilhoit 54:44

Awesome. So I guess another question that I've gotten a lot is, is there any part of a health department that you don't see this being applicable towards?

Jefferson McMillan-Wilhoit 55:04

I don't think so. Brett Nebu.

Brett Emo 55:09

I, you know, having been an administrator and what always got me was the HR, consulting HR manual. It's a long document, looking for what you want. I wish I had something like this. I...

Brett Emo 55:23

It would have been extraordinarily helpful in many ways. And in my practice, I can't think, I just can't think of a circumstance where it would have not fit into some aspect of the workflow, whether it's research, communication, or just checking what I had written, like just... I can't ask a member of my staff to dedicate their time to checking my work. They don't have time for it.

And so I can build my confidence too in what I'm saying if I could have AI check it against some standards and policy. So I really struggle to find an area like in my practice where it would not have had some contribution.

Jefferson McMillan-Wilhoit 56:03

Yeah, I mean, I, the first one that, like, every time I get asked this question, like, the first thing that comes to mind is like my facilities team. So when I was the CIO for a local health department before coming to F&T, and I did a lot of work with my facilities team.

But even there, like, there are definitely ways that like work order management andprobably like helping to get through technical manuals on equipment quicker. Like even in facilities, there's a use for this. So I just, I'm with you, Brett. I struggle to really think of an area where you couldn't use something like this.

Nebu Kolenchery 56:42

I don't have a cannot, I do have a should not, and that's probably for clinical decision making. So PH360 does not make clinical decisions for you. So a doctor should do that.

Jefferson McMillan-Wilhoit 56:47

M. That's fair.

Nebu Kolenchery 56:52

That's the only, and what we mean is like if you say I have the fever, a rash, and I haven't eaten for five days, what disease do I have? If that's what you ask, I'll say go to a doctor, which is what everyone should say. So.

Juliana McMillan-Wilhoit 57:08

It also won't give you like a chocolate chip cookie recipe because it knows public health. But yeah, I mean, I just imagine with a new employee, like being able to upload your different internal policies or just like, right, you know, the document that you have that's like, you know, go to these different places to get these different things answered.

And being able, instead of having to have your new employee be stuck and afraid, not afraid to ask you a question, but where you are no longer the thing that's holding them back from like moving forward in that onboarding process. So then you can focus on the heart, the relationship of. getting to know them, getting to understand their goals, getting to understand what really makes them excited.

I think that that is just an incredibly powerful opportunity of something like this across health departments. Well, anyway, thank you all so much for your participation and just everyone's engagement today. We are just really, really thankful forUm... Yeah, for everyone, for everyone's participation.

So you'll be getting, as Nebu said earlier, you'll be getting a follow-up email in a little bit from us that will both say the winner of the t-shirt because Teams is not letting me see sort of all the participants at this point. So for full transparency, that will be included in the email, just that person's first name. and as well as a link to the recording.

But what I just want to end with is you matter and your work matters. AI is not going to replace you. The vision that we have of AI is a tool that can come alongside and assist you in your work and in your workflows and to help you do the work that really matters to you and your communities. So I am just personally

Brett Emo 58:54

The.

Juliana McMillan-Wilhoit 59:06

thankful for you and just thankful for the work and the vision that each and every one of you have. So thank you so much. stopped transcription