WEBVTT
00:00:05.120 --> 00:00:09.679
This is the Beyond Data Management Podcast powered by HS GovTech.
00:00:09.919 --> 00:00:12.560
Now here's your host, Cameron Garrison.
00:00:12.800 --> 00:00:17.199
Welcome to a special presentation of the Beyond Data Management Podcast Channel.
00:00:17.280 --> 00:00:18.399
I'm Cameron Garrison.
00:00:18.559 --> 00:00:31.280
We were recently at AFTO, the Association of Food and Drug Officials conference, and they had a really neat event on the Friday of the conference beginning, and it was the AI in Food Safety workshop.
00:00:31.519 --> 00:00:44.000
There was a lot of panelists from around the country, industry, and regulatory that joined to provide presentations on challenges that they're solving in food safety with AI technology.
00:00:44.240 --> 00:00:51.039
And there was also a panel where you all got to ask questions in the audience and we discussed difficult use cases for AI.
00:00:51.280 --> 00:00:55.439
I was really honored to be asked to be one of the speakers and panelists on that.
00:00:55.840 --> 00:01:03.520
And so what we're going to play for you here is the topic that I chose to present on using technology and AI for food safety.
00:01:03.759 --> 00:01:05.599
It's called closing the blind spot.
00:01:06.319 --> 00:01:34.640
It's a new way to, with AI, bring the inspections that are done every day by the dedicated public servants and health departments and agriculture departments around the country, making that available easily and readily to industry partners and particularly their QA and food safety teams who are also on the same mission that the regulators are, which is good quality and food safety.
00:01:35.280 --> 00:01:37.280
Inspections are really important.
00:01:37.760 --> 00:01:49.920
Some people say they're just a snapshot, and that may be true, but they're important, and that data should be available to be actioned on by big brands and chains that are being regulated.
00:01:50.159 --> 00:02:00.000
So here from the AFTO AI in food safety workshop, my presentation on closing the blind spot between regulators and industry.
00:02:24.319 --> 00:02:27.439
HealthSpace is currently, I just did it.
00:02:37.840 --> 00:02:49.599
Currently it's uh just over a thousand fifty, as the main source of truth collecting all of their data, everything from complaints, inspections, permitting, billing, all of those types of things.
00:02:49.759 --> 00:02:50.719
And this was hard for me.
00:02:50.800 --> 00:02:56.080
I've had up until about a week and a half ago, I had three separate case studies working.
00:02:56.639 --> 00:02:59.599
And one of them actually is one of the workshop topics.
00:02:59.680 --> 00:03:06.080
How do you get use inspection data to do better usage of your resources with limited inspectors?
00:03:06.159 --> 00:03:07.840
You never have as many as you need.
00:03:08.159 --> 00:03:12.000
We've been doing that with the state of Colorado since 2020.
00:03:12.159 --> 00:03:22.479
So then the other one that I really love that I've been involved in analyzing deeply lately is foodborne illness complaint intake and AI agents there.
00:03:22.960 --> 00:03:32.400
But I decided at the end to settle on this one because it's actually one that me and a lot of people, even people in Jorge's position, have been sort of chasing for years.
00:03:32.639 --> 00:03:45.520
And that is the public regulatory data and the difficulty creating a blind spot for industry to get it timely and to be able to utilize it.
00:03:45.759 --> 00:03:48.000
And this has been, again, a long time going.
00:03:48.080 --> 00:03:54.240
Back in 2005, I so I first company in the space uh DHD.
00:03:55.120 --> 00:04:03.520
And in 2005, we worked with them at DHD, and I worked with uh Darden and Yum brands and some of the larger ones.
00:04:03.759 --> 00:04:13.520
DHD at the time had about 720 agencies, and the agencies were very willing to say, yes, you can share the public record aspect under these conditions.
00:04:13.840 --> 00:04:20.000
But it still didn't turn out to really accomplish what needed to be done because we were using ETLs.
00:04:20.079 --> 00:04:32.720
I will try not to go into nerd speak here, uh, but we were using processes that ended up just being expensive, manual, and then the chains or the brands would have to have a lot of people to actually normalize that data.
00:04:32.959 --> 00:04:36.079
So, this is what we'll talk about here today, is closing that blind spot.
00:04:36.160 --> 00:04:41.360
A lot of times there's a mock headline there, health inspectors cite serious violations at popular chains.
00:04:41.759 --> 00:04:54.879
A lot of times, unfortunately, the first time corporate or industry will know at times of a really bad violation, or even if it's not that bad, uh a tantalizing one is the media.
00:04:55.199 --> 00:04:59.120
You get called for comment, or you'll just see it on the news.
00:04:59.519 --> 00:05:01.199
This has been an interesting evolution.
00:05:01.279 --> 00:05:19.040
So the case study problem for this is how can a restaurant chain, and not just chains all of them, but chains that's more acute, receive local health inspection results in real time, interpreting them consistently, even though there's different sheets, different scoring methods, different food codes being implemented.
