Webinar Recap · with NACCHO August 18, 2026
PH360™ AI your health department can actually use
Watch the recording
The short version
- AI built only for public health. It complies with the HIPAA Privacy and Security Rules, so real case data, PHI included, belongs in it.
- You upload your own documents. State code, local ordinances, SOPs. It answers from those first and cites the section it used.
- It tells you when it does not know. No guessing, and no answer without a source behind it.
Read the transcript
Welcome
Jenny Bengston, Manager of Affiliate Business Partnerships, NACCHO
Good afternoon, everyone, and welcome. Today’s session will introduce you to PH360 and explore how AI can support local public health professionals in their day-to-day work.
Nebu Kolenchery, Chief Revenue Officer, F&T Labs
Flourish and Thrive Labs is a consulting company that works with state and local health departments, and we pride ourselves on being for public health, by public health. What that means is that all of our folks are former public health officials just like you, and come into any engagement understanding your work. Immediately prior to this, I was the communicable disease director at a county health department.
Jefferson McMillan-Wilhoit, Chief Executive Officer, F&T Labs
I’ve been in public health for 16 years — every time I have to say this, it shocks me. My background is as an epidemiologist and biostatistician by training, and immediately prior to this I was the chief information officer for a local health department in Illinois.
Public health must use AI
Nebu Kolenchery
No ifs, ands, or buts — our perspective is that public health must use AI. It’s an imperative, and here’s why. Public health today is being asked to do more than it ever has, with fewer people and less money than we’ve had in decades. The math doesn’t math. AI is one of the only tools we’ve seen that can help.
For years, the fixes we actually needed — more funding, more staff, modern systems — have been promised, but they’ve never really arrived. AI could be the first tool that lets all of us move the needle together without waiting on an external miracle. So the question isn’t whether. It’s how, and what kind.
If you’ve been told to go look into AI and haven’t found a path you’d actually trust, that’s who this webinar is for. I hope you walk out of here with one concrete step for you and your agency — whether or not you ever become a PH360 customer.
What an AI tool needs before public health should touch it
Nebu Kolenchery
From our perspective it’s three things. First, it should be purpose-built for public health, so you don’t have to vet as much of the information you’re getting. Second, guardrails that run both ways — every answer checked from you to the tool and back at you, so every answer can carry a cited source and can tell you when it doesn’t know. Third, consistency: the same answer every time, tied to those sources, so you can trust what you get back.
You are the public health professional. We often tell people: use AI for the things that don’t require your head or your heart. That’s you — your relationships, your brain. That stays with you. On the other hand, AI can be a really great intern that you supervise. That’s how we think about AI use in public health, and that’s the basis for how we built PH360.
Every public health workflow is the same five steps
Nebu Kolenchery
Think about every single workflow you do — a workflow being the series of steps you use to execute a task. Say you’re dealing with a cyclospora case. A doctor reports a patient: that’s intake. You triage and sort it: that’s process. You may consult an external guide — your state’s communicable disease manual, your department’s SOP: that’s reference. You produce something, a response to the provider or a case surveillance form: that’s output. And then you send it back out: communicate.
Every public health workflow runs on some version of those five steps. What we’re asking is that you think about using AI inside that constraint — to make each step faster, or to help with the handoff between steps.
Maple and Oak
Nebu Kolenchery
All of our modules are named after trees, because we like trees. I wish I had a better answer for you.
Maple works like any chat-based AI tool you might use today — you put in a question or a task and it responds. But straight out of the box it comes with those public health guardrails: cited sources you can trust, consistent answers every time. Oak is a no-code agent builder that helps you execute that five-step process. You identify a workflow, give it instructions, and have it draw from the specific bodies of knowledge you reference daily, so those steps get faster.
What’s next
Nebu Kolenchery
Once you have that foundation — Maple, Oak, and Birch for administering users — you can build apps on top of it. We’re starting with Willow and Spruce. Willow is clinical case management: public health nursing, vaccine visits, billing, so your department can do what your community wants and recoup the revenue without spending an arm and a leg on a full-blown EHR. Spruce takes that five-step process into environmental health inspections — restaurants, sewage, wells — AI-native at the baseline so inspectors are better supported in the field.
