with 🎙️ Victoria Edwards – CEO & Co-Founder of FIDO Tech
💧 FIDO AI is a data-as-a-service (DaaS) end-to-end leak detection solution that identifies leaks, sizes them, and tells you where to dig to repair them
What we covered:
🎵 How water networks sing, and how you can leverage this for leak detection
💧 How non-revenue water is a threat in a world of rising water scarcity and why we shall fight it
🔥 How no other industry would actually allow losing 40% of its production without reacting
🤖 How artificial intelligence can turn leak detection around by drastically increasing accuracy and reducing false positives
🖥️ How monitoring a leak’s evolution is almost as important as detecting it, and why you’d probably not want to repair all of them straight away
🧰 How despite its power, AI won’t replace humans and hence never is a threat but rather another tool in the water toolbox
🌐 How FIDO ambitions to save the planet one megaliter at a time by taking a disruptive sensor agnostic and AI approach to
🎹 How Victoria’s musical background is an asset in that endeavor, even if she no longer plays Rachmaninov
🆕 How you need your innovation first to help and solve a challenge, then be transformational, for it to make a dent in the utility world.
🤙 How being sensor agnostic and adopting an open approach is the best approach.
🪃 How it all started for FIDO and what step they already underwent
🤑 How FIDO removes adoption barriers with its CAPEX Free / Data as a Service
🏭 Training on some of the World’s most complex networks, striving even with little knowledge of the network, fixing leaks as an alternative to increasing production capacity, building hardware, pivoting the original concept… and much more!
🔥 … and of course, we concluded with the 𝙧𝙖𝙥𝙞𝙙 𝙛𝙞𝙧𝙚 𝙦𝙪𝙚𝙨𝙩𝙞𝙤𝙣𝙨 🔥
Teaser: 92% accurate leak detection
🔗 Have a look at FIDO’s website
🔗 Come say hi to Victoria on Linkedin
is on Linkedin ➡️
Infographic: Highly accurate leak detectionInfographic-Victoria-Edwards-FIDO-Tech-Accurate-Leak-Detection-Artificial-Intelligence
Quotes: Accurate Leak DetectionQuotes-Victoria-Edwards-FIDO-Tech-Accurate-Leak-Detection-Artificial-Intelligence
Table of contents
- What we covered:
- Teaser: 92% accurate leak detection
- Infographic: Highly accurate leak detection
- Quotes: Accurate Leak Detection
- Full Transcript:
- Saving the World one mega-liter at a time
- Starting a AI Water Company from scratch
- No other Sector would accept a 40% loss like the Water Industry
- How to take off in the Water Industry on the fast track
- Leak Detection today is quite anachronic
- How to remove adoption barriers with creative business models
- Repair only the leaks that make sense, when it makes sense
- Coin Flip vs 92% Accurate Leak Detection
- Rapid fire questions:
- Other Episodes:
These are computer-generated, so expect some typos 🙂
Antoine Walter: Hi, Victoria. Welcome to the show.
Victoria Edwards: Thank you for inviting me on.
Antoine Walter: We had a very interesting discussion just before kicking that famous red button. And you told me something about the place you rats, which makes for a perfect postcard, and I’m really going into its traits without even seeing you high or welcoming anything because you have something really, I would say, not so common about the place you’re at next to Oxford, which is Vista, which I’m probably pronounced very, very bad, but what can you tell me about Vista, which I would ignore by.
Victoria Edwards: One fact for a pub quiz Vista village was just the closest rail station to us is home to Vista village designer, retail outlet. It’s the most visited location in the UK after Buckingham palace. And it’s unique because it’s the only railway station in the UK that announces that they’ve got to that station in it’s first language, which is Mandarin, not English.
So. So that
Antoine Walter: was setting you for a kind of international path somehow?
Victoria Edwards: Absolutely. I don’t know which one came first, but yeah, we certainly hit the global stage. Right.
Antoine Walter: There’s another thing, another elephant in that room, which I have to address right now, so that it’s done your company’s called Fido. And I guess if you just type Fido into Google, it’s probably not your company, which is going to come up first.
So what does Fido standard. Yeah,
Victoria Edwards: it doesn’t come up first yet, but it will. We have grand designs on that. So originally a typical startup new marketing budget, and it’s actually an acronym for free, intelligent domain observers. So it will look at any environment and make predictions based on that. But actually I’ve got a Mudder Irish red setter dog called committees.
So we put the two together. Hence the name five.
Saving the World one mega-liter at a time
Antoine Walter: And if we were to meet in an elevator, you know, that’s really the standard classical question you have just 20 seconds to tell me what Fido is doing, what you would be saying.
Victoria Edwards: Very simple. We’re saving the planet one mega liter at a time by taking a disruptive sensor agnostic and AI approach to non-revenue water when
Antoine Walter: it was preparing for the discussion.
Wondering, I mean, I’ve seen the two cases in the people which were on that microphone so far, there were people inside the industry, which happened to find out there’s a problem. And then that developed a solution to that problem. And there were people from outside the industry, which come with a certain tech and say, Hey, by the way, that sounds something that could be applied into water.
Let’s try it out. Which one would you. Oh,
Starting a AI Water Company from scratch
Victoria Edwards: well, absolutely water virgins. We started two and a half years ago. Um, but our background, I’m a musician and I majored in acoustic musical analysis. So leaks, acoustics, the noises they make are in my DNA, but we’ve had an eclectic career dealing with big data analytics, but also operational field people.
And it was a purely chance conversation. One of our long-time friends phone. The CTO and said, I’m trying to sell a fact book solution to utility companies. And all they want to talk about is leakage. Can you think of anything? And our CTO was stuck on the motorway on the MCX in the days when we have traffic jams and started thinking about all the analysis he’d done and came up with Fido, AI and the form of differential analysis to find leaks.
And that’s where we see.
Taking the leap of faith and entering the Water Industry
Antoine Walter: At that very moment where you say you’re quite Virgin to wardrobe, what do you know about this leaks? This non-revenue water? Is it something which is mainstream enough that you’ve heard of it or is it totally new?
