Hello and welcome to get flushed. I'm Pete. Today's episode is all about artificial intelligence, or AI. I'll be joined today by Vladislav ostremski from Airvote. We'll hear from Vlad in a moment, but you may remember Airvote from the episode I made with Angelique, one of the founders of Airvote. I also made an episode with Roy Behring from texan site Services, who described the efficiencies gained by his team when they deployed Airvote in their restrooms at the Valero, Texas open.
I'll put links to both those episodes in the notes for the show. For newer listeners or those not familiar with Airvote, it's a very simple but effective system. The operator places an airvote sticker with coloured QR codes in each restroom cabin. When toilet users scan the sticker with the camera on their phone, the system launches a web based interface that allows the user to leave feedback. That feedback can be sent directly to the restroom operator on regular routes.
It allows users to let the operator know if something needs attention. And at events, the system allows the operator to find out about and respond to any problems in real time. A couple of weeks ago, Angelic from Airvote got in touch to let me know that the company is doing some exciting work with AI.
At that point, I really didn't know much about artificial intelligence other than what I've seen on tv or in social media, so I spent the last couple of weeks trying to learn more and actually try some AI for myself. It turns out that it can do some pretty incredible things if you follow my social channels. You may have seen the photograph of my new bright yellow restroom inspired podcast studio.
You may have heard the trailer I shared where last week's episode about tenders and bids was discussed enthusiastically by two reviewers. And you may have seen the transcript I've been adding to episodes over the last few weeks. All of those things were produced by AI, and it took no more effort from me than a couple of clicks of my mouse. Honestly, I don't know whether to be amazed, relieved, or scared.
To help us understand how AI can benefit restroom operators, I spoke with Vlad, one of the joint founders and the head of research and development at Airvote. Hello. Hey Vlad, how are you? Great, and you? I'm good, thanks. Welcome to get flushed. It's great to have you on the show. Nice to meet you. I'm guessing you're in the Ukraine. Are you looking at your Facebook profile? Yes, yes, I'm in Ukraine. In Odessa actually in the southern part of Ukraine. All good, not too impacted with the war.
I would say 50 50 because I'm originally from Mykolaia. It's a bit eastern, like 150 km eastern from Odessa and there was war there. But everything is okay now and it's pretty quiet there right now. Can we just cover your role at Airvote, Vlad? How long have you been there and what exactly do you do in your day to day job for Airvote? I'm actually co founder and head of R and D at Airvault and we started Ayervote with Dmitry and Angelic.
In the beginning, I was mobile developer and I developed applications on iOS and Android, and Dmitry found me on one freelance platform. We met and talked about the idea that Dmitry had to create that feedback system. The idea was to put QR codes everywhere to get as much feedback as possible. And he thought that the mobile application is needed for that. And we talked maybe one or 2 hours about that, discussing it from different angles.
And my understanding was that you don't need mobile application for that. You cannot expect end user to have Airvote application installed. To use airvote application to scan an Airvote QR code. That's why our end users just use their phone camera to scan a QR code. Then they get website opened and they leave a feedback, write a comment or add a picture from there. So they don't need application, just use a phone camera and default to a browser on their phone.
I think that's a really elegant solution, Vlad. I think really elegant. And as you say, you can't expect people to download an app before they give you the feedback. So you've circumvented that and made it very, very easy to use. That was the point that I was a mobile developer developing mobile applications for iOS and Android. And I told Mitri, I'm a mobile developer, but you don't need a mobile developer for this system. Well, it sounds like you talked yourself out of a job there, Vlad.
That sounds like that, yes. That was funny, because when we talked to each other, we first felt like we like to work together. And we started from that point, but now we have a team, we work on envote, we develop new features, we get new customers, and we are pretty happy with what we have right now. I did some work with Angelique and she's very keen for you to tell the listeners all about the use of artificial intelligence, both in Airvo and then in a wider context for restroom operators.
Recently released that feature we call it AI assist. Right now it's not wide functionality, but we are continuously working on it. And I'll describe how it works right now and our plans for future and what portable restroom operators can get from AI in general because it really can help them. I remember it all, that you used AI, that you tried it and it really helps you with your podcast transcriptions and all that stuff. Yeah, I released a trailer today for the episode that comes out on Friday.
I used Google. Is it Nolm? I pushed my transcript to no lm and it created an interview between two people who I presume don't exist. But they reviewed my podcast and I used a clip from that transcript and I don't know. Can you tell it's not real people talking? I don't know. It sounds like real people to me. But that is actually crazy because you can just use that and everybody listening to your podcast will say, okay, nice, good, but that will be not real people, but just, just AI.