Demo — starting with what it will not do
Jefferson McMillan-Wilhoit
Much like we like trees, we also like the Wizard of Oz — so we’re going to pretend we’re at the Emerald City Health Department today. It’s always fun to have a somewhat fictitious health department to work out of, given that we’ve all worked in actual ones for a while. I’m going to wear a couple of different hats and tell you who I am at each point.
Jefferson McMillan-Wilhoit
But I always like to start with: what can this thing not do? So I’m going to ask Maple to plan a retirement party for our finance director, with a funny toast. And what you’ll notice is that Maple says that falls outside its abilities — because we have locked PH360 down to public health only.
Every other general-purpose AI assistant on the market today would instantly and cheerfully respond to that. Ours doesn’t, and there’s a reason. Part of our challenge when we think about using AI in public health is making sure there are good AI practices. To expect everyone in public health to become a good prompt engineer — to learn what to ask it and what not to ask it — is a lot to ask of people who are already doing three jobs. This is supposed to make things easier.
Demo — the public health reasoning engine
Jefferson McMillan-Wilhoit
… what the requirements are for a temporary food establishment, what notification obligations might look like, and what we should be capturing right now for the after-action review that we may not be able to reconstruct. You’ll notice that as Maple builds, it is showing you what it is building — and once everything turns green, that’s how you know it’s finalized.
Jefferson McMillan-Wilhoit
Our public health reasoning engine is built off of national standards — what’s published from CDC, what’s published from FDA, model practices. Those all feed into the reasoning engine. I like to think about it as your generalized public health friend. It doesn’t know anything about your local establishment. It doesn’t know anything about your agency and its policies. All it really knows is national best practice.
If I had to build this myself, it probably would have taken me at least an hour, maybe an hour and a half. It gives you the epi workflow and immediate priorities, cohort and case-control design, laboratory information, then the environmental health workflow — vendor inspection priorities, temperature documentation, cold-holding requirements. And it is very clear that this reflects the FDA Food Code framework and not what your state adopted. This is the national version.
Jefferson McMillan-Wilhoit
It is not claiming anything it doesn’t know. That’s what’s important about the reasoning engine — everything is grounded. We’ve built this at multiple levels to reduce and mostly prevent hallucination. It’s always going to tell you when it doesn’t know something, and at what level it knows something. Down at the bottom it will say when an answer is based on the model’s general knowledge rather than on grounded sources.
Demo — drafting a situation report
Jefferson McMillan-Wilhoit
Maple can draft Word documents, which is a really common thing for us in public health. I’m going to ask it to draft an incident situation report for the outbreak. I gave it a date, an incident number, six sections, and some facts I want it to use.
What’s unique about the way Maple works: you want to be able to work with the document alongside the AI. It’s not that you want the AI to produce the document, then download it, then edit it, then go back to the AI. We want you to use AI inside your workflow so you’re never leaving it. So Maple generates the first draft, and then you open it and review it. We don’t allow straight downloading of Word documents without somebody reviewing it — that’s the design decision.
Demo — your own documents, cited
Jefferson McMillan-Wilhoit
I’m going to drag and drop a few documents in — extreme heat guidance memoranda, my municipal code, and my emergency operations plan for extreme heat response. I’m going to ask it to answer specific questions using only those three attached documents, and to cite them. These are large PDFs, and yours probably are too. It’s breaking each one down into its individual parts so it can grab and cite those answers.
Jefferson McMillan-Wilhoit
Here are the approved cold-holding methods, and it’s citing exactly where it got them — municipal code 12.4.0 — and it does that for each one. If we had time I’d open those documents and show you that is actually the code. You could cross-reference and cross-check each one, and it’s going to be accurate, because PH360 has those cross-checks built in.
Demo — building an agent in Oak
Jefferson McMillan-Wilhoit
That’s Maple without any localization at all. But we all know public health is inherently local. So how do we help local health departments add that local context to the reasoning engine? Through our app called Oak.
Agents are just containerized ways of working with AI. Normally, if you wanted to build one, you’d need Python or all kinds of other things. We guide you through it without needing any code at all. It’s a six-step process. We give it a title and a description. Then we tell it what the agent should do — here, that it’s an environmental health and communicable disease advisor, and that it should answer from the knowledge base.
