From AI Insights to Autonomous Action

Hello, everyone. Welcome. Thank you for joining our five two webinar on integrations and the connected smart factory. I'm Andrew Murray, an AI product manager on our Spot and Orbit product teams. Today, we're gonna be talking to you guys about the connected smart factory and what the heck that even means. We're gonna talk about our upcoming five two feature launch, which we're really excited about and is jam packed with a ton of features. We're also gonna be talking about how customers are using Spot and Orbit today in order to integrate into their facility for a much deeper and richer experience. If you guys have any questions, please feel free to leave them in the chat. We have moderators who will be answering them in real time, and we'll also get to a few of them at the end. I'm joined today by two of my favorite colleagues, Ashley and Ben. Hi there. My name is Ashley Armandi. I manage our enterprise sales engineering team here at Boston Dynamics, and I work with a variety of our different customers across many different industries to help support their Spot journey. Hey, everyone. I'm Ben Gowoski, and I work on Andrew's team as a product manager supporting the Orbit and Spot Enterprise platform. Awesome. Thank you guys for joining me. I thought I'd start us off with something that we've probably all heard our customers express to us, which is this concept of tool fatigue. I think it's very true of any product today that customers and facilities are using tons of different platforms to manage what's going on on their site. And then we come and really we're introducing two new products, Spot and Orbit. And they want to know how is this going to work in our facility? Is this going to be another isolated, siloed platform that we're going to have to log into? And I think we've done we've really taken that to heart. And we've tried to make Orbit a bidirectional tool, Something that can take the data from our robots, plug into their systems, but also vice versa, take data from their systems and potentially have our robots do something interesting with that information. So Ashley, I'm interested to hear about how our customers are maybe using our products today to build and embark on this journey with us on this kind of smart connected factory. Yeah, absolutely. Our customers, the most successful customers that we work with, they're not using Spot and Orbit as an isolated tool. They're really making that interconnected journey and starting that data flywheel. And so one of the most compelling customer journeys that I think about when I think about that connectedness is the customers in our semiconductor space. They often come to us with a pain point of you think about all this peripheral equipment that's there, and it's not economical to outfit those with static sensors all over the place. The journey that we go on with them is we can bring the sensors to those equipment, and then we can enable them with of the data that's going on about all of those different assets that they can connect to their existing systems. And so ultimately, what they use that for is then to create these more not only preventative maintenance programs, but predictive maintenance programs with the use of all of this data that they're able to capture and integrate into their existing systems. Yeah. So you're talking about data in and data out and creating that data flywheel that our customers like. But you need to trust that data, and you need to know you have enough of it. Too much, you can't find the signal in noise. So Andrew, you wanna talk about some of the new five point two features and how we can kind of get there to build that flywheel for our customers? One hundred percent. Yeah. Five point two is really building on the ethos of a smart connected factory, but really, it's building on this ethos of Spot semantically understanding a customer's space and telling them what they need to know when they need to know that information. So how do you do that? Right? They want customers today expect Spot to be able to walk around the facility. It looks at this motor every single day. It inspects it and alerts me when it's too hot. Why can't Spot go around and say, that's a motor and it's in an anomalous state right now? We're not gonna get all the way there in five two, but we're building the foundational blocks and the building blocks that are necessary to build that blue sky vision of semantically understanding your space. To do that, we've started by integrating Gemini into our product suite, which has been fantastic. Not only are we getting much richer insights on the inspections that we perform today, but we've been able to add new inspections to assets that we'd never supported before. Things like gauges, sight glasses, these highly industrial assets that, Gemini is giving us much richer information on those assets and the alerts that are associated with it. We're also able to add modalities to our AI suite. We can do AI inspections on video, which has been something that has been highly requested. Right? We can take a look at leaking pipes, leaking equipment, flashing safety indicator lights. My personal favorite, conveyor belt sag, conveyor belt damage, so that your entire production line doesn't go down. Right? These are things that incorporating Gemini into our suite has just made our AI tool much more sophisticated, which is great. We're also in introducing a new concept, I have coined encounters, which is really building that blue sky vision of semantically understanding the world. Spot sees a puddle on the ground. Spot sees a person in an area they shouldn't be in. Immediately flag an anomaly that that is something that should not be occurring on my site. If I'm a factory worker, it's something that I can immediately see and say, this is not right. I have to go and fix it. So now Spot is understanding these things and being able to tell you, hey, Something's going on in your site. You should go take care of it. We're also making Orbit more asset centric, which is really exciting. If, again, I'm a facility manager, I don't necessarily care if Spot's performing the fortieth thermal inspection. Right? But I do wanna know if my motor bearing on