Token and Vlad, I pushed in the transcript and it gave me a 45 minutes podcast. They spoke about my podcast and the script, and then they used the AI function in the background to generate more content about tenders and bids, which was the subject of my podcast. It's just nuts. That's crazy. So airvote then, how have you applied AI to airvote and what will it do for your users? First, we're thinking about what if we implement AI in air vault and how we can benefit from it.
Because everybody is using AI here worldwide. And we thought, okay, airwods don't use AI, and that's not good. We started with the simplest thing we can do in Airvault. We have something called user vote, that end users scan our QR code and leave a vote. And they can add a comment, they can send a picture or a few pictures. And that's what our customers see and what our customer analyze. And we started from there. We asked AI to analyze comments and pictures.
It can be done in a different way, so we can even just ask, just describe us what's on this picture. And I can describe, okay, this is this, and this on this picture. But this is not what we wanted. We went a different way and we asked a list of questions. So for each picture or comment, we created a list of questions that we ask AI to answer. These questions are simple, like yes or no, like does this picture show deutsche restroom? Or does it show full restroom?
Or does it show restroom lane on its side? Or does it show that toilet paper is missing? Or maybe something is broken. So all these questions and more and more questions we ask and we receive answers like yes or no. And from that point we collect the data and we have information about each picture and comment. We started from filtering out useless pictures and comments so our customers don't get them at all so we can save their time. Sometimes we just use AI to rephrase comments.
If they contain something inappropriate, that's helpful because in the past we just removed the words that were inappropriate, but now we just rephrase them. And that's better because people can get the full idea of what user set without the stars and all that stuff. So you're interpreting the comments and the photographs that users upload. And are you doing that in real time, Vlad? Yes, it's done in real time.
So your customer, your airvote customer then would get a notification that will say something along the lines of, there's a toilet has tipped over. This is what we're working right now. So now we collected that information and what we have right now, we remove useless pictures.
So our customers receive alerts with information and they don't get pictures that are useless, that contain faces, human faces, or maybe just a QR code, because many people, many users that scan a QR code, and when we ask them to take a picture, they just take a picture of a cure code that they scanned. And that's completely useless for us and for our customers. And yeah, that's funny. That's funny because there are a lot of them doing this.
And that was one of our goals, to remove that useless pictures. And it works. Actually, it works. You're already identified, you get some information that you don't want to collect and you're using the AI to filter that out and remove it so that your customer doesn't get inundated with just nonsense. Rubbish. In the beginning we were like, oh, what if we have many, many votes containing data that's fancy or inappropriate, how we can handle that so our customers don't receive that.
And then we understood that we don't see that over using, so we don't see that big amount of offensive comments or offensive images. And we're happy with that because end users who scan and leave votes, they don't write a lot of offensive comments, don't send a lot of offensive images. And that is great.
That shows that the people who use like restrooms or who use pure codes in different industries, airboat is not only in sanitation, but in airports, healthcare, and even in restaurants, cafes and all that kind of places. So we see that people don't overuse them, and that's great. But in that some cases when we get QR codes or offensive pictures or comments, we can remove them or sometimes just rephrase and show them to our customers.
It's really striking how airvote has taken off since I did that first interview with Angelique in 2021. It's just gone from strength to strength, and the uptake's been really impressive. People have really taken to it. The operators love it. And I think customers enjoy having that direct feedback channel. You can actually send feedback in real time to the people that matter using airvote, and it's absolutely tremendous. Yes, that's beautiful.
But it depends because different businesses have different models. For some businesses, the real time feedback is not so important because they get too much feedback and they cannot operate it in real time. Especially that businesses like airports. So they mostly use weekly or monthly reports and go from there. So they analyze reports, they analyze votes and make decisions from that point. But some airports have teams that monitor feedback in real time and fix if something is broken.
Or sometimes some restrooms get clogged. It depends actually. But for portable restroom operators it's different because it has large businesses and small businesses, and sometimes we have a dedicated team or it's just somebody who analyzes votes. Sometimes business managers owners do it by themselves. They get all the alerts and analyze and do stuff and in it. But here we come to the next stage that we are planning to implement.
Having the data, we can filter votes and categorize and prioritize them. This is what we don't have right now, but this is the next stage and that can be interesting to learn more about it. And we want to hear from our customers, especially from portable restaurant operators, what they think about it. Especially for bad businesses who cannot handle all the feedback in real time.