Jefferson McMillan-Wilhoit
More importantly, we need to tell it what to do if it doesn’t know something — and that’s probably the most important part. Say so plainly, never guess, name a document, and if you don’t know any of it, send the user to the EH director for code, the epi manager for cases, or the health officer for legal questions.
Jefferson McMillan-Wilhoit
Then you upload your documents. I can also upload Excel files — and if you give it an Excel file, it can actually do math on it and run statistics. AI inherently does not do math well, but we’ve built an engine to help it, and it’s repeatable and reliable. So I’ve uploaded a salmonella line list and an inspection report.
Jefferson McMillan-Wilhoit
The next thing you have to do is test it. We require it. I always ask what knowledge is in your knowledge base — that’s just good AI practice. We want to make sure the instructions and the knowledge are correct for what we want to use. The last part is to activate it, which means we can share it — just with ourselves, with specific people, or with the entire organization.
Demo — running the agent from Maple
Jefferson McMillan-Wilhoit
Once the agent is built in Oak, we can access it in Maple — that’s the cool part. Let’s go to the PHAB Accreditation Assistant, which comes out of the box. For anyone going through PHAB accreditation, you’ll be able to see this is correct. We’ll ask what measures are needed in Domain One for an initial accreditation. It goes and looks at the standards and measures for 2026 plus a whole bunch of other documents.
Jefferson McMillan-Wilhoit
That’s more than a Control-F on a PDF. It’s taking all those pieces of information and putting them together into a coherent answer I can use to make a decision. Notice it says it’s verifying its sources. Here are all my measures — and as importantly, it cites the source and gives an evidence grade. It says how confident it is in the evidence it’s providing.
Where this is going
Jefferson McMillan-Wilhoit
That’s a whole bunch of what Maple and Oak can do. And to Nebu’s point — once we got that down, once we could do secure, reliable AI, we can now start doing really cool things like building Willow. A light electronic health record that we all wish we had in public health, that doesn’t require all the things an ambulatory or hospital EHR would have. And it all runs powered by that AI interface, the source engine, the public health reasoning engine, and the agents.
Nebu Kolenchery
If folks would like to try this for free, you can scan this QR code or go to the website and sign up for a free two-week trial. I would love to get you all on the platform and get your thoughts on it.
Q&A
Nebu Kolenchery
Can it accept state-level regulations as an upload? Yes — you can either upload it as part of Maple or create an agent with it. Both were shown in the demo.
Jefferson McMillan-Wilhoit
Does it pull from the scientific literature? Today the public health reasoning engine does have some authoritative scientific literature, like MMWR. It’s all curated by those of us at F&T — it goes through an editorial process. It’s not going to access PubMed or everything in the literature, but where we’ve felt it’s important to have certain pieces, it has them.
Jefferson McMillan-Wilhoit
Can it highlight what changed in a document? You can ask it what changed and it’ll tell you, or you can see it on the preview — but it doesn’t highlight on the preview itself. Does it cite the tool in the report? In exported Word documents there’s a footer that says “co-created with PH360 AI assistance.” Whether you leave it in is up to you, but it comes with it.
Jefferson McMillan-Wilhoit
Word only, or PowerPoint too? Currently it’s live with Word documents. PowerPoint is on the roadmap. It can generate HTML code, so if you’re thinking about graphics to go in a slide, it can generate that.
Jefferson McMillan-Wilhoit
Can we build an agent that includes our policies and procedures, even if some content isn’t exclusively public health? Yes, but it will only answer public health related questions. Even if you create an agent, all questions and answers go through the reasoning engine — which has all of those guardrails in it.
Jefferson McMillan-Wilhoit
What’s the difference between uploading documents to Maple and building an agent? In Maple, uploading documents has a limit of 10. Agents take an unlimited number of documents — and the agent keeps memory. It keeps track of what it answered previously so it doesn’t contradict answers it gave before.
Jefferson McMillan-Wilhoit
Is PH360 an ONC-certified solution for HL7 transmission? It is not ONC certified, because that would mean essentially being an ambulatory EHR. However, everything underlying it is HL7 on FHIR — all of its APIs are FHIR-exposed. I’d love to have a conversation about why that decision was made and what it does have that allows for HL7 transmission.
Jefferson McMillan-Wilhoit
When people add their own documents, does that get incorporated into the overall knowledge base? It does not. Every health department gets its own little PH360 house to live in. None of your information goes anywhere else, and no one else’s information comes into you. If you upload a document, and it’s not an agent you shared, no one else can access it. That’s part of our security infrastructure.