motor five of my conveyor line is in an anomalous state or if it's overheating. And so we've switched how we convey this information to customers within Orbit so that everything is much more asset centric. And that plays a lot nicer with your asset management tools, which you utilize to manage your facility every single day. Getting outside of Orbit, we're also introducing a new gas payload, which is really exciting. We can now sense twenty different gases, real stream data just like all of our other payloads, and this is flying off the charts. So we're really excited about this new payload. And all of these features not only make Spot and Orbit smarter, but it's part of building this really sophisticated data set that you can then incorporate into your factory, into your other systems, and build this smart connected factory that we're going to talk about more today. Yeah, that gas sensor is huge. We have customers that have the gas sensor, and they have the robot, and they've been screaming from the rooftops, Why can't we have this in one place? That's right. And now we have that for them. Yep. So it's a huge value unlock for those customers. And again, tying that data into one place, and then feeding it all back up to your systems that have been there for a long time. You're not replacing anything new. You're not investing in this kind of like new silo you have to manage. It's coming back right into the system that's been there the whole time. So let's talk more about the platform features. So one of the on prem kind of customer ask we've had is to have everything funnel through Orbit, and that means robot logs. So one of the huge things we did in five point two is funneling spot robot logs through Orbit. So now you can have robot logs and Orbit logs all in one place. And this may not be painful if you just have one robot, but if you have three, five, or fifty robots, this is unmanageable, and you almost have to have a full time product manager monitoring this stuff and making sure you can see what is happening, getting these logs over to support if anything goes wrong, looking back in time. So this is a huge relief for those customers. Some of the features that I'm a little bit more excited about is we have sensor integration. So what this means is you can integrate PLCs, thermal sensors into Orbit. And then what happens then? So now you have a thermal sensor connected to orbit. You can set a threshold with orbit decision logic, and you can say, hey, when this sensor goes above this temperature, immediately send a robot to go inspect that asset. So while you have one sensor saying there's something probably wrong, immediately send the robot out, get some thermal scans, get a secondary check, and this is huge. You don't have to wait for the next mission. You can immediately have that robot go out. So if the sensor triggers on Sunday, and you weren't gonna send a robot out until Tuesday, this gets that in real time. It goes right to the machine, does the inspection, comes back, and that is huge. You don't have to have a schedule for it. It is just on the sensor schedule. Yeah. This is really great because everything I was hitting on prior is taking data from Spot, pulling it into a system, getting more data into your systems that you already have. But this is really exciting because now customers can take the systems that they already have, have that feed into Orbit, and then have our robots, like Spot, perform an action. And I was actually at a customer site the other day. We were talking about this feature, and he drives two hours to get on-site. If there is an alert that gets triggered by their system and he has to go there just to find out that it's not actually a critical alert he had to deal with, he's now wasted two hours of his day. Right? Huge bummer. So if Spot can actually get that notification, go out, perform a first primary check, and then say, hey, you actually don't have to come on-site. I've checked this out. You're fine. Or yes, this is critical. Please come onto the site and deal with this issue. This is a really, really cool feature where we talk about that bidirectional communication with Orbit. We're taking data from these systems to now have our robots go and perform actions. So Andrew, a lot of those AI inspection features you're talking about, I've seen a lot of ROI capture for customers on this, especially in the EHS space. There was one customer that had almost half a million dollars in liabilities that they found just from these inspections. Those would have been fines. And then we had another customer find an oil level sight glass that just had a little bit of a discrepancy. They had a maintenance engineer go check it, save them thirty thousand dollars. This was all within the first month of having the robot. So these are just small stories that add up. You don't just have to use the robot for EHS, or for asset maintenance, or for security. You can use this robot for many things twenty fourseven in your facility, and find ROI compounding across those different verticals. One hundred percent. And this is something that we talk a lot about with the AI tool. We just had a customer come on-site literally the other day where we deployed it in this vertical with this team, and they said, we introduced the AI tool, and now security wants to come talk to us. Safety wants to come talk to us. Our executive team is coming to talk to us, and they want to apply the AI tool across all of our different teams now, with Spot being the data collection platform, and I use the AI to get more insights out of it. Ashley, have you heard that across any of the customers that you engage with? Absolutely. I think this is that concept of that one robot multiple jobs story that we always kind of talk through at a lot a variety of different customers that we talk with. And one of the things that comes to mind when I talk about that is the mining industry. The heavy Emshaw fines for things like housekeeping that are easily findable but maybe not done in a consistent way with humans in the loop right now. And so Spot, as it's going around and doing those typical industrial inspection checks like the thermal auditing of their equipment and looking for pneumatic