Victoria Edwards: It was totally new for us, but it was quite simple. We’d, we’d always come from a very inventive backgrounds companies.
And I actually looked at the Sunday times newspaper and I got so fed up of our CTO, looking on the business page and going, oh, they’ve just raised $20 million for an idea. Right. I thought about that three years ago. So I hit him around the head and said, if you ever say that again, I said, you come up with an idea that I could be really passionate about.
Has a need is something that we can develop. I’m going to shoulder in and I’m going to give it my all and fighters. One thing I could see because without water people die, it’s quite simple. We’re not making widgets. We’re not making something go faster. We’re not trying to make something necessarily more efficient.
We are doing something that helps. The water chronic water shortage this day, zero that we’re facing also in the UK. We’re lucky enough to see those a real need that was driven both by the utility companies wanting to fix the problem, but also UK regulation. So there was a perfect storm. There was regulation and desire, and there was an opportunity because the problem wasn’t being fixed, there’s some great technology out there, but they’ve all got the same fundamental.
Taking human errors out of leak detection to make it accurate
That all proprietary. So they don’t talk to each other. Don’t share data and they all, at some point require human analysis to find out if there’s a leak. So we said, if we’re going to do this. It’s a world’s problem. It’s something we can be proud of something. I call the Fido face that we are tackling non-revenue water and water leakage, water scarcity.
We’ve got to make sure that we are non-proprietary and we’re going to remove human error from the leak detection cycle. And that’s what we set out to do. And that’s our pre. How can you have a world where four out of 10 people, and we’re not just talking about Sub-Saharan Africa, we’re talking in the UK and Europe and the USA are facing waters insecurity and water scarcity.
No other Sector would accept a 40% loss like the Water Industry
It’s in conscionable. I mean, there’s no other industry in the world that would produce something. And then lose 40% of it. Can you imagine BMW building a phenomenal factory producing a hundred Supercars and then burning 40 on the forecourt? Yep. The water industry does that. So we saw that as the problem, we saw it as the opportunity we saw that we needed to be different because we can only tackle this global crisis by being collaborative, by being agnostic, by being open data.
And that. Y we said, this is going to work. We’re going to make Fido work. We’re going to take this and run with it really hard with a mixture of good luck timeing and the genius of the team that I’ve got here at Fido, we seem to be getting somewhere
Understanding the water networks
Antoine Walter: when you’re saying proprietary and you on the other hand being agnostic, what are you referring to?
Is it software? Is it hardware? What is usually the closed and.
Victoria Edwards: Pipes make noises don’t they leaks actually make noises. So if you want to find a link, you’ve got to listen to a network and networks can be very noisy places that can be generating noise, drilling noise, trams ahead, owls, hooting, and all sorts of different things.
Pressure management valves, that mask the noise of a leak. So there’s great hardware out there that listens for leaks, but they all require human analysis. And we said, well, look, let’s build them on. Let’s build an AI model that can take any audio file from any hardware and tell you if there’s a leak.
Okay. So we’ll come on to how we got this weather, United utilities, innovation lab, which was, was huge for it. So we built a mobile and we built it on an absolute truth. And there’s only one absolute truth in linkage is that you’ve dug a hole and you found him. Well, you’ve dug a hole and you haven’t. So we manually verified every piece of data.
We fed our training library because you can’t have good AI without a good solid training library. And so Fido learnt to be able to take audio files from any piece of hardware and find not only if there’s a leak, but most importantly, the leak something. Now for us to be able to tell you where the largest leaks are, first is a phenomenal.
Planning your leak reparation works
You don’t want to dig the road six times to find six small leaks, dig a road once, fix the big leak, let the others track their degradation till they become economic to repair. And we said, actually, this AI can be sensor agnostic, and actually must be because you were not going to solve this global crisis.
By trying to, we’re not giving away the keys to the executive washroom. We’re here. We’re saying to the hardware manufacturers come work with us. We can together really solve this problem. And you’ve still got your business model, but we’ve tipped that on its head and gone. We’ll take any audio.
Antoine Walter: There’s a lot to unpack in what you just said.
And I have to take notes because really there’s a lot of the directions I’d like to explore, but I’d like to take you to the very beginning of what you said about United utilities, because your drawings, their innovation lab, if I’m right in 2018, what I’m wondering is your product, what does it look like at that stage?
And what does United utilities do in terms of leak detection at that moment? How is that encounter? Is it like a win-win and an epiphany or is it. Build that a bit more.
How to take off in the Water Industry on the fast track
Victoria Edwards: Well with all the typical arrogants of a startup, that’s just really got a concept. We rolled up to a 20, 65 conference in Sheffield, promising the world, and we’ve got three little balls, plastic balls in a cardboard box.
Cause originally we were going to go in pipe. We’re going to collect the data by being in. Flowing down the pipe, listening to the leaks. And I remember Kieran Brocklebank. The head of innovation is United utilities was also at the conference and he came over to our trestle table. He looked the banner up and down, looked at us, looked at the plastic balls and went, Hmm, I think you need our innovation lab.
He said apply. And literally I’m terrible at filling in forms. Right. And I was sitting on a train from. Oxford to Glasgow. And I thought procrastination time over I’ll fill in the form. Cause Karen kept, you know, say you should apply. You should apply, you know, give it a go. And I’ve read about their innovation lab.
And it’s actually one that does something. It doesn’t just do it for ticking the box. It actually really works with it’s very intense. I thought, oh, we won’t get it. We’ve just got this idea. And so I applied. And we were lucky enough to win a place. Um, what that innovation lab gave us was two very important things.
Fail fast and early!
One, it gave us the actually the environment to fail, which was really important. We went through eight iterations of our offering, completely changed. By the time we came out, we came out with a sensor agnostic AI, and we came out with a methodology of finding leaks without ever having to go into. Cause putting things in the Bolton network in pipe is not a particularly good idea.
So we said, actually, we tried it and we lost one. I mean, we’re not doing it anymore. We are going to find a way that we can. Any data from the external of a pipe or from a hydrophone or a lager, but we will never go in pipe and we will find both if it’s a leak, the size and the location. And what you you did is they had complete faith in us.
