From Vision to Reality: The Senseye Product Evolution - with Adam Poole
Show notes
Welcome to the Trend Detection podcast, brought to you by Senseye Predictive Maintenance – which gives you visibility and insights into all your assets, from single machines to full plants to help you reduce downtime, increase knowledge sharing and accelerate digital transformation across your organization.
You can find out more about how Senseye Predictive Maintenance can reduce unplanned downtime and contribute towards improved sustainability within your manufacturing plants, by visiting: www.siemens.com/senseye-predictive-maintenance
Transcript
We never build anything in a bubble. I think our biggest advantage is that we build really quickly. We've got, so in that software development world you talk about agile development, whereas you do things very quickly, you respond to things that come up immediately and fix things rather than planning it all ahead of time. Welcome to Eye Opener, powered by Siemens. And now your host, Gugan Sanga. Welcome to trend detection powered by Siemens. In today's episode, we explore the redesigned dashboard, reporting capabilities, and new features such as knowledgeBuilder, which has shaped the user interface over the last 18 months. On today's episode, I'm delighted to be joined by Adam Paul, um, who is going to be discussing this topic with me. And if you could start off by introducing yourself to our audience, that'd be great. Yeah, I'm Adam Poole. I'm product design manager at Sensai. I've been here for nearly 8 years now. Um, before that, I worked for 17 years in aerospace. So, it's quite a lot of experience of all kinds of things now and um, Even in the last year, I mean, things have changed in the design world so much with all the G AI stuff that's come on board. So it's been an exciting time for us. Amazing. So thank you so much again for joining me on today's episode. And let's get straight into this topic. So I know from a sense side predictive maintenance perspective, there's been a lot of changes which have been made, a lot of advancements to the application, um. And I guess the first question I have is around the dashboard and the user interface. So how has that changed? Well, it's, it's wouldn't be short to say that in the last two years we've completely redesigned the whole application, like top to bottom, with the most change happening in the last year, actually, it's kind of, it's kind of crazy and our pace of development is so fast. Um, and a big part of that. is having more users, more users across many different industries across the whole world, different nationalities, different languages to support. But there's always a common requirement for what we call like reporting. So you've got these management level people, they want, they might use the application only a few minutes a day, but they want a quick place to see the top level things that are of interest to them, the machines that are going to break, uh, imminently, that kind of stuff. So we wanted to put something in front of them. So we put this new homepage, which is like a top level dashboard, um, all of that information's there for them, customizable, um, and they can export it, schedule email, all of that kind of stuff very quickly, um. And in terms of user experience, we had elements of this within the app already, but they were kind of buried. And one key thing of like building a better product is just like make it really simple for people. It's, it's no good having all the features unless they're usable. So if it's 3 clicks away, make it like 2 clicks away or 1 click away, make it really straightforward. Um, and really listen to people what, what people want from you as well, is, is a good thing to do. OK, cool. So you touched on um the topic of just making it simple and making it easier. So I can imagine you've taken into account, you know, user feedback of the application. Um, just leading into my next question focused around You know, can you give a bit more information around what drove those changes to be made? Sure, I mean, we never build anything in a bubble. I think our biggest advantage is that we build really quickly. We've got, so in that software development world, you talk about agile development, whereas You do things very quickly. You respond to things that come up immediately and fix things rather than planning it all ahead of time. In the old days it used to be called Waterfall where you'd, we'll plan everything, we'll design it, we'll make the whole thing ahead of time and then we'll go and build it. And as soon as you do that, you'll find you were wrong. Or technology has changed with beneath you. If we'd made plans a year ago, they would have all been wrong because this little thing called Gen AI came along and changed everything. So We do it like incrementally. We work in like 8 week um cycles. And even within that kind of cycle, we might change what we're doing. So you can kind of pivot onto something different. Um, and a big part of that is listening to what the users say. So there's lots of mechanisms for people to do that. We have a teams on the ground who are talking, you know, working in the factories with people or going to visit them, um, very closely and they'll be telling us exactly what they like and what they don't like. And we capture every single thing. So we've got like a register of like, it's probably thousands of things now. Um, every idea that comes in, we look at and um that forms the shape of what we build next. And we, and also internally we have a culture, we've always had this culture of like encouraging product ideas. We, we have a channel in our, um, Slack called product ideas and anyone, anywhere can put an idea in there. And the first thing about doing that kind of creative exercise is like, there are no bad ideas. So let's just put it down. I mean, the truth is there are bad ideas, but even within a bad idea, there's a germ of a good idea. So, and that you don't even know how that might be used down the line. So it's all there. We read them all and we turn it into and shape it into features that come along and they might not come along for a year or two, but they all get kind of integrated somehow. So we have that. And also I think we're probably unusual as a product team, we're kind of central. So we, we get all the business requirements and you get the customer requirements and we do regular customer calls as well. So we're talking directly to the people who use the software every day, and having an honest conversation about them. And we never mind bad feedback. It doesn't matter to us. It's like, just tell us what you want. Um. And then you kind of look at the consensus of where things are and you identify, ah, we should really be doing this. We're totally missing this