00:05:19.120 --> 00:05:20.639
You have quite an issue with that there.
00:05:20.720 --> 00:05:23.360
So that's sort of the problem in a nutshell.
00:05:23.600 --> 00:05:25.439
And it's really two sides of the same mission.
00:05:25.600 --> 00:05:29.519
So environmental health departments, there's 3,000 and change.
00:05:29.839 --> 00:05:37.680
And then you've got federal regulators, you've got state departments of agriculture, we work with a lot of them, all of them generating data.
00:05:37.920 --> 00:05:41.600
At a top level, they're generating the exact same data.
00:05:42.240 --> 00:05:46.000
But they're not, if you need to analyze it consistently.
00:05:46.160 --> 00:05:51.759
Uh, millions are spent by chains in pursuit of this data and on other quality assurance programs.
00:05:51.920 --> 00:05:54.800
They spend a lot on internal quality assurance audits.
00:05:55.040 --> 00:05:56.879
But there's really one goal at the end of the day.
00:05:57.040 --> 00:06:08.319
These are not adversarial relationships, these are two partners on they have different roles, but are serving on the same mission, which is you don't want your customers to get sick.
00:06:09.199 --> 00:06:12.399
And that's what the health department does as well.
00:06:13.680 --> 00:06:16.160
So the scale of the problem is pretty, pretty big.
00:06:16.480 --> 00:06:22.399
So it's it's hard to actually do a graphic that to represent just how messy it is.
00:06:22.639 --> 00:06:31.759
But you've got 3,000 or so data silos or databases that are controlled and configured and customized by all the local health agencies.
00:06:32.720 --> 00:06:40.240
And if you're a chain, it's a map of data points, but no one way to grab them and to be able to interact with them.
00:06:40.879 --> 00:06:43.680
And and uh the current state's also so fragmented.
00:06:43.839 --> 00:06:57.439
So you've got you know almost 1,100 of them on one platform like HS Cloud Suite, but then that still leaves you know a couple thousand more, a little bit less, uh, that have a variety of things.
00:06:57.600 --> 00:07:01.600
In-house, believe it or not, there's still a lot of agencies using paper.
00:07:01.920 --> 00:07:03.600
Uh there really is.
00:07:03.839 --> 00:07:09.839
And and of course, that's an even bigger problem because no matter you can't even ingest that data if it's on paper.
00:07:10.079 --> 00:07:12.639
Some people have websites, so almost universally.
00:07:12.720 --> 00:07:17.199
So this is public record data, uh, generally speaking, in every state.
00:07:17.519 --> 00:07:27.680
And the states and local departments have put a huge push over the last 20 years to put that data online for the public, which is great for the consuming public.
00:07:28.079 --> 00:07:31.680
And it's good for the chains, except they have a scale problem.
00:07:32.000 --> 00:07:35.199
If you've got how many facilities did you say you had, Jorge?
00:07:35.279 --> 00:07:35.839
I should remember that.
00:07:37.279 --> 00:07:38.639
See, 7,000 stores.
00:07:38.720 --> 00:07:41.040
That's a lot of inspections rolling in.
00:07:41.360 --> 00:07:45.759
You cannot staff up to read and make analysis of them by by humans.
00:07:45.839 --> 00:07:46.879
It's just not possible.
00:07:47.040 --> 00:07:48.480
And I'll skip a couple there for time.
00:07:48.639 --> 00:07:51.680
The heterogeneity challenge is really what this is.
00:07:51.839 --> 00:07:55.439
Uh, and I used the handwritten form as an example there.
00:07:55.600 --> 00:07:58.160
You have everything from handwritten to more standardized.
00:07:58.240 --> 00:08:07.439
Uh, we do have 14 states that have standardized on HS Cloud Suite, meaning at least in those 14 states, all of the local jurisdictions have a shared data model.
00:08:07.600 --> 00:08:14.959
So there is progress being made, but if we wait for that to happen, all 50 states agree and come to one standard.
00:08:15.120 --> 00:08:21.279
We're all going to be old, gray, dead, and gone by the time that there's something actionable out of that.
00:08:21.680 --> 00:08:40.879
What chains need is a dashboard to their brand standards, which are probably going to be tied to the FDA probably 2022 food code, and at least in many ways, so that they can view 7,000 inspections from 7,000 locations from different agencies in one format and be able to do weighting and things.
00:08:41.039 --> 00:08:43.600
Um, the AI-powered solution is really the only way to do it.
00:08:43.679 --> 00:08:55.600
So when I worked with Darden 21 years ago and Young Brands and some others, but those were the largest, it was really just a down payment on what could be, but the tech just wasn't there.
00:08:55.759 --> 00:08:56.559
Well, it's here now.
00:08:56.639 --> 00:09:01.600
And there's actually some people that are working on the that problem from an AI perspective.
00:09:01.840 --> 00:09:12.879
And now that you've got a huge amount of the country's data consolidating to at least one standard platform that can be shared more easily, uh, the pieces are here and the time is now.