Jefferson McMillan-Wilhoit
How big is the context window? Is the back end using RAG? Document upload is 25 megabytes per document, and there’s no limit on the number. We’re running frontier commercial models inside our secured environment, so their context windows apply. But it’s not just RAG. We’re using full vector searching, so before going to the model it does a vector search and pulls only the context that’s needed. It’s not overloading the context window.
Jefferson McMillan-Wilhoit
Can it hold PHI? Yes — that line list has PHI in it. That’s one of the big parts about PH360: you can put PHI in it, you can put PII in it, it is fully secured. We are audited to SOC 2 standards, and we support everyone’s HIPAA compliance. We sign a BAA with our clients. That’s also how we can support things like Willow. Your staff doesn’t have to think about whether they can put something into an AI — the platform is secured, so you can put pretty much anything into it.
Jefferson McMillan-Wilhoit
Does our PH360 live on your servers or in my department? We use AWS as our hosting partner, inside our virtual private cloud. Every tenant lives inside that VPC — but your database is your own database with your own key, and everything passing between that database and the agents is its own thing. Every agent you build has its own runtime in our fabric, so you’re not sharing it with anyone. And your data is your data — when you’re done, we don’t keep it. It gets deleted on a retention schedule, because we have to maintain HIPAA compliance we keep certain things for six to seven years, but most everything gets deleted.
Jefferson McMillan-Wilhoit
Who owns the information uploaded, and the output generated? The department. You own it. We don’t own it. Even as an admin, I can’t access your database. That’s just a thing we keep security on.
Jefferson McMillan-Wilhoit
Could we build agents for community partners, or would they need a subscription? Today they would need a license, and license types are on our website. But it’s come up a lot recently, so we are investigating pushing that out to community partners. A lot of what we’re building — Willow, and Spruce for environmental health — already has that functionality, because they have community portals. Making that available through Maple is definitely something we’re considering.
Jefferson McMillan-Wilhoit
Can local documents override national guidelines? Yes, within the context of Oak. The reasoning engine looks for a local requirement first — something localized to you through an agent or an uploaded document. If it finds it, it uses it. If it doesn’t, it goes out to the national corpus, and if it finds something there it cites and uses that. If it doesn’t have something in either one, it will say it doesn’t know and stop. Think about it as concentric circles the reasoning engine works from.
Nebu Kolenchery
After the free trial, what does it cost annually, and is there a limit on users? There is not a limited number of users. Our pricing is on the website — the same page you go to for the free trial. Scroll down and you’ll see we have population-based pricing, which means it’s priced to be accessible to smaller health departments, and you can get a reasonable estimate there.
On August 18, 2026, Jefferson McMillan-Wilhoit and Nebu Kolenchery spent forty minutes with NACCHO, where F&T Labs is a Diamond Affiliate Business Partner, showing PH360™ to local public health staff.
Transparent Pricing is published on our website.
What is PH360?
PH360 is a platform built for the work health departments actually do: case investigation, inspections, accreditation, preparedness, communications. Underneath every answer sits a public health reasoning engine. That engine is what separates it from a general chat tool with a public health label stuck on the front.
Our team curates that engine by hand from CDC guidance, the FDA Food Code, MMWR, and model practices. Nothing is scraped. It also works in layers: your local rules come first, national guidance second, and if it has neither, it tells you so and stops.
Maple Chat
The chat side of PH360. Ask a question in plain language and the answer cites its source down to the section and document name, with an evidence grade attached. Three assistants are ready out of the box: case investigation and surveillance, PHAB accreditation, and the FDA 2022 Food Code. It drafts Word documents you edit in place, and nothing downloads before a human reviews it.
Above: measles case response cited to Wisconsin Administrative Code.
Oak Build
The builder side of PH360. Build assistants for the jobs your team actually does, whether that is accreditation, food code, communicable disease, or comms, and they appear in Maple for everyone. Documents are unlimited, including spreadsheets it can run real statistics on. You tell it who to refer people to when it does not know, test it, then share it department wide.
Above: the six-step builder. Name it, write instructions, add knowledge, test, activate.