And we had to prove that we were worthy of their trust and they worked very hard, but they gave us access to people. They gave us access to build something. That they actually needed that the industry needed. I mean, lots of startups have some great ideas, but they come with the product, go knocking on the door and go, we’ve got this one.
The tricky business of starting a Water Business
You used it for, you’re really stupid of what United utilities did we say, actually, we like your idea. We think it’s got, these are the challenges that we have. How can you deliver that? And I was actually determined because people say, don’t walk in the water industry, it’s taken you two and a half years to get a contract.
You should be taking you to all and gas. And I said, well, no, because water for me is a matter of life and death. This is what’s important to me. And I said to the team, when we want to place on the innovation lab is 10 week accelerator course. The. And we, you know, we tried to pretend we were bigger than we were at.
We, you know, but we decamped to the cheapest flats, we find near their head office. And I said, right, my promise to you guys, you’re the geniuses that have invented this. You’re the ones that are going to make it happen. But I promise you, I will not leave that lab until I’ve got a contractual agreement and everybody was going to get mad at you, man.
And you, you were so fantastic. I think I squatted for three months afterwards and wouldn’t leave the lab, but we’re there. We now have a, long-term not only a business as usual for Fido AI, but a development long-term development contract with them. And I all credit to them because we were taking time away from people’s busy day jobs, but they helped us craft something that we can apply globally and we continue to learn and we continue to do that.
Leak Detection today is quite anachronic
Antoine Walter: But you mentioned that they had challenges. So what were these challenges at United utilities and what did you solve?
Victoria Edwards: Well, I think the challenges, it’s not just, I think it’s to do with all the utility companies. They’re massive, massive organizations. And so a lot of information is. So actually we’ll you need a joined up approach.
So if you’ve got people trying to sell you things in new tech all the time, there are some things that work in some environments and some things that don’t work as well. But what the problem is is you can be inundated with too much. So you’ve got all these systems telling you where there could be a leak here, or when it rained a bit over there, or it might, you might get a leak over there next year.
You’ve got people still out in the street using listening sticks and where it is so critical and it’s very brave of utility companies to make a change. And if they’re going to make a change, it’s got to be a transformational one, not a transactional because the risk is too. If the prize is big enough, if they make that leap, if it’s only a transactional, they’ll say, actually, we’ll stick with what we do.
We’ll just do more of the same, but they were open to realizing they probably could get better value out of the equipment and the hardware they’d already spent CapEx programs on. And what we said to you, we promise not to give, but Fido, just more data, but that doesn’t do anything. We’ll give you actionable data.
How FIDO intends to shift the paradigm
So we will say this is the result. And therefore. This is the next we will give you prioritization. So we’ll tell you it’s a leak. We’ll tell you the frequency of the leak, but will tell you the size. So you will be sending your engineers out and especially during the COVID pandemic. And you’ve got to think about people’s health and safety.
You don’t want them wandering the street with a listening stick. You want them to go out to where they know there’s a leak in that facility and it’s a large one, so you can prioritize. So I think that’s how. I mean, you have to speak to them about it, but we’re, you know, we’ve been in their annual report and we worked really closely with them.
And I think that the challenge that they had that most other utilities challenges have is there’s a lot of great technology out there, but how do you use it and how do you use it in a way that just doesn’t mean just more data that a human still has to process. So here at Pfizer, we said, we’re going to give you actionable data that removes human error from the leak detection cycle, which allows humans to go and do the more complex.
It’s not to take their jobs. And that’s the big thing about AI, which we probably discuss now, but people think you’re going to lose your jobs. You’re not, it’s going to allow humans to do the more complex connections and work that they need to do rather than the mundane day in, day out with uncovering accuracy.
And that’s what AI does.
Antoine Walter: I have a curve ball for you at that stage because whatever you’re explaining right now, it makes me think that maybe the UK is a bit specific. The facts that you have this limited number of utilities, which then have quite a bunch of technologies and things rolled out because in many places around the world, you don’t drone in data.
You just have no clue about your network. Would that be an issue for
Victoria Edwards: you? Very simple. And we have came at this through bitter experience of running field operation teams in the Wi-Fi space. So if you’ve got a pipe and you’ve got water in that pipe, Fido can find the leak and rank it by size for you. So during this terrible, terrible pandemic, we’ve launched in places such as the democratic Republic, South Africa.
Yes. They’re not digitized network. So say how do we know? We don’t know how many leaks we have. We can’t even tell probably where part of our network is. And so we thought we’d sit pretty it a good curve ball question. We thought there’s loads of hardware in the UK. We can just analyze all the audio files.
How to remove adoption barriers with creative business models
Hardware agnostic and we’ll do the same globally. So we met really looked outside. The UK went, oh, we haven’t, can’t see this, this hardware, where are the locals? Where are the hydrophones? Where are they? What’s happening? So we went and asked and we got the same question back time and time again, one, it was very CapEx heavy, and two, they haven’t got the humans to do the analysis.
So we said, okay, if we remove those adoption barriers, If we smashed them open, would you like to know where your leaks are? And they went well? Yes, but how can you, so we, those little plastic balls that we’re going to go down inside the pipe, we turn them into a little Fido bug and we give it away for free.
So we charge no hardware. We just say it actually it’s like the sky concept you pay for the content, not for the boxes, delivering the content. And we’ve developed this little bug and it’s always on, and there’s no technical input. You just put it on a pipe. We’ll put it in the chamber on a stock Tupper on the customer’s meter box to put it on.
Press record overnight or an hour, however long you want it to, it goes to record foreign mobile phone goes up to the cloud instant results from Fido AI. It’s a dumb device, the clever stuff’s done by the AI. And they said to us really well, how can you give us leak size? Cause we don’t have DNAs, you know, district meters.
So we don’t know what’s coming in and what’s going out. How can you tell me that the noise you’re you’re listening to. And I said, well, you have to remember that Fido AI, what it does is trained on some of the most complex networks in the world, which have DMAs. So we could verify our leaks when we’re saying larger.