leaks. It can also go and look for housekeeping items and alert using that Gemini AI to the teams that, hey, this is a safety risk. And it can do things like that. It can look at things on the ground. It can also look for things like not finger safe grading that are on the sides of these equipment. And so they're solving a problem with not getting full coverage on their assets, but they're also solving this other problem that's adding up in these Emsha fines and realizing real value from limiting those. Hundred percent. One of the things, you know, we hear I heard a lot about before joining Boston Dynamics is Boston Dynamics robots are general purpose. And then you get here, and you see, holy crap. We have robots in textiles. We have dog food customers. We have customers who are paper and pulp. We're everywhere. And these tools that we have been building allows all of these different customers with all of these different value props to deploy Spot and use it in their facilities and get actual value out of our Orbit tool. So it's incredible to see. Ben, I'm interested you know, one of the things that we've talked a lot about is ingesting data into these systems. Can you maybe talk to me a little bit more about pulling data out of those systems and having Spot perform more of these tasks? Yeah, definitely. So one thing that we've invested a little bit more in is our Orbit API with this release. So you'll see there's a v zero and now we have a v one API. And what that really means is we've just made it more reliable and scalable, and we've designed it with purpose to integrate with third party systems. So you don't have to come to the Orbit UI for everything. As long as you set up this plumbing in the right way, you can get all this data fed into your MES, your SCADA systems right away. So you can get all your inspection data fed into there. And with the new asset data platform that we have, this new kind of architecture, we can meet you where your asset hierarchy is, and make that process painless. And then Ashley and team and our field engineers can work with you during setup to make sure we map those assets together. And let me just tell you a little story about how this can kind of like all tie together with sensor integration and our APIs. So let's say you already have a water pump with a sensor. That sensor goes off and it's connected to orbit, and it triggers an alert to say, We have a leak. You immediately send Spot to go out to where that leak area is, but along the way, it encounters a puddle. So like Andrew was saying, now we have this encounter. We are gonna immediately send a webhook alert, push that to the customer. That can be a text, an email, and it says, oh crap, we have water on the floor. We need to do something about it. So then that customer can then go through the normal maintenance process and react. But then Spot's going to continue going along, and it's going to do its inspections, and it's going to push that other data back up into orbit. And then what you can do if you need to do the report, have your system pull all those images, any thermal scans, image scans, and then you have a full report of, hey, this system alerted us. Spot confirmed it and texted me that there is an ongoing leak and I sent a maintenance engineer out there. And then I used the API to pull all the other inspection data from the previous day, and then the current day to see what happened and when, all through this like new API. So it's a huge value unlock. And we meet you where you are with the current systems, so you don't always have to be fumbling with the Orbit UI if you don't need to. Yeah. If I'm a facility manager and, you know, none of us are engineers, probably listening and thinking, oh my god, this sounds so complicated to set up. Can you maybe talk about, like, who this is meant for? How do you build these APIs and integrations? Like, how does someone get into Orbit and actually, go to the process of having Spot then go and perform these if then type of decisions on-site? Definitely. So for sensor integration, it's probably a little more involved for just the average maintenance engineer. You're going to need a network professional, an IT professional, or a contractor that you have on-site. But if you're already working with these MES and SCADA systems that are pulling PLCs, seeing when lines are on and off, you most likely have that at your plant. Once you do that initial kind of legwork to get there, then it is just kind of selecting some switches, doing if this, then that logic like Andrew was saying, and then saying, what do you want the robot to do when you see this alert? And it does not interrupt your existing mission flow and schedules. All it does is it works with it. It works around it, and you can still have your day to day operations. But if something just alerts out of the sudden, immediately goes off and does that. And then we can kind of work back to the webhook integration that pushes all these alerts to email and text. And that's what we've had since five point zero, and has been really well adopted with our customers. Yeah. I hear all the time different integration type of tasks that customers are asking us for. Ashley, do you have any examples maybe you can run us through of customers asking for, I want Spot to get this information and go do this thing, or I want Spot to collect this data and plug it into my system? Like, have you heard that at any of your sites? Absolutely, yeah. One of the most common industries that I find that out in is food and beverage manufacturing. And so when I go and I talk to these food and beverage manufacturers, a lot of these processes that they run are batch processes. And so if we use a traditional method of having Spot go out and look at a variety of different drive components, we don't know if those are going to be on and off based off of demand or downstream feedback from other process units. And so one thing that we can do with a more robust API, even better, is to connect to those SCADA systems and control centers for different pieces of equipment. And they can trigger Spot to say, hey, go out and inspect this piece of equipment. It's on now. What that ends up having additional benefit for is data