Three programs, preloaded
- Accreditation
PHAB Standards & Measures - Communicable Disease
CDC guidance and case definitions - FDA Food Code
the 2022 edition, in full
PH360 knows public health. You give it your jurisdiction.
Every PH360 starts the same way. By day one your team is working with the FDA Food Code, CDC communicable disease guidance, and the PHAB Standards & Measures. Those are identical in every jurisdiction, which is why the first useful answers come the week you start rather than a quarter later.
Then you make it yours. Your state code and administrative rules. Local ordinances and variance decisions. The SOPs, standing orders, and outbreak playbooks that only exist inside your department. Your own administrators add all of it without help from a developer, and PH360 then reasons across the federal baseline and your local context together.
A general purpose tool can tell you what the FDA Food Code says. It cannot tell you what your county adopted, where your board granted a variance, or who signs off on it. That is the gap this closes.
FDA Food Code · CDC communicable disease guidance · PHAB Standards & Measures
- State code & administrative rules
- Local ordinances & variance decisions
- SOPs, standing orders & playbooks
When someone leaves, what they built stays.
The knowledge your team layers in lives in the platform rather than in one person’s head or one person’s spreadsheet. Any administrator can see and edit any agent. When the sanitarian who built your inspection workflow retires, the workflow stays behind, and so does everything they taught it.
AI only works if you can trust it
Nebu argued in the session that three things have to be true before public health should trust an AI tool. Most platforms manage one of them. We built PH360 to hold all three at once.
Good tools
Purpose-built, not adapted
We chose and tested the tool for public health so that you are not the one carrying that risk.
Good guardrails
Bidirectional, by design
Every answer is checked in both directions, cites its source, and tells you when it is unsure, including when it is working from general knowledge rather than a grounded source.
Good knowledge bases
Same answer every time
Federal frameworks arrive preloaded and your local rules layer on top, so the same question returns the same answer every time.
You stay the public health professional. PH360 is the tool you supervise. It never declares a case, never acts without you, and shows the evidence as you work.
It fits the work you already do
Every public health workflow runs the same five-step cycle, whether it sits in epidemiology, environmental health, immunizations, preparedness, or communications. PH360 is built around that cycle rather than around one program area, which is why the same platform serves your whole department.
01
Intake
Something comes in. A report. A call. A complaint.
02
Process
Sort it. Triage it. Understand what’s actually being asked.
03
Reference
Match it against the knowledge base: codes, SOPs, expertise.
04
Output
Draft. Report. Decision. The thing that gets handed off.
05
Communicate
Close the loop. Back to the public. Back to the case file.
Results from the field
A Wisconsin county health department measured an 83% workload reduction across two AI use cases and reached full team adoption in 30 days. Adoption held because the tool sat inside workflows the team already ran, instead of asking anyone to learn a new system first.
Where PH360 is going
Maple and Oak are what you use today, and they are also the foundation for what comes next. Jefferson put it this way in the session: once we could do secure, reliable AI, we could start building the things public health has actually been missing.
Two purpose-built apps, three shared engines, one secure platform.
Willow
Care & case management
Clinical care, billing, and registry reporting. The light EHR public health has always wanted, without everything a hospital system drags along.
Spruce
Environmental health
Inspections and environmental health programs, including the community-facing portals partners need.
Willow and Spruce run on these engines
Maple
Reasoning app and engine
Drafting, policy answers, and analysis, all grounded, cited, and graded for confidence.
Oak
Agent app and engine
Plain-English updates that act across connected systems. No code, unlimited documents, persistent memory.
Birch
Admin app and engine
Users, roles, audit trails, and workflow configuration.
Because everything shares those engines, the guardrails hold everywhere. A question asked in Maple, in an Oak agent, or eventually in Willow runs through the same reasoning engine and comes back cited and graded the same way. What you buy today is the floor for everything on this map.
Point it at your own codes and manuals, free, for two weeks.
Start your free trialBuilt to pass your security review
Most AI pilots in government stall because the security review starts after everyone is already excited. Start it earlier instead. Forward this section to whoever owns that review at your agency. Everything below is verifiable at our trust center, and your IT team can contact us directly.