00:09:13.120 --> 00:09:23.039
So old, old, old school, we would use you can ingest that data electronically through a REST API into a dashboard for your chain or your company.
00:09:23.279 --> 00:09:34.559
And all of that regulatory data, it also would allow you then to lay in the internal QA audits and standards that you do, and to be able to have one ecosystem where you can deal with all of that.
00:09:35.440 --> 00:09:41.039
I I'm not even gonna spend much time here because I think it's kind of obvious, and I don't even believe my own number on the slide there.
00:09:41.120 --> 00:09:42.559
I think nine days is generous.
00:09:42.639 --> 00:09:48.159
Um, to have to do of every inspection at every one of your locations within nine days.
00:09:48.320 --> 00:09:53.519
I I actually think that's a really optimistic uh uh analysis of that.
00:09:53.840 --> 00:10:09.360
So it really would be a game changer, though, thinking of partners from public side and the industry side, the ability with AI to bring all of this together, and I will get you a national unified command center if you're on the brand side.
00:10:09.600 --> 00:10:12.879
So you can set your what you what's important to your brand.
00:10:12.960 --> 00:10:19.679
What are the things that if the department marks it, we always want to action it immediately in the following workflow?
00:10:20.240 --> 00:10:28.320
And having that for a thousand and ultimately more agencies will be able to work without human involvement.
00:10:28.480 --> 00:10:46.960
This hard slide's a little hard to see, but to give you an idea from a brand perspective, you can on the left-hand side there, where there are the different, those are actually like a brand standard tied to FDA food code, which is going to reflect back to what inspectors are marking in the field at the regulatory level.
00:10:47.120 --> 00:11:00.720
And you can actually review for each category, thinking again, FDA 22, if you standardize to that, that every one of your locations or regions, the locations are there, and then you can literally have them scored however you want.
00:11:00.879 --> 00:11:07.600
The scoring is something that is a big problem because some agencies, well, North Carolina, where I'm from, they do uh 100.
00:11:08.000 --> 00:11:10.080
Uh they count down from 100.
00:11:10.240 --> 00:11:18.720
And so if you get a 70, what's funny about that is a lot of consumers actually, uh without some education back in the day, thought a B was okay.
00:11:18.879 --> 00:11:25.919
And the logic would be well, if I got a B in school, I didn't get grounded, I didn't have privileges taken, and I I turned out okay.
00:11:26.080 --> 00:11:29.919
Um so and then some agencies do demerits from zero up.
00:11:30.159 --> 00:11:32.159
You can't handle that if you're a chain.
00:11:32.320 --> 00:11:34.159
You need to be able to develop your own weighting.
00:11:34.240 --> 00:11:37.679
If and it's always fun to pick on the roaches and mice because that's what the media does.
00:11:37.840 --> 00:11:45.759
So if we have roaches in our facility, what do we want to weight that as a corporation according to our standards internally?
00:11:45.919 --> 00:11:53.759
And so you can take all the agency data in and put your own consistent weighting on it, which does make it a lot easier to actually action on.
00:11:54.000 --> 00:11:58.639
And, you know, it's instant national visibility for the chains, which they don't have well.
00:11:58.960 --> 00:12:01.039
I did found, I think I failed to mention the top.
00:12:01.200 --> 00:12:02.399
I worked with EcoLab.
00:12:02.720 --> 00:12:04.639
So I've been chasing this for a long time.
00:12:04.879 --> 00:12:11.039
I worked with EcoLab in 07 and 08 to found their HDI, Health Department Intelligence arm out of EcoShore.
00:12:11.200 --> 00:12:12.799
But we ran into the same problems.
00:12:12.960 --> 00:12:19.679
The data couldn't be normalized efficiently or cheaply, and you couldn't just put it on autopilot and have it show up.
00:12:19.840 --> 00:12:21.039
But that's that's here now.
00:12:21.200 --> 00:12:23.440
So chains will get visibility they've ever had.
00:12:23.759 --> 00:12:26.720
You'll it ties into a food safety culture.
00:12:26.879 --> 00:12:27.600
You hear that a lot.
00:12:27.759 --> 00:12:33.519
We've all been to those seminars and workshops, but there is, it's real, and this really helps reinforce that.
00:12:33.759 --> 00:12:36.320
And more importantly, faster corrective action.
00:12:36.399 --> 00:12:40.399
And that's where agencies are in league with industry.
00:12:40.720 --> 00:12:50.320
If they'll start actioning on certain critical items you find because they know about them quickly, that's a win for the regulatory side.
00:12:50.639 --> 00:12:57.360
And of course, for the public, I think the the uh information is uh or the benefits are quite uh self-evident.
00:12:57.679 --> 00:13:14.480
So, sort of in closing, and just in one slide to sort of wrap up what we said, like why up until, even though I kept chasing that whale, we couldn't quite get it where it needed to be, is because ETLs and dumping data to an FTP and then parsing it in, it just doesn't scale.
00:13:14.720 --> 00:13:24.480
But AI, and one of the big things is we struggled this before in the first iteration of this years ago, is if you're doing it based on keyword, it just leaves a lot of gaps.