That may be useful to analyze feedback, maybe on a weekly or monthly basis in reports, but receive alerts about critical problems in real time. For example, if they could receive alerts only if maybe water is leaking or toilet is laying on its side, then they will only receive alerts about that stuff about dead votes, but don't receive alerts about dirty restroom or no toilet paper or things like that, so they can analyze them. What's the best way for people to send you that feedback?
Vlad we would be happy if they share their thoughts on Teamair Dash vote.com. I'Ll put a link in the notes and then people will be able to follow up on that. Teamair Dash vote.com yes, I think it's really exciting that you are asking customers how they would like to see the product develop and then using AI to help you achieve those outcomes. I'm guessing that that's the world we live in these days, Vlad, that things are moving so quickly, but AI itself hasn't got all of the answers.
It needs to be directed and told what you want it to achieve. Yes, this is what we faced when we started. AI can do many things, but first we need to set a goal, okay, we need to do this, and then that's very important to give good, clean instructions, ask good questions that AI can answer you and give you information that you need. It can give you a lot of things, but it can lead you to that point when you're not doing what you expected.
So you first wanted, okay, let me do this and achieve these goals, but then you, you understand, okay, that's AI. It's so powerful and we can do this and this and this and then, okay, please stop. We just wanted to achieve these goals and that's it. We don't want to dive so deep to do all that stuff because we just don't need that. The artificial intelligence is intelligent, but it still needs to be directed to achieve a particular outcome.
Yes. Knowing that, I was going to meet you today and record this podcast I've been playing online with the Google LM interface and with chat GPT, and I've been asking them to make me pictures and produce transcripts and do all sorts of things that I didn't know I could achieve through AI. The one thing that struck me is that you ask it questions using normal language.
My knowledge of computer programming in the past was that you needed to learn code, and it's such an accessible tool now that you don't need to write in special computer code or anything, you just ask it questions in plain language. You can communicate with the.
I just like with somebody sitting opposite of you or just using your phone, even using chat GPT application, you can talk to AI, ask different questions, or just ask, tell me something, or let's discuss this and this, and it will answer you in a human language, in any language you want. And that's it. So you don't need to learn something or have any special to use it? Yeah, yeah. So you don't need to learn any code or anything like that.
As a developer, how confident are you in the accuracy and the reliability and the efficacy of the information that you get back from artificial intelligence? Do you trust it? So I don't trust AI, I don't trust even what I can read Internet and even what I hear from somebody, so I can trust her 100%. So I'm the kind of man that should check everything. If I hear something, I know that, okay, this is just a thought or just a fact or just opinion.
And if this is information that I should rely on, I need check it. I need check it in one more source to make sure that that's correct information. That's more about how you use Google. So we used to use Google just asking maybe some questions or just type in, okay, what's this? And this? And Google gives you some results and you click on first link, second link, and you go through pages and you analyze that stuff. But this is almost a way how AI works.
So you don't spend time on visiting that links and going through the website to get information. So you just ask AI and AI does it for you. It takes information that it has. Sometimes it searches Internet for some information and it gives it to you. But this same when you google something and you open the website and you see some information, can you rely on it?
If you're making critical decisions, you just have to make sure that you've based that on something that's actually accurate and reflects the data that you've got yourself. Where do you think it will go from here for restroom operators then, Vlad, what do you think the future for AI in the restroom industry looks like? I think that in the nearest future, portable restroom operators will use AI more and more. Maybe it can come first from CRM systems.
Maybe they will start using AI first because that's where it can come from. Because there are systems developed by programmers that know more about AI and then can implement the different AI features that will be used by portable restroom operators. But maybe some of them who are more in that technical world or who want to learn more about all this stuff can use them on their own. For example, routing.
When we talk about routing systems, some portable restaurant operators use full cycle systems to track for customer management, routing jobs and all the things. But some of them use different systems for routings for customer management and other tasks. And routing system is something that can use AI for sure, because AI can analyze not only how to reach the point, to have less distance, smaller distance, and that's it.
I think that the routing system use many different input points to build their routes, but AI is something that can analyze a large amount of data and provide the result based on that. And even if somebody needs to include some crazy things like like weather or maybe any situations in that point. AI can even take this into analysis and provide routing based on the weather.
At the minute, most operators will go on a fixed term basis, so they'll go once a week or once every ten days or once every two weeks to empty the restroom. And some of them will be very full and some of them won't have been used very much at all.
I would love to see a world where AI, with appropriate sensors in the tank, could record the data from each restroom and then recommend a service schedule and an optimized route that is the most efficient and effective, not just for the restroom operator, but for the customer as well. And I think we're a little bit away from that because we've not got the readily available sensors that can measure the, the capacity of a tank or how much waste is in a tank.