- SOC 2 Type II audited · complies with the HIPAA Privacy and Security Rules · we sign a BAA
- Tenant-isolated on AWS · US data residency · your own database and your own key
- Your data is yours. You own the inputs and the outputs, and it never trains anyone else’s model
- Full audit logging for FOIA and sunshine-law requests · source citations on every output
If your department is still writing its AI policy, our free AI policy template is a head start.
Questions from the session
What is PH360, and where is it heading?
PH360 is a secure AI platform built specifically for governmental public health. Today it gives you Maple, a chat interface backed by a public health reasoning engine, and Oak, a no-code agent builder for your own documents. Those are two of three shared engines, with Birch handling users, roles, and audit. Together they are the foundation for purpose-built apps: Willow for care and case management, and Spruce for environmental health. Learn more on the PH360 platform page.
Can we upload our state regulations, local codes, and SOPs?
Yes, two ways. You can attach documents directly in Maple (up to 10 at a time, 25 MB each), or you can build an Oak agent, which takes an unlimited number of documents and keeps memory so it doesn’t contradict answers it gave before. Either way it cites the specific document and section it drew from.
Can our local rules override national guidance?
Yes, and that is the design. The reasoning engine works in concentric circles. It looks for a local requirement first, from an agent or an uploaded document. If it finds one, it uses and cites it. If not, it goes to the national corpus. If it has nothing in either, it says it doesn’t know and stops.
Can PH360 hold PHI and PII?
Yes. The salmonella line list used in the live demo contained PHI. PH360 is SOC 2 Type II audited and complies with the HIPAA Privacy and Security Rules, and we sign a BAA with every client. The point is that your staff shouldn’t have to stop and wonder whether something is safe to put in.
Who owns our data and our outputs?
You do. The department owns both the inputs and the outputs. Your documents never join a shared knowledge base, and nothing you upload trains anyone else’s model. Even F&T Labs admins cannot access your database.
Where does it run, and what happens to our data when we leave?
PH360 is hosted on AWS inside our virtual private cloud, with US data residency. Each tenant has its own database with its own key, and every agent you build gets its own runtime. When you’re done, your data is deleted on a retention schedule. Certain records are held six to seven years to maintain HIPAA compliance, and most everything else is removed.
Is it RAG? What’s the context window?
It’s not a plain RAG pipeline. PH360 runs a full vector search before the model call, so only the context actually needed gets passed rather than overloading the window. Documents upload at up to 25 MB each with no limit on count. The underlying models are frontier commercial models running inside our secured environment, so their context windows apply.
Is PH360 ONC certified for HL7 transmission?
No. ONC certification would require PH360 to be an ambulatory EHR, which is deliberately not what it is. But everything underneath is HL7 on FHIR, and all APIs are FHIR-exposed. If this matters for your program, book time with us and we’ll walk through the architecture.
Can community partners use agents we build?
Today they would need their own license. It’s a request we’ve heard a lot, and it’s under active investigation. Willow and Spruce already include community portals, so extending that through Maple is on the table.
Does it draft documents, and does it disclose AI assistance?
Maple drafts Word documents you review and edit in place, so you never have to download, edit elsewhere, and come back. Nothing downloads without a human reviewing it first. Every exported document carries a footer reading “co-created with PH360 AI assistance,” which you can keep or remove. PowerPoint export is on the roadmap.
Do I need to be a prompt engineer?
No. Maple is plain-language chat, and you do not have to explain that you work in public health. Oak walks you through six steps, including the most important one: telling the agent what to do when it doesn’t know, and who to send the person to instead. If you can drag and drop a file, you can build an agent.
How does pricing work, and is there a user limit?
There is no limit on users. Pricing is population-based, so it stays accessible to smaller health departments, and it is published openly. see the pricing section on the PH360 page and estimate your own jurisdiction.
How do I try it?
Start a free two-week trial. Bring the workflow that’s eating your team’s time and point it at your own documents. That is the fastest way to know whether it fits. If you hit any trouble creating an account, email information@fandtlabs.com and we’ll sort it out.
Try it on the workflow eating your team’s time
Two weeks free, unlimited users, and we do not ask for a card. Upload your own code, your own manual, your own line list, then check whether the answers hold up against the documents you already trust. That is a faster test than any demo we could give you.
Pricing is population-based and published openly. Estimate your jurisdiction on the PH360 page. Questions? information@fandtlabs.com
Ready to point it at your own documents?
Start your free trial