thing here, we need to switch over to that and make that happen for people. So I guess the biggest theme is about being flexible, not being rigid in your thinking. Um. I'm probably answering lots of different questions here, but one key bit I think it's sometimes we get people saying, what's your 5 year roadmap? And it's like, it's really hard to have a 5 year roadmap. We might think in terms of like a 3 year or a 1 year. But again, you have to have that flexibility to know that down the line something. Something might, some technology might come along that you don't know about and you've got to be able to jump on that as quickly as possible. Because it could be, you know, the game changer, basically. That's what I was just going to mention. So when it does come to this, we, we just touched around, for example, generative AI and this is a trend now. I say it's a trend because it's part of everyone's day to day activities, whether that's work related or personal. Um, so I'm just interested to know how much of a role that played in when it came, came to implementing some of these features of it, it's been huge. It's been the biggest kind of unanticipated change I, I've ever known, I'd imagine in terms of revolutionary technology, you know, it's huge. Our research team, we have a like a small research team internally who are always looking at what's coming along and they were on this very early and telling us. We need to get on this, we need to get on this. So then as a product team, we'd say, right, you go and, go and test some things, go and prove it to us. And then they go and prove it and make little, little apps on the side to test things out. And then we start bringing those features forward, one thing at a time. And what you find is when you're on the cutting edge of stuff like that, especially in a big company, you're kind of paving the way. So you've got to jump through all the legal hoops as well. You've got to understand, anticipate how people might be using it and do it really incrementally. Um, it's tempting to dive in and just do everything, but I think we've done it the right way. We've done like one feature at a time and they're slowly building it out toward this kind of Where it's becoming even more useful. What we haven't done is made it the whole product. I think the mistake you could make, and some people do is like, oh, I love chat GPT. That's our product now. But We've made it like a sub-feature. It's just a feature of what we're already doing. It, it complements like predictive maintenance. There are things it can do, it's really good at like analyzing data and looking for patterns, analyzing what our users have added and being able to extract useful kind of human readable information out of it, but it's not the whole product. Um, one thing I always thought was like, A prompt is not a product. Like if you just see a prompt and you're expecting people to ask you questions, it's like you haven't really done any work. You're just, you're kind of using Church EBT and If chat GPT went away, you no longer have a product. You know what I mean? Whereas if chat GPT went away for us, we'd still have the product we established over, you know, 10 years. So if we look at chat GBT and Sensei as in the application, I think it's beneficial in ways where it can help to support and track issues with machines and assets. Um, but as you mentioned, you know, it's not something that we're heavily reliant on. Um, we can still have the sensor application because you can still have human interaction with that. Yeah. And we've always been based around, well, it was always based on machine learning, which was the old fashionable name for AI. And then JennyI is kind of the next thing that's come along. But what it's really good at is, you know, we, we're good at, we're good at operating at scale. So like you might have a small number of users in a factory, but they're dealing with so much data, like the amount of data we deal with every day is just unbelievable. And you might only have a team of like 3 maintainers. There's no way you could look at all of that data manually. So you've got the machine learning doing the heavy lifting and then the general on top of that analyzing all the historic stuff and looking for patterns that we previously couldn't have spotted and saying, well, this is happening now and the last time this happened, you did this and the time before that you did this. And then It's encouraging more human input, which is what I've always wanted, is because the more human feedback you get, you know, you, because it's very good at doing kind of sentiment analysis, understanding the tone of what people add, you see that coming back out of it again. And once customers see that, I think you want to add more and more. And we're beginning to see it more and more now with the um, the knowledge building side of things. So that's a term we've come up with is like adding your own. Knowledge and documents and files into the system and then having them interpreted by the generator to help you understand the machines better. And it's all, it's all about kind of customize, it's like the ultimate customization for them, really. It's made us more of a platform, I guess, really. So all within the secure environment. So we're thinking they might have operational procedures that are common across like the whole factory. So in, within the app you can put those documents in there and then it filters through everything, and then when they're looking at a particular machine, they should say, how should I operate this, and it will know that document's there and it will give you a summary back from it. And really importantly, it will always like reference where it's come from, which is a huge thing. Um, you can see people coming reliant on a chat GPT output without knowing. It's provenance, like, where did this come from? They'll just believe it. But we can't do that in an operational world because there might be safe, you know, we have to be very careful with like safety and things like that. We're not going to tell people to do anything that might be dangerous. We're incredibly careful about those things. So it'll always tell you where it's come from. So you've got the operational side, but you've also got, you know, the individual machines, this, you know, you add that information in there rather than having to go and look up the manual. The manual manually, you can, you can just ask it and it'll tell you what, what you need to go and do. So, and that's the real growth, I think that's the next thing we're seeing. And you're seeing these features in lots of software now, it's becoming, these are becoming kind of core feat standard features. It's not really a differentiator for some people because it's just going to become expected. Even