Correct. And we’ve done all that verification. So when we hear a noise and Fido says, it’s a leak, there may be a leak hidden behind something else, but there’s a leak there. We give it a unique Fido ID number and we turn that leak into a living, breathing. And we don’t forget about it. And we track it all the way through wherever it is in the world.
You could be running fighter bugs on the shams Elisa. You could have a logo in Rio de Janeiro. You could have a hydrophone in wet and windy Wigan. As long as it takes an audio recording, we give it that fighter ID number. We track it through its repair detection, repair circle, and we have the continuous feedback loop for the AI.
So it gets more and more accurate. And that’s a true. Global collaborations, except he says, oh, we do collaborate in the water industry and I’ll go, do we not so sure. Anyway, so what Fido AI gives you is the ability to raise your network to the level of the best. And that’s what we’re really proud of that we can do.
The perks of being CAPEX free
Antoine Walter: You mentioned DMEs, which is the opportunity for me to just direct the people, listening to that to two other episodes I was doing with Olivier Narbey in season three and with Luke Butler in season four, where we covered those topics of DMAs. So we don’t have to recover it, go and listen to those. It’s going to be, I guess, pretty interesting.
Plus Luke butter adds this element of open source. We also have that discussion about the closeness of data, which you mentioned. So I think that is really complimentary to the discussion we’re having. Right. What they were mentioning as well, is that with this topic of a non-revenue water, even if you remove that boundary of CapEx, because your CapEx free with your solution, there still that element of when is it worth it?
Solving for leak. And when is it simply not a better, but probably a simpler solution to put a new treatment plant and add more water in the system. I know that’s not sustainable. It’s just, if we are realistic and honest with the water industry. It’s what happens most of the time you’re missing water, you produce more water and then only if really you’re out of water in the full region, maybe you consider addressing the leaks.
I’m painting it a bit in black right now. But I think I’m not that far from the reality. There’s one thing which was a bit almost shocking when I was looking for. Statistics to prepare for all those discussions. It’s that there’s not even a worldwide acception of how much water we’re losing. So it sounds like something which is a bit like up in the cloud and read the estimates.
They go from simple to double to triple my question here is when you’re meeting a utility the first time. What is their states of information? Do they tell you I have 37% or 12% of leaks and like to solve half of them? Or are they really looking for an advice coming from you where you tell them what they have to do?
Victoria Edwards: It depends on the maturity of the utility company. I think whether the, any customer or utility that you want to work with, you have to gain through. So there’s no way in a first meeting, they’re going to say we’ve got a real problem here. And we don’t know where our network is and we have loads of dry decks and we’re in the wrong place.
They go. No. Well, I know exactly where my network is. We never over dry dig and we’ve got leakage under control. We have to get around that leakage is seen as a failure. It’s a fact of life. It happens on your network. And I saw a comment about plastic pipes. I think that’s hilarious. They don’t leak well, they do.
You just can’t hear them, but Fido can hear them if they’re
Antoine Walter: installed the right way.
Repair only the leaks that make sense, when it makes sense
Victoria Edwards: It’s the fittings, not the pipe. Um, you’re quite right. When you’re looking at the utility companies, some have great data, know exactly what’s happening and they can say by leakage, but it’s 5.8% and you know, But it’s also really good to work with ones that say, if they’re honest and say, well, we don’t know.
And we said, well, let us give you a snapshot of we, how bad or good we think it is. And there’s no way you can get to that economic level of zero leakage. It’s not economic to do so, but surely just by fixing your large leaks, which can save you X percent is going to be cheaper than any building extra capacity.
And that’s a real Sopot moments for me. I am against building extra capacity. Where you can’t fix the water you’ve already produced when 4% of all global electricity and power is spent producing portable water that’s outrage. You can’t just do more of the same or say, well, we’ll just find more. We’ll forget about the stuff we’re losing.
Now we know we’ve got the technology. We know we’ve got the expertise and I’m not talking fight. I’m talking as an industry to fix these problems. You fix what you’ve already produced first. You don’t just go looking and desalination and you know, abstraction is absolutely. Second choice. It should be a choice of last resort.
Well, you should be honest and learn what you’re losing and then come up with a program and a policy to fix that first.
Antoine Walter: Yeah. If I’m coming back to my question about the data you have, when you’re starting, if I get your rights, you’re doing kind of a miracle because. You could be having a network where you have parts of the network, which are no longer in service, but still are still in the network.
So those pipes are somewhere. Sometimes you even have water, you don’t really know where it comes from. You just happily test it. And it’s still good. So you’re like, okay, I don’t want to understand what’s happening here. And you can have like a piece of ductile iron, and then you have a change and then it’s plastics, and then you have new change.
And it’s, uh, I mean, Russia to that extent is very special because I think they even have stainless steel. Pipes in the ground, which is a thing, the only country in the world, which is doing that. But, but you can have all this changes in material a bit everywhere. And I’m wondering how can you overcome the facts that you have no clue about what’s there in case you have.
Accurate Leak Detection… and material agnostic
Victoria Edwards: All right. So yep. You can always go to a new environment. So the way that the AI modeling works is it looks for exceptions as well. So we don’t cut into our AI. There are very clear reason why we don’t, because we’re not a simple pattern recognition technology. We’ve come at it from a different way, looking for best outcomes at each decision-making point.
And therefore we don’t need to know the pipe material. That we’re hearing the order. We’ve learnt that Fido AI has now learned the individual characteristics of a leak in different environments. Cause acoustic propagation is a huge topic and I would encourage you to read, we’ve got a brilliant article from a PhD, a KTP associate this working full five day local jail ed about acoustic propagation.
And, and if that is typical. So when fighter goes to a new. And it collects files that fall outside our clustering. It sees it as an exception and we take that and we use it as part of the training model and we train it out. So for instance, you may get copper somewhere. So for us, it doesn’t matter that, and this is when we come on to exact location because we can give exact location.
It doesn’t matter if you’ve got cast and then you’ve got ductile and then you’ve got plastic. You’ve done repairs with different. It can do that calculation. So it’s really important. And the thing that’s the beauty of the AI, because we only feed it verified data. And we trained on these networks that it now can see what’s an exception go.