But I think the day will come fairly quickly. Yeah, that's the next thought that I had in my mind, that the easiest way that pros can achieve that is we have drivers. Each pro has drivers. And some of them required drivers to take pictures of restrooms after they done with their service job. And what do they do? Pictures before they do their service job. AI could then interpret the picture and calculate how full the waste tank was before they started to do the service.
Yes, yes, this data can be analyzed later. Yes, this is just one picture before doing service job. It sounds like, okay, it's just one picture, but actually drivers need to do that picture. It takes some time, but it can be used later because when you have pictures that have timestamp on that, it shows the whole restroom. It shows if it has graffiti, if it has toilet paper or paper towels. Is it dirty or not? Is it full or not?
So sometimes it may require not just one picture if they want to get more detailed information and more precise analysis. That's an incredibly easy fix. I think. Yes, I think that this is something that portable restroom operators should work on or companies providing services for them, because it cannot be hundred percent precise with that information because it depends. Sometimes there are ten people using one restroom in a week and sometimes there is one man using restroom per week.
But in standard situations, AI can analyze information based on photos of pictures and provide that data. From that you can see that over the course of a few months or a year, AI has got the intelligence. I can't think of another word, but AI is smart enough to be able to work out when that restroom should be cleaned in order to maximize efficiency for everybody. I agree, I agree with that. This is something that I'm thinking about. And this would get wows from customers.
I'm pretty sure about that. We shouldn't kid ourselves. We're not going to see a point where AI gets out of the truck and holds the wand and pumps the toilet, but it will certainly help the restroom operator work in a more efficient and effective way. Yes. So, Vlad, you said when you first met Dmitry, you talked him out of developing a mobile application. Do you think there'll ever be a time where you need to actually develop that application for Airvote to move forwards?
For a few years, I worked more on managing the project, managing the R and D side of the project, and working on business development. But a few months ago, we decided that we need mobile application. And that's not about users to use mobile application to scan feedback QR codes. It's more about our customers. We have different models in our system called service log. It works pretty simple. It's not a CRM, it's just a model that helps to log services for drivers.
The same QR codes that our end users use for feedback, same QR codes are used for drivers. But drivers, when scanning, sends a different screen that shows different options, like serviced or cleaned or delivered or picked up, and that's it. If drivers need, they can take pictures or write a comment. So it's one QR code. People who use the restroom use that QR code to give feedback.
But you're saying that drivers through the app can use that to record services, which then presumably get sent to the office automatically and then can trigger the next billing cycle. Correct. That's how it works, yeah, that's clever. QR code is just an entry point, and it can be used as entry point to different doors. So that's why we decided to develop that feature based on existing QR codes.
By the way, just thinking about the different doors that QR code open, we have customers working with us who prefer using that QR codes not just for feedback for end users, service login for drivers, but also for their customers. Use the QR codes to give information about the status of the restroom at that time. So the customers can visit different restrooms and say, okay, this restroom is a bad state, has like a bad condition, and that's it. But that's the sidetracking a bit.
So let's, let's get back to service login. And what we found, that service login works well. But with that scanning QR codes, using phone camera and doing all that in browser, it has some problem points that we needed to resolve.
So we decided about developing a mobile application because with mobile application drivers can use a dedicated mobile application and don't use their phone camera, but they use same a road QR codes that end users and they use GPS model in their phone and will have more precise geolocation data. And what a lot of our customers told about is offline mode because some of them don't have cell network where they service units and they need that. In remote areas and stuff.
Yeah, these are just a small amount of problems that we can solve with the mobile application and we're going to release it soon, I think in a week or two. And some of our customers using service log system by Airvault will start working with application and we'll see how it goes. Hey, I've really enjoyed that conversation, Vlad. You've helped me understand AI and what it can do for restroom operators. Thank you very much for taking the time to talk with me.
Thank you, Pete. It was nice to meet you. I'd like to thank Vlad for taking the time to record this episode with me. It's still quite early days, but I'm very excited about the way that AI is already enhancing Airvote and will go on to help improve efficiencies for restroom operators. And of course, it won't stop there. The possibilities for artificial intelligence are endless. If you're already an Airvote client, you'll no doubt get to try their new app in the not too distant future.
If you're not, jump onto their website, air dash vote.com, and sign up. Tell them you're a friend of Pete from get flushed. I'll put links to the Airvote website and both the episodes I mentioned in the notes for today, but that's all I've got. Thanks for your time. I've been Pete and you've been listening to get flushed, the world's favourite sanitation podcast.