00:13:24.720 --> 00:13:38.320
But when you have models that understand semantic meaning, they know, okay, it's 16B on this county's checklist, but it's really this under the FDA food code, and we're going to reflect it that way for consistency for the chain.
00:13:38.480 --> 00:13:47.120
Uh, and you don't have to go through hand, uh, you know, unstructured data, PDFs, copies of things, and you don't have to do public information requests to get the data.
00:13:47.200 --> 00:13:49.440
Uh, so it really gives them a pipeline.
00:13:49.600 --> 00:13:51.840
Uh, we I know there'll probably be some questions on this.
00:13:52.000 --> 00:13:57.679
Health departments do have to maintain control of what they're sharing and who it's being shared with.
00:13:57.840 --> 00:13:58.879
That is a given.
00:13:58.960 --> 00:14:02.559
We can't just fling your data out there and into the blind.
00:14:02.720 --> 00:14:06.879
And so it would have to be approved by an agency to join that network and do it.
00:14:07.039 --> 00:14:09.360
And only the change should have visibility to it.
00:14:09.519 --> 00:14:12.399
Anything that you want to communicate to the public should come through your website.
00:14:12.480 --> 00:14:21.440
And most agencies are already doing that, and I don't think you want to water that down because that then distracts from the brand and being out front as your agency.
00:14:21.600 --> 00:14:24.960
Uh, you don't want to have that really kind of be behind a third party.
00:14:25.200 --> 00:14:28.480
And of course, it's auditable and you can do phase rollout.
00:14:28.559 --> 00:14:30.559
Uh, so there's going to be a lot of fun there.
00:14:30.720 --> 00:14:36.639
I know I'm long, and before we go to questions, I took a note on that podcast tool because that's that's really cool.
00:14:36.879 --> 00:14:39.679
Um, I was just plugged to a degree.
00:14:39.759 --> 00:14:47.600
It's a passion project started about a year ago, and uh HSGov Tech was very kind in uh in jumping behind it and helping make it.
00:14:48.559 --> 00:14:52.960
But this is a uh podcast for the regulatory community, particularly EH.
00:14:53.039 --> 00:14:54.399
It's pretty heavy on food safety.
00:14:54.480 --> 00:14:56.720
You'll recognize Steve and his bow tie there.
00:14:56.879 --> 00:15:02.879
Uh he did a uh quite a good rundown of a bunch of things that were going on and a preview of the conference for that audience.
00:15:02.960 --> 00:15:04.639
But commend it to you if you'd like.
00:15:04.720 --> 00:15:05.679
There's been some great guests.
00:15:05.759 --> 00:15:07.440
In fact, this guy right here has been on before.
00:15:07.519 --> 00:15:08.879
We had a long conversation.
00:15:09.039 --> 00:15:19.200
So, really, it's just sort of a water cooler where you can come uh learn about different things other people are doing around the nation and uh and and build a little bit more community, which has always been hard with this.
00:15:19.360 --> 00:15:20.320
Thank you all.
00:15:22.000 --> 00:15:23.759
I hope you enjoyed that presentation.
00:15:23.919 --> 00:15:31.440
There's a lot of work that has been done, a lot of work still to do to make this ultimate dream a reality.
00:15:31.600 --> 00:15:36.720
But if you'd like to know more, I've got my contact information down in the description below.
00:15:36.879 --> 00:15:42.480
Please feel free to reach out to me if you've got any ideas, suggestions, or would like to be involved in some way.
00:15:42.559 --> 00:15:45.440
I'd love to talk with you and appreciate you watching.
00:15:45.679 --> 00:15:50.000
Special thanks also to AFTO again for putting on a really great event.
00:15:50.159 --> 00:15:56.639
There's a lot uh of great learning there for myself included, and I want to thank them for doing that and for inviting me.
00:15:56.720 --> 00:16:00.320
We look forward to doing some follow-ups on it with AFTO again in the future.
00:16:00.480 --> 00:16:02.559
Until next time, have a great day, everybody.
00:16:04.960 --> 00:16:09.039
Quick break here to introduce Wiley from our friends at AFTO.
00:16:09.360 --> 00:16:10.320
Need answers fast?
00:16:10.720 --> 00:16:19.679
Wiley helps you find guidance, training, contact information, subject matter experts, and AFTO resources in seconds powered by AI.
00:16:19.919 --> 00:16:28.480
It can also help draft checklists, SOPs, training plans, action plans, webinar summaries, and even approval of emails for your supervisor.
00:16:28.720 --> 00:16:37.600
Whether you're in the office, in the field, or preparing for your next training, Wiley is there to help you spend less time researching and more time getting things done.
00:16:37.840 --> 00:16:41.200
Learn more at afto.org Ask Wiley.
00:16:41.279 --> 00:16:45.440
It's a brave new world of AI, and Afto's leading the way with Wiley.
00:16:45.600 --> 00:16:50.399
Afto.org askwily or scan the QR code on the screen.