This is an exception. This is an outlier. I need to drag that out and look at what it is. And then we feed it back in. It gets trained out and then gets added in. Into part of the decision-making process. And, uh, and I think that’s the beauty. So I find it makes me smile. People say to me, I mean, my mother is very upset with me because I was trained as a classical pianist and she had great hopes that I’d be, you know, well, acoustics in leaks is the nearest thing I’ve got to that mother.
So I do apologize, but I think comes serving the world better than trying to play dodgy rendition of Rachmaninoff. Second piano concerto.
Antoine Walter: When I hear you describing the AI, I have the feeling that you’re your biggest assets. If you say the AI itself is not patented, is the library you have inside the AI. Am I fully. It’s
Victoria Edwards: not, no, because it’s like, it’s like a good version. I don’t know if you’ve ever seen little shop of horrors as a plant that keeps bringing him to be fed by Seymour.
And he keeps dripping in more, more of himself. Is that Fido’s is it’s a thing of beauty, actually the Fido AI, it started out learning it learned on that. Absolute true. So it built up this library and that library is constantly growing is over two and a half million files, but it’s learning the characteristics of it.
So it’s adding into that library, but it’s also not a simple pattern recognition, so it can take the AI and say, these are part of the characteristics I’d expect to find in this leak and have a leak of this size in this environment. But these are some other characteristics that are unknown. If there are enough of the unknown, it then gets fed out into an exception.
So it’s not as simple. And it teaches itself with that unique Fido playback number. It knows then what’s going into its training module. We do a lot of work on false positives and false negatives to check that we’ve got that accuracy level. We’re at 92% accuracy. Now that we’re contractually. We stand by that we’re actually sort of more than that unofficially, but I’ll, I’ll get shut down in flames.
If, if I give you that figure and we’re constantly improving anybody that said it perfect is finished, is lying. You know, AI by its nature. If it’s built properly, the concept on which it’s built, it’s built on proper foundations and not on sand and Spiros data will always be. So AI, we started, we could just tell that there’s a leak or not.
We then craft the leak sizing, which is the holy grail. We’ve cracked leak sizing. We can tell pipe material. We can. Now we’re about to look at things like we’re going to tell you what type of leak it is. So you know what equipment you need to go and repair it from doing that. You know, we’re building this knowledge for you as your utility comes.
Cause we don’t own the data. We’re not one of these companies say is your data, your network, all we’re doing is processing it and giving it back to you as actionable insight with the results that you need. And that’s a fundamentally different approach. And I think it’s one that people need to take. So we’re very good at saying we keep.
Coin Flip vs 92% Accurate Leak Detection
Antoine Walter: We’ve covered two of your perks. I would say the first is accuracy. You guarantee 92 person is probably a, both, but you go into 92% just as a reference. You, you mentioned if there’s a human involved, what would be the typical accuracy you would expect?
Victoria Edwards: Um, well, it depends on your level of skill, but it’s about tossing a coin, but it’s also speed.
So maybe they get it right 60%, but what they can’t tell you. And as a game we often play, as I often say to people, I’m going to play you two noises, tell me which one’s a leak. Okay. And the room, the audience. So we normally get a fairly even split now. And I said, well, you know, the ones that said noise, one was a leak.
Well done, Bravo. Right? Who can tell me this? Nobody can tell you the size humans can’t determine the size. So what fighter AI is doing, so you don’t want to human sitting there listening to it. You want another ring, accuracy of leak, no leak. And you want to know this. And the frequency of that leak. And then you’ve got enough information to really find it.
So humans need to do the more complex things. So I don’t know if you’ve ever seen the analogy. There’s a picture, isn’t it? The 12 evolution steps of man, it’s all done. Everybody’s t-shirt at the seaside. You can go and buy that, you know, and actually out of this 12 steps, AI is only at step two. It’s not going to replace humans.
It’s there to remove from humans. What is a repetitive task, but to do it with accuracy and at speed and at scale. So the humans can concentrate on the more neural network intensive, complex relationships that the human brain can muster.
Antoine Walter: I’m going to check in post production. If I add here the noise of a leak, because I have to be transparent with you at the very beginning.
AI doesn’t replace humans: it enhances them on their limitations
What I wanted to do is prepare you some noises, which are leaks, which I found on your website in which I found left and right. And some other noises, which are ASM videos on YouTube. And it’s very clear. But I think it would be an unfair game because that would just demonstrate that the human doesn’t detect the difference as well as an AI would, which by the way, would be proving a point.
Victoria Edwards: Absolutely. And we’re, we’re, we’re happy to say that, but in certain materials, you know, the human air. So a lot of people still using people to walk up and down the street with the listening stick. Right. And we say actually, but human ears can’t hear. So the largest leaks of the quietest and especially on plastic, you’re never gonna hear.
The leak. Yeah, because if it doesn’t fit, there are leaks, maybe not in the pint meter, but in the fittings thoroughly. So see it, it’s something you’re not going to hear large leaks. So you need a dumb recording device that records everything that doesn’t sit and wait for threshold because a large leaks are often not that loud and let the AI do that.
And that’s what it’s trained to do. That’s what it will do day in and day out.
Antoine Walter: To bring it to my truck. The first Pyrrhic was this accuracy and compared to the human, which might be a bit better than a conflict, but still there’s a way to go. The second was the size. And you’re just explain how that is difficult to humans to determine the size.
Whereas that is like daily bread for the AI. The third, which if I get right is pressing you and your offering is dig direction. Can you tell me what Dick direction.
Victoria Edwards: You have to walk in the shoes of the person that’s using your technology with your results or your data or your hardware. And I’m a great believer.
If you haven’t done it yourself, it’s not right to then enforce it from on high. So we spent a lot of time out on networks and a lot of time with the U U engineers and others in the UK. And we noticed when they got to dig, you know, so an engineer has been out using whatever technology and the blue bar says.