like normal day to day office users, you're expected to be able to do these things. So we have to support those technologies now. Um, but then think about how it really works for our product. Um, and we've just come out with those kind of features recently, so people will be able to upload. Whatever files they want and choose to add it to the co-pilot effectively and make it accessible. So it's still in their power to choose what they want to do. Um, and now we're excited to see what they actually go and do with it. Yeah, exactly. Well, this is it. It's aligned with how, um, you know, current trends are with AI and you, you mentioned around the co-pilot, so within Sensei and the application. And it's interesting because, you know, speaking to customers and users, they seem to be more inclined by just asking, you know, the, the co-pilot information about what's going on with their assets. And I think maybe 11 reason for that is because it's so intrusive and it's so quick to get a response. So, yeah, it's very quick and the interesting side again is the human angle. One reason we wanted to do this kind of knowledge side of things was It's a very We noticed a trend that there's aging workforces within our customers. You've got very experienced engineers kind of aging out of the workforce, so they're retiring. And they've got 40, 50 years of knowledge that we need to capture and pass on to the younger generations. And this is like an amazing way to do it. But then the compliment to that is you've got the younger workforce coming on board who don't have that experience. But they know about these technologies. So they understand these new interfaces better than the older workforces. So they, they will start kind of educating the people above them and, and how to use these things. So it's funny because you get that sometimes that resistance when you put like a prompt and you can ask your questions and stuff and they're like, there's a bit of grumpiness there. And yet the younger people would be like, oh yes, this is, I do this all the time, it's fine. And then they'll start passing that back on and then it all starts to kind of cycle around a bit. And now they're asking like, can you do video as well? Can you extract transcripts from video? Can you do um translation and all these kind of things. So it's amazing side of things. And like that localization thing, I said, we've got customers all over the world. And we support, I think 12 languages in the app now, which is a part you kind of sometimes forget about. So the way you write things it becomes important, it needs to work in all those languages. And you almost get it for free with, with, um, that chat GPT integration as well. So it will automatically translate things for you as well, which is a, it's kind of a superpower, you know, it kind of Bridges some barriers. You know, we've got big multinational companies where that can be an issue. We, you know, we're still quite a small team in some ways. You can't necessarily expect to support all those languages, but we can kind of inherently do it now. It's, it's scaled us even more, I think. Exactly. I think that's a huge benefit as well. If you think about, you know, all the industries that we work in and, um, you know, like Sensi is on a global scale. Um, it's quite beneficial to have that because we have to bear in mind that English is not someone's first language. So having that availability within the application does make it easier because you've got this, you know, flexibility, going back to this flexibility and understanding your machines and assets and having it in different languages definitely helps. What's amazing is we got, you got like the formal language which is in the documents. And then we've got the informal language which is written by users. And you know, that might be formal, but sometimes you, I've seen like slang terms for machines, um creep in because that's what people use. And then the uh co-pilot will use those to learn that that's the term for that machine and use it back to people. And that's kind of fascinating. So when users notice that they're like, oh my God, it's using the term I use for that machine. And that's kind of an unanticipated like bonus. But again, it, it's all about encouraging people to Use it as much as they can and really, really get the benefits of it. So they kind of make bridge that gap in their mind. They like, oh this is. This is really listening to me and understanding what we're doing. Well, that's it. It's just that like you've mentioned, just listening to me, so making it more of a personalized experience. I feel like, you know, like I said, chat GBT just with anything has pros and cons. Um, but within this application, you know, referring to key terms or slang terms and for the co-pilot to then pick that up. You kind of think to yourself, oh, hang on, it's actually listening to what I'm saying and it's recording the information that I've shared. Yeah, and historically that that's really complicated because you might have a really strict like naming conventions for your machines, which, you know, companies do. So you might have like a 20 digit alphanumerical name plus something like this and plus something on the end, but we also call it this. And it'll understand all of that. And, but that's the kind of institutional knowledge that you would have otherwise lost, um, from those people aging out of the workforce. But if you kind of capture it in place, you can, you can pass it on and understand it. So it's just, those are unanticipated things, I think that's the real um benefits to it. But a key part is like it's not going to do everything. We're just doing like feature at a time because the one thing you see is when you see like a prompt, like if you could, I could ask it anything, but that you can also be paralyzed by that. Like, I don't know what, what can I ask her? I don't know what to ask it. So, and if you ask it something and then it kind of can't answer it, then you've lost them. So in terms of user experience, we, we've been very careful to manage that. You know, the broader chat GPT will just answer everything obviously, but within our use case in protective maintenance, obviously we, we've turned off the less relevant things, um, and made it very specific to the tasks in front of us. We don't want people going off and doing things they shouldn't do, but also mindful of the secure environment we need to operate in. But again, they need to know what, what I can ask it and what response I can get. So that's what we're kind of managing and then eventually build out all those features to a We call it an app wide co-pilot where it will know what you were looking at, it will know everything you've done and all that kind of stuff, but it'll take a little while. Yeah, …