00:00:05.120 --> 00:00:09.679
This is the Beyond Data Management Podcast powered by HS GovTech.
00:00:09.919 --> 00:00:12.560
Now here's your host, Cameron Garrison.
00:00:12.800 --> 00:00:17.199
Welcome to a special presentation of the Beyond Data Management Podcast Channel.
00:00:17.280 --> 00:00:18.399
I'm Cameron Garrison.
00:00:18.559 --> 00:00:31.280
We were recently at AFTO, the Association of Food and Drug Officials conference, and they had a really neat event on the Friday of the conference beginning, and it was the AI in Food Safety workshop.
00:00:31.519 --> 00:00:44.000
There was a lot of panelists from around the country, industry, and regulatory that joined to provide presentations on challenges that they're solving in food safety with AI technology.
00:00:44.240 --> 00:00:51.039
And there was also a panel where you all got to ask questions in the audience and we discussed difficult use cases for AI.
00:00:51.280 --> 00:00:55.439
I was really honored to be asked to be one of the speakers and panelists on that.
00:00:55.840 --> 00:01:03.520
And so what we're going to play for you here is the topic that I chose to present on using technology and AI for food safety.
00:01:03.759 --> 00:01:05.599
It's called closing the blind spot.
00:01:06.319 --> 00:01:34.640
It's a new way to, with AI, bring the inspections that are done every day by the dedicated public servants and health departments and agriculture departments around the country, making that available easily and readily to industry partners and particularly their QA and food safety teams who are also on the same mission that the regulators are, which is good quality and food safety.
00:01:35.280 --> 00:01:37.280
Inspections are really important.
00:01:37.760 --> 00:01:49.920
Some people say they're just a snapshot, and that may be true, but they're important, and that data should be available to be actioned on by big brands and chains that are being regulated.
00:01:50.159 --> 00:02:00.000
So here from the AFTO AI in food safety workshop, my presentation on closing the blind spot between regulators and industry.
00:02:24.319 --> 00:02:27.439
HealthSpace is currently, I just did it.
00:02:37.840 --> 00:02:49.599
Currently it's uh just over a thousand fifty, as the main source of truth collecting all of their data, everything from complaints, inspections, permitting, billing, all of those types of things.
00:02:49.759 --> 00:02:50.719
And this was hard for me.
00:02:50.800 --> 00:02:56.080
I've had up until about a week and a half ago, I had three separate case studies working.
00:02:56.639 --> 00:02:59.599
And one of them actually is one of the workshop topics.
00:02:59.680 --> 00:03:06.080
How do you get use inspection data to do better usage of your resources with limited inspectors?
00:03:06.159 --> 00:03:07.840
You never have as many as you need.
00:03:08.159 --> 00:03:12.000
We've been doing that with the state of Colorado since 2020.
00:03:12.159 --> 00:03:22.479
So then the other one that I really love that I've been involved in analyzing deeply lately is foodborne illness complaint intake and AI agents there.
00:03:22.960 --> 00:03:32.400
But I decided at the end to settle on this one because it's actually one that me and a lot of people, even people in Jorge's position, have been sort of chasing for years.
00:03:32.639 --> 00:03:45.520
And that is the public regulatory data and the difficulty creating a blind spot for industry to get it timely and to be able to utilize it.
00:03:45.759 --> 00:03:48.000
And this has been, again, a long time going.
00:03:48.080 --> 00:03:54.240
Back in 2005, I so I first company in the space uh DHD.
00:03:55.120 --> 00:04:03.520
And in 2005, we worked with them at DHD, and I worked with uh Darden and Yum brands and some of the larger ones.
00:04:03.759 --> 00:04:13.520
DHD at the time had about 720 agencies, and the agencies were very willing to say, yes, you can share the public record aspect under these conditions.
00:04:13.840 --> 00:04:20.000
But it still didn't turn out to really accomplish what needed to be done because we were using ETLs.
00:04:20.079 --> 00:04:32.720
I will try not to go into nerd speak here, uh, but we were using processes that ended up just being expensive, manual, and then the chains or the brands would have to have a lot of people to actually normalize that data.
00:04:32.959 --> 00:04:36.079
So, this is what we'll talk about here today, is closing that blind spot.
00:04:36.160 --> 00:04:41.360
A lot of times there's a mock headline there, health inspectors cite serious violations at popular chains.
00:04:41.759 --> 00:04:54.879
A lot of times, unfortunately, the first time corporate or industry will know at times of a really bad violation, or even if it's not that bad, uh a tantalizing one is the media.
00:04:55.199 --> 00:04:59.120
You get called for comment, or you'll just see it on the news.
00:04:59.519 --> 00:05:01.199
This has been an interesting evolution.
00:05:01.279 --> 00:05:19.040
So the case study problem for this is how can a restaurant chain, and not just chains all of them, but chains that's more acute, receive local health inspection results in real time, interpreting them consistently, even though there's different sheets, different scoring methods, different food codes being implemented.
00:05:19.120 --> 00:05:20.639
You have quite an issue with that there.