And so we said, actually, instead of aborting that dry dig you’ve in this note, Can we make it so simple to allow them a methodology to carry on digging. So it’s going to be either maybe digging in the middle of the night, it’s wet, it’s cold, or it’s in the middle of the day, it’s boiling hot. They don’t want to plug data in and have to filter things.
It’s a big deal to be wrong when fixing a leak
Right. And they get it right. But sometimes they get it wrong. It’s really expensive to get a whole crew out, to come and dig it, upsets customer service, you know, the roads blocked again, or you’ve dug it and you can’t find it. And I’m still getting, and you’ve been sending me a bill and you have a 50 leakage.
That’s not their skillset. Their skillset is to get that leak repaired as quickly and as efficiently as professionally as possible. So we said, right, just take a bug. And it’s always on, never goes flat, right? Because equipment always goes flat in the back of an engineer’s ban. You know, the only thing that’s normally charged is the mobile phone.
… so AI also has your back there
So we take a mobile for you. Take a little bug, put it on one end of the exposed pipe where you’ve dug that it’s dry, but another bug on the other end and you press dig direction on your phone, on the Fido app. And it says the leaks that way from point a keep digging. So it allows you to keep digging, but we get lots of uses for that.
We then are asked sometimes to actually, can you tell if the repairs been done? So, what you do is you just put a buck down again, where the repair has been done, records, everything, very simple device, send it to AI, the Fido AI, and find the way I says, lick noise has disappeared, repair complete. So, you know, actually it’s been repaired properly and that the noise has dissipated.
So you’ve got a bit of sort of quality assurance there, but I think the big one for us is our, as our correlation, we can tell you the exact location of the leak without human analysis.
Antoine Walter: So with just two bugs, you don’t need to try and get. So you put two books and what’s the distance you put between those two.
So the difference between
Victoria Edwards: the digging. So the bit is we’ve told you the engineer from the sensor, whether it’s a Fido bug or an acoustic lager or a hydrophone, there’s a leak and the size of a leak sets a large leak. You now need to narrow it down. So you can mark that one meter bar, or sometimes in some areas in the world is a box or it’s a different color so that the DIC team know where.
So an engineer will go out into the field. He may use the listening stick, see if he can hear the noise. And where is it? Loudest, top sounding it’s called, or he may have a ground mic, or he may have a correlating device where he has to put in the pipe material. It’s all an edge device. And we said, no, no, no.
That doesn’t work to get an exact location. It’s the speed. And the correlation between two noise sources. Okay. So if you can imagine there are 85 billion milliseconds a day. If you’re recording of those two different noise sources is out by more than maybe eight milliseconds. It’s going to throw out where the pinpoint of the leak is.
So we said you need cloud computing power to do that. You’ve got to do it in the cloud. You’ve got to do it with Fido AI. So we said, let’s make it simple for the human. Remember no human analysis. So you take your. And you tap it together and you put one bug down where you’ve been sent to saying that we can hear a leak in that vicinity.
You put another bug at another asset. You can find, press correlate on your mobile phone app. It does a recording of the two sounds, sends it to Fido in the cloud, the cloud. We do the algorithm, the mathematical calculation, and it’ll tell you it’s 126.3 meters from buggy. That should dig. You don’t have to tell us apart material and it’s designed to be simple because what you’re doing is you’re using the power of cloud computing and removing the ability for human error.
So I think that’s one of our biggest achievements as well, is that we are empowering. And this is what I’m passionate about. We’re empowering any utility, regardless of the skill of their work force. To tackle this global problem of water, scarcity and insecurity. So we’ve got some great in the UK and some of the other countries and in the far east and Singapore and places like that, really skilled technicians.
And that’s great. So they can delve into Fido AI. They can look at the results. They can listen, they can see more analysis, but if you haven’t got that knowledge, we’re not going to deny you the fact that you can’t fix leakage. We’re going to say, we’re going to make it simple. We’re going to walk you through the steps.
So I think. That’s what I’m proud of. It’s clever. Yes. Great. Is it usable? Absolutely. And that’s the bit, I’m most proud of that we’ve taken it and said actually it should be available to everyone.
Antoine Walter: You mentioned cloud computing. Let’s explore your vertical. If you will elect to understand where it starts and somehow where it stops, you mentioned the hardware.
So if I take the vertical analogy, that would be the bottom. So you go from your bugs, but it could be also something else than the bugs that feeds into your system because you’re agnostic and then it feeds into the AI. So the AI, can it be on premise or is it always in the.
Victoria Edwards: It’s always in the cloud. I mean, there are obviously in different parts of the world.
There are different data requirements. Maybe we don’t own the data and we don’t store the data. We just deliver it back. The conversation you have to have is if it’s on premise, it can be done, but you’re not benefiting because Fido is learning every day from every file anonymized file that it takes. If you’ve got an on-premise.
Then it’s more static. So we would say, actually think about that. We don’t own the data it’s anonymized. It comes to as an accuser, we just give it back and we give you instructions to do to that. So why not benefit from that? Where everybody benefits, but you know, so it is a plan and we push it out to any platform.
It’s your data, how many you want to deliver it is up to you. So when you’re talking about that next level up in a way of, we’re not a platform from provider, we’ve got no interest in being a platform provider, we will deliver. The data to you and whatever shape format you want.
Antoine Walter: So when you say pushing to a platform that would be a GIS system, that would be, what could it be?
Victoria Edwards: be anything it’s, it’s all the big, the big players, your name, the nosy, um, Salesforce, which has done enough for SAP or a core could be the utilities own. It could be their data lake. There’s a lot of move for utility companies. You know, to have their own data lake, because then they have ownership of the data.
So we say, yeah, absolutely good idea. You know, you pull it in the new, interrogate it in the way that you want to do. And I think it’s important to do that. So we’ll work with any platform or any utility company to deliver that. Which
Antoine Walter: brings me to an acronym, which I discovered on your website, I’m sorry, I’m a lay man.
It might exist. And you might not be the only one, but that’s the first time I came across it, which is data as a service, because it was trying to understand your business model because if I get it right, you’ll know CapEx, which means the bugs come for free and you push the data to a bigger platform. So it’s not like you’re bringing everyone to your proprietary environment and then selling add-ons or whatever, which you could have inside your closed system.