00:05:20.720 --> 00:05:23.360
So that's sort of the problem in a nutshell.
00:05:23.600 --> 00:05:25.439
And it's really two sides of the same mission.
00:05:25.600 --> 00:05:29.519
So environmental health departments, there's 3,000 and change.
00:05:29.839 --> 00:05:37.680
And then you've got federal regulators, you've got state departments of agriculture, we work with a lot of them, all of them generating data.
00:05:37.920 --> 00:05:41.600
At a top level, they're generating the exact same data.
00:05:42.240 --> 00:05:46.000
But they're not, if you need to analyze it consistently.
00:05:46.160 --> 00:05:51.759
Uh, millions are spent by chains in pursuit of this data and on other quality assurance programs.
00:05:51.920 --> 00:05:54.800
They spend a lot on internal quality assurance audits.
00:05:55.040 --> 00:05:56.879
But there's really one goal at the end of the day.
00:05:57.040 --> 00:06:08.319
These are not adversarial relationships, these are two partners on they have different roles, but are serving on the same mission, which is you don't want your customers to get sick.
00:06:09.199 --> 00:06:12.399
And that's what the health department does as well.
00:06:13.680 --> 00:06:16.160
So the scale of the problem is pretty, pretty big.
00:06:16.480 --> 00:06:22.399
So it's it's hard to actually do a graphic that to represent just how messy it is.
00:06:22.639 --> 00:06:31.759
But you've got 3,000 or so data silos or databases that are controlled and configured and customized by all the local health agencies.
00:06:32.720 --> 00:06:40.240
And if you're a chain, it's a map of data points, but no one way to grab them and to be able to interact with them.
00:06:40.879 --> 00:06:43.680
And and uh the current state's also so fragmented.
00:06:43.839 --> 00:06:57.439
So you've got you know almost 1,100 of them on one platform like HS Cloud Suite, but then that still leaves you know a couple thousand more, a little bit less, uh, that have a variety of things.
00:06:57.600 --> 00:07:01.600
In-house, believe it or not, there's still a lot of agencies using paper.
00:07:01.920 --> 00:07:03.600
Uh there really is.
00:07:03.839 --> 00:07:09.839
And and of course, that's an even bigger problem because no matter you can't even ingest that data if it's on paper.
00:07:10.079 --> 00:07:12.639
Some people have websites, so almost universally.
00:07:12.720 --> 00:07:17.199
So this is public record data, uh, generally speaking, in every state.
00:07:17.519 --> 00:07:27.680
And the states and local departments have put a huge push over the last 20 years to put that data online for the public, which is great for the consuming public.
00:07:28.079 --> 00:07:31.680
And it's good for the chains, except they have a scale problem.
00:07:32.000 --> 00:07:35.199
If you've got how many facilities did you say you had, Jorge?
00:07:35.279 --> 00:07:35.839
I should remember that.
00:07:37.279 --> 00:07:38.639
See, 7,000 stores.
00:07:38.720 --> 00:07:41.040
That's a lot of inspections rolling in.
00:07:41.360 --> 00:07:45.759
You cannot staff up to read and make analysis of them by by humans.
00:07:45.839 --> 00:07:46.879
It's just not possible.
00:07:47.040 --> 00:07:48.480
And I'll skip a couple there for time.
00:07:48.639 --> 00:07:51.680
The heterogeneity challenge is really what this is.
00:07:51.839 --> 00:07:55.439
Uh, and I used the handwritten form as an example there.
00:07:55.600 --> 00:07:58.160
You have everything from handwritten to more standardized.
00:07:58.240 --> 00:08:07.439
Uh, we do have 14 states that have standardized on HS Cloud Suite, meaning at least in those 14 states, all of the local jurisdictions have a shared data model.
00:08:07.600 --> 00:08:14.959
So there is progress being made, but if we wait for that to happen, all 50 states agree and come to one standard.
00:08:15.120 --> 00:08:21.279
We're all going to be old, gray, dead, and gone by the time that there's something actionable out of that.
00:08:21.680 --> 00:08:40.879
What chains need is a dashboard to their brand standards, which are probably going to be tied to the FDA probably 2022 food code, and at least in many ways, so that they can view 7,000 inspections from 7,000 locations from different agencies in one format and be able to do weighting and things.
00:08:41.039 --> 00:08:43.600
Um, the AI-powered solution is really the only way to do it.
00:08:43.679 --> 00:08:55.600
So when I worked with Darden 21 years ago and Young Brands and some others, but those were the largest, it was really just a down payment on what could be, but the tech just wasn't there.
00:08:55.759 --> 00:08:56.559
Well, it's here now.
00:08:56.639 --> 00:09:01.600
And there's actually some people that are working on the that problem from an AI perspective.
00:09:01.840 --> 00:09:12.879
And now that you've got a huge amount of the country's data consolidating to at least one standard platform that can be shared more easily, uh, the pieces are here and the time is now.