So, where do you make your money? Is it on that data as a service? I guess? Yeah. It’s on
Victoria Edwards: data as a service, we take the data, you you’ve given us the either we’ve collected or third-party has collected and we pushed out the results and also the insights you get from that, those results. And it’s different from software because software is a, is a licensed model to install something on your system.
We’re not installing anything. We’re just delivering data into yours. And it’s a common term data as a service. So you will see that more, but that’s very new for the utility industry. Very new for the water industry. They used to often CapEx. So when we first started, um, they said, no, no, no, we buy hardware.
I go, oh gosh, how are we going to do this? And we said, should we package AI into a big black box and just give them the black box because you know, how do we explain what it does? But the utilities company knew you’ve been great at leading this. And they’re, you know, others, you water and set the trend to people in the UK and DC water outside and out in Australia.
And they get the subject. The data as a service works so we can give them the immediate data, you know, the leak, the size of the leak, you know, the frequency. We can give them exact location if you’re using the bugs, we can do, but we’ll also do more with that data so we can feed into your strategic.
Because we can look at now, this is what’s happened on your network with absolute certainty. Could we guaranteed it all the way through? We’ve tracked it through the leak. We’ve tracked it through the repair. We’ve given you the volume metrics saved. So we produced an impact report for them. I’ll never be arrogant enough to say we can do predictive analysis.
And I think it’s very dangerous thing to say. You can do it because you can only train your AI to do predictive analysis. If you’ve had the results back to do that, what I could see is it. And if you predict 30 years in the. Well proven it yet. So what we say is we can’t give you predictions, but we can give you absolute verified data about trends, about how your network degrades or acts at certain periods in time.
So you can spot patterns and therefore you can then feed it into your model and your strategic planning, your consumption demand model to make informed decisions about whether you need to build desalination plants, whether you need to do pipe repair, whether you need to build more capacity. I
Antoine Walter: think we should have an offline discussion about predictive analysis, because I know that plastic pipes don’t leak.
Even if you’re saying they might a and plastic fittings even less. I mean, I’m joking. It’s just that I have to defend the bits of my business, you know, but at GF piping systems we’ve been working on time-of-flight diffraction and then building an algorithm to try to do that. Predicting fades. Detecting some stuff, insides, a fitting or some stuff inside a pipe, and then saying, okay, that is going to break in so much time with a level of probability.
So I don’t know for the sheeting materials, which are not plastic, but for plastics,
Victoria Edwards: but you need to still build that predictive model on verified data. Absolutely. Otherwise,
Antoine Walter: fully agree. Coming back to what you said about selling something very new in utilities and. I would say even worse, you’re selling something which is very new, but replaces something which is existing.
Because if you come with something which is fully new, then, I mean, it’s a different discussion, but here you’re turning something which used to be a CapEx into an OPEX in a word where CapEx is the king. I mean, a utility can always defend the budgets and then say why they need to do some CapEx. It’s much more difficult for them to come and say, I have to justify my OPEX.
And it’s really something that. Flipped around from the industry, which is the exact opposite. So you’re going into that most difficult environments to build up a startup. You explain us how you were able to sell in weeks. What takes most of this industry years, if not decades, what’s your secret.
Victoria Edwards: Uh, I don’t think it’s a secret.
I think it’s a combination of many things. Isn’t it? Timing, opportunity and belief. We have what we call the Fido face. We will do what we say we can do. I think a lot of startups make the decision. They go in and say, well, you’ve developed this thing and you’ve got to use it because it’s better. Why are you being so stupid with what you’ve been doing before?
I said, well, that’s not wrong. It’s not a stupid decision. It’s probably the best decision they made at that period in time to deliver or solve a problem or a need that they had. Can we enhance. So we came at it very much at the model of saying, actually you’ve made these CapEx investments. Fido is an overlay to give you a better ROI on those decisions that you’ve made.
And I think by having that and being able to demonstrate that by turning each leak into an asset and tracking it through, it’s a logical decision for them then to have in the fact that it’s. Um, where they haven’t made the cap X because they don’t have a Catholics budget because, you know, outside the UK, then the, actually the conversations are easier to have.
So I think that’s why we’ve, we’ve got that traction because we’re giving a value add rather than replay. I’m a
Antoine Walter: firm believer in that concept of the AR funnel. You know, this video with Dave Bachler, which is saying our, and then explaining everything by the acronym. So this acquisition activation revenue, retention, and referral, and just to switch from the aide to the RSO, starting with the revenue, there’s this wall effect, this activation trigger.
What is your activation trigger at Fido? What is the thing in your offering where customers look at that? Oh, I got it. That’s what I need. That’s the value in it.
Victoria Edwards: It’s visual. And, um, for the people that have already got hardware, we just put up a map of all the alarms they’ve had that day, which may be. And go, you think you’ve got a team of 110 engineers out there trying to find those 72% of them.
Aren’t leaks look, fighters told you there’s only six of which three of them are large. They’re the ones you need to focus on. So what it does, it’s very clear return in terms of driving. Cause the only thing you can affect in leakages run time and volumetrics and fighter AI with an O-ring accuracy delivers volume metrics and the ability to hit run.
To affect it. And that’s the simple they get that we then also can do all the other impacts or what other impact does it, does it have means you dig less means you don’t have to do nighttime working. It means that you know, where you should be doing your capital replacement programs more efficiently. It also makes sure that the repairs are done properly.
That’s, you know, in the leak that was there. It has been fixed as, you know, the large leaking. It’s not the small one, which is the stock tap, which is the easy one to fix. There’s a lot more insight. And I think that it’s that movement. We call it the wheel of fortune. We touched the business and about five or six different areas, finding the leak is just the start point.
But I think it’s that ability. That’s where we set. And that’s why we’ve got so much traction.
Antoine Walter: And if that’s just a starting point, is there a portion in you which things, why not building something larger and going into the next steps?
Victoria Edwards: Oh, gosh, yeah, that that’s outside my remit, but my, uh, the guys here, um, the guys and girls have just have the most phenomenal brains and that we create an atmosphere where we accept failure.
Right. And, and I say to some of the, some of the brightest that we’ve got, I say to them, look, your first job is not meant to be like this. It’s meant to be grunt work where you do the same thing as a bit boring. Do the same thing. The only thing you have to worry about at Fido is who is going to play you in the Netflix film..
Right. It go and think we have some wonderful ideas that are coming out. And, but we’re passionate about this water scarcity. We believe that we are here to do a job and that is our job. So in terms of the next things that are happening, we have to make a change, a fundamental change in how people perceive water.
I’m a great believer in giving community agent. You’ve got to give the community agency is then it’s being affected. It’s then it’s going to run out for fix. Then that’s going to have, you know, the bills increasing, how do you help them understand what’s happening beneath their feet? So we’re looking at how do we get communities involved to understand.
The value of water, democratize water, how they can get involved in a gamification way to understand leakage of water in their community and how they can be empowered to work with a local utility company to solve that problem. So we’ve got that it’s in the horizon, we’re playing around with some wonderful ideas.
So the only point we need to human now is at the point of dig and people are saying to us, well, come on fight. I don’t want to have to do. Surely you can do it from the inside. So watch this space.
Antoine Walter: So that’s a hint at what you might be building. Okay. I’m sorry. But I have to ask you this closing question for the deep dive.
I know you’ve been hearing it quite a lot, but what does success look like for you in let’s say five years?
Victoria Edwards: I think success is quite simply that Fido is given the. Platform and the space to be able to deliver to anybody regardless of network maturity, regardless of workforce, that actually Fido can be to leakage what Google is to search by being non-proprietary by being open data and by being collaborative.
And I’d like to get, give the world the opportunity to tackle this global problem before discussing.
Antoine Walter: I think that was the perfect conclusion for the deep dive. I propose you to switch to the rapid fire questions.
Rapid fire questions:
Antoine Walter: So in that last question, I try to keep the questions short and you have to pledge to keep the answers short. Of course I’m never cutting the microphone. My first question is what is the most exciting project you’ve been working on?
Victoria Edwards: Community agency making leakage, visible and fixable tool.
Antoine Walter: Can you name one thing that you’ve learned the hard way?
Victoria Edwards: Don’t take your eye off the ball.
Antoine Walter: Is there something you are doing today in your job that you will not be doing in 10 years?
Victoria Edwards: Uh, I’m trying to convince the world that solving leakage is difficult and unachievable.
Antoine Walter: Do you still have to do that today?
If you have a proportion of people, what would be. 50 50, 80% of the people are not convinced that leak detection is something which is easy. I’m sorry, I’m sidetracking you, but I’m just
Victoria Edwards: curious. We get caught in the minutia of economic levels of leakage. Well, let’s just fix all the large leaks versed in that, then that stuff like discussion.
But yes, I think over 50%, I still thinking it’s problematic.
Antoine Walter: What is the trend to watch out for in the water sector?
Victoria Edwards: Oh, for all the wrong reasons desalination. And why would you say. But because that that’s the utility, you know, the answer to everything at desalination obstruction. Well, we haven’t got enough water now, so we’ll just find some more than most expensive way possible.
Let’s fix what you’ve already produced. Let’s stop losing it. Then. That’s how that conversation. If you
Antoine Walter: were a word political leader, what would be your first action to influence the fate of the words? Water challenges?
Victoria Edwards: Well, you can make everybody give away the first 50 liters for free. And then they can start charging afterwards to each household per day or some limits.
It will soon have a global conversation about protecting the water they’ve already produced. Um, and one global body that actually, there’s not one global voice about water is the world’s most precious resource. And we have nothing that’s cohesive and coherent that has teeth.
Antoine Walter: It’s very interesting. The answer that you give here, because it’s a Cintas.
Many answers I’ve got here. I had Minnie on that microphone explaining that the climate change has this zero carbon net zero, and that is the rally call for everybody. And we don’t have that in water. And then the other part of it, the first 50 liters, which are free is really something. There are two teams on that microphone.
One team is we shall charge much more for water and the other team is the essential needs of water. Be covered for free. And then you should be walking the talk and adding some more layers of protection for the next liters, which are not supposed
Victoria Edwards: to be absolutely. I mean, everybody’s jumped on the net zero bandwagon, the politicians, well, actually we’re facing days zero.
We won’t even get to our net zero goals. Day zero is the day the world runs out of water. Happened in Johannesburg, back in April, 2018, it’s happening all over the world by 2030, 40% of will. Outstrip demand. Outstrips supply is a question we have to have, and we need a big solution to achieve that, to stop these ever happening.
Antoine Walter: Well, do you have someone to recommend me that should definitely invite on that microphone as soon as possible?
Victoria Edwards: I think you’ve got to find somebody that’s led the industrial revolution in another state. So, you know, if they were alive, you’d get, you’d have a mixture of Joan, of arc and Basil’s yet, wouldn’t you man, who invented the service in general arc?
Who said, you know, this was what I was put on us to do. This is my faith. This is my belief. I’m not going to stray from it at the moment. We’ve got to find somebody that can treat water as the most essential resource on us that we’ve got. And then we’ll take it by its horns. So we need to freeze. Well,
Antoine Walter: Victoria, it’s been amazing discussing with you over this small hardware.
If people want to follow up with you, where shall I redirect them?
Victoria Edwards: Um, obviously I do quite a few social posts on my LinkedIn. Um, the company, LinkedIn, there was a subscription LinkedIn newsletter at which I’d invite people to come and read. I’d like to start a conversation, some really strong opinions.
Let’s have those discussions. Let’s keep walk to where it should be, which is at the forefront of everybody’s mind not chasing, you know, this technology that can’t be implemented without water. So treat it with a precious resource that is.
Antoine Walter: Well, it was a fascinating conversation with my ends. I’m sorry for the Lehman questions, but really it was, it’s a, it’s a fascinating field and yeah, thanks a lot for sharing and for the openness, it was really a pleasure.
Victoria Edwards: Thank you very much.