00:09:13.120 --> 00:09:23.039
So old, old, old school, we would use you can ingest that data electronically through a REST API into a dashboard for your chain or your company.
00:09:23.279 --> 00:09:34.559
And all of that regulatory data, it also would allow you then to lay in the internal QA audits and standards that you do, and to be able to have one ecosystem where you can deal with all of that.
00:09:35.440 --> 00:09:41.039
I I'm not even gonna spend much time here because I think it's kind of obvious, and I don't even believe my own number on the slide there.
00:09:41.120 --> 00:09:42.559
I think nine days is generous.
00:09:42.639 --> 00:09:48.159
Um, to have to do of every inspection at every one of your locations within nine days.
00:09:48.320 --> 00:09:53.519
I I actually think that's a really optimistic uh uh analysis of that.
00:09:53.840 --> 00:10:09.360
So it really would be a game changer, though, thinking of partners from public side and the industry side, the ability with AI to bring all of this together, and I will get you a national unified command center if you're on the brand side.
00:10:09.600 --> 00:10:12.879
So you can set your what you what's important to your brand.
00:10:12.960 --> 00:10:19.679
What are the things that if the department marks it, we always want to action it immediately in the following workflow?
00:10:20.240 --> 00:10:28.320
And having that for a thousand and ultimately more agencies will be able to work without human involvement.
00:10:28.480 --> 00:10:46.960
This hard slide's a little hard to see, but to give you an idea from a brand perspective, you can on the left-hand side there, where there are the different, those are actually like a brand standard tied to FDA food code, which is going to reflect back to what inspectors are marking in the field at the regulatory level.
00:10:47.120 --> 00:11:00.720
And you can actually review for each category, thinking again, FDA 22, if you standardize to that, that every one of your locations or regions, the locations are there, and then you can literally have them scored however you want.
00:11:00.879 --> 00:11:07.600
The scoring is something that is a big problem because some agencies, well, North Carolina, where I'm from, they do uh 100.
00:11:08.000 --> 00:11:10.080
Uh they count down from 100.
00:11:10.240 --> 00:11:18.720
And so if you get a 70, what's funny about that is a lot of consumers actually, uh without some education back in the day, thought a B was okay.
00:11:18.879 --> 00:11:25.919
And the logic would be well, if I got a B in school, I didn't get grounded, I didn't have privileges taken, and I I turned out okay.
00:11:26.080 --> 00:11:29.919
Um so and then some agencies do demerits from zero up.
00:11:30.159 --> 00:11:32.159
You can't handle that if you're a chain.
00:11:32.320 --> 00:11:34.159
You need to be able to develop your own weighting.
00:11:34.240 --> 00:11:37.679
If and it's always fun to pick on the roaches and mice because that's what the media does.
00:11:37.840 --> 00:11:45.759
So if we have roaches in our facility, what do we want to weight that as a corporation according to our standards internally?
00:11:45.919 --> 00:11:53.759
And so you can take all the agency data in and put your own consistent weighting on it, which does make it a lot easier to actually action on.
00:11:54.000 --> 00:11:58.639
And, you know, it's instant national visibility for the chains, which they don't have well.
00:11:58.960 --> 00:12:01.039
I did found, I think I failed to mention the top.
00:12:01.200 --> 00:12:02.399
I worked with EcoLab.
00:12:02.720 --> 00:12:04.639
So I've been chasing this for a long time.
00:12:04.879 --> 00:12:11.039
I worked with EcoLab in 07 and 08 to found their HDI, Health Department Intelligence arm out of EcoShore.
00:12:11.200 --> 00:12:12.799
But we ran into the same problems.
00:12:12.960 --> 00:12:19.679
The data couldn't be normalized efficiently or cheaply, and you couldn't just put it on autopilot and have it show up.
00:12:19.840 --> 00:12:21.039
But that's that's here now.
00:12:21.200 --> 00:12:23.440
So chains will get visibility they've ever had.
00:12:23.759 --> 00:12:26.720
You'll it ties into a food safety culture.
00:12:26.879 --> 00:12:27.600
You hear that a lot.
00:12:27.759 --> 00:12:33.519
We've all been to those seminars and workshops, but there is, it's real, and this really helps reinforce that.
00:12:33.759 --> 00:12:36.320
And more importantly, faster corrective action.
00:12:36.399 --> 00:12:40.399
And that's where agencies are in league with industry.
00:12:40.720 --> 00:12:50.320
If they'll start actioning on certain critical items you find because they know about them quickly, that's a win for the regulatory side.
00:12:50.639 --> 00:12:57.360
And of course, for the public, I think the the uh information is uh or the benefits are quite uh self-evident.
00:12:57.679 --> 00:13:14.480
So, sort of in closing, and just in one slide to sort of wrap up what we said, like why up until, even though I kept chasing that whale, we couldn't quite get it where it needed to be, is because ETLs and dumping data to an FTP and then parsing it in, it just doesn't scale.
00:13:14.720 --> 00:13:24.480
But AI, and one of the big things is we struggled this before in the first iteration of this years ago, is if you're doing it based on keyword, it just leaves a lot of gaps.
00:13:24.720 --> 00:13:38.320
But when you have models that understand semantic meaning, they know, okay, it's 16B on this county's checklist, but it's really this under the FDA food code, and we're going to reflect it that way for consistency for the chain.
00:13:38.480 --> 00:13:47.120
Uh, and you don't have to go through hand, uh, you know, unstructured data, PDFs, copies of things, and you don't have to do public information requests to get the data.
00:13:47.200 --> 00:13:49.440
Uh, so it really gives them a pipeline.
00:13:49.600 --> 00:13:51.840
Uh, we I know there'll probably be some questions on this.
00:13:52.000 --> 00:13:57.679
Health departments do have to maintain control of what they're sharing and who it's being shared with.
00:13:57.840 --> 00:13:58.879
That is a given.
00:13:58.960 --> 00:14:02.559
We can't just fling your data out there and into the blind.
00:14:02.720 --> 00:14:06.879
And so it would have to be approved by an agency to join that network and do it.
00:14:07.039 --> 00:14:09.360
And only the change should have visibility to it.
00:14:09.519 --> 00:14:12.399
Anything that you want to communicate to the public should come through your website.
00:14:12.480 --> 00:14:21.440
And most agencies are already doing that, and I don't think you want to water that down because that then distracts from the brand and being out front as your agency.
00:14:21.600 --> 00:14:24.960
Uh, you don't want to have that really kind of be behind a third party.
00:14:25.200 --> 00:14:28.480
And of course, it's auditable and you can do phase rollout.
00:14:28.559 --> 00:14:30.559
Uh, so there's going to be a lot of fun there.
00:14:30.720 --> 00:14:36.639
I know I'm long, and before we go to questions, I took a note on that podcast tool because that's that's really cool.
00:14:36.879 --> 00:14:39.679
Um, I was just plugged to a degree.
00:14:39.759 --> 00:14:47.600
It's a passion project started about a year ago, and uh HSGov Tech was very kind in uh in jumping behind it and helping make it.
00:14:48.559 --> 00:14:52.960
But this is a uh podcast for the regulatory community, particularly EH.
00:14:53.039 --> 00:14:54.399
It's pretty heavy on food safety.
00:14:54.480 --> 00:14:56.720
You'll recognize Steve and his bow tie there.
00:14:56.879 --> 00:15:02.879
Uh he did a uh quite a good rundown of a bunch of things that were going on and a preview of the conference for that audience.
00:15:02.960 --> 00:15:04.639
But commend it to you if you'd like.
00:15:04.720 --> 00:15:05.679
There's been some great guests.
00:15:05.759 --> 00:15:07.440
In fact, this guy right here has been on before.
00:15:07.519 --> 00:15:08.879
We had a long conversation.
00:15:09.039 --> 00:15:19.200
So, really, it's just sort of a water cooler where you can come uh learn about different things other people are doing around the nation and uh and and build a little bit more community, which has always been hard with this.
00:15:19.360 --> 00:15:20.320
Thank you all.
00:15:22.000 --> 00:15:23.759
I hope you enjoyed that presentation.
00:15:23.919 --> 00:15:31.440
There's a lot of work that has been done, a lot of work still to do to make this ultimate dream a reality.
00:15:31.600 --> 00:15:36.720
But if you'd like to know more, I've got my contact information down in the description below.
00:15:36.879 --> 00:15:42.480
Please feel free to reach out to me if you've got any ideas, suggestions, or would like to be involved in some way.
00:15:42.559 --> 00:15:45.440
I'd love to talk with you and appreciate you watching.
00:15:45.679 --> 00:15:50.000
Special thanks also to AFTO again for putting on a really great event.
00:15:50.159 --> 00:15:56.639
There's a lot uh of great learning there for myself included, and I want to thank them for doing that and for inviting me.
00:15:56.720 --> 00:16:00.320
We look forward to doing some follow-ups on it with AFTO again in the future.
00:16:00.480 --> 00:16:02.559
Until next time, have a great day, everybody.
00:16:04.960 --> 00:16:09.039
Quick break here to introduce Wiley from our friends at AFTO.
00:16:09.360 --> 00:16:10.320
Need answers fast?
00:16:10.720 --> 00:16:19.679
Wiley helps you find guidance, training, contact information, subject matter experts, and AFTO resources in seconds powered by AI.
00:16:19.919 --> 00:16:28.480
It can also help draft checklists, SOPs, training plans, action plans, webinar summaries, and even approval of emails for your supervisor.
00:16:28.720 --> 00:16:37.600
Whether you're in the office, in the field, or preparing for your next training, Wiley is there to help you spend less time researching and more time getting things done.
00:16:37.840 --> 00:16:41.200
Learn more at afto.org Ask Wiley.
00:16:41.279 --> 00:16:45.440
It's a brave new world of AI, and Afto's leading the way with Wiley.
00:16:45.600 --> 00:16:50.399
Afto.org askwily or scan the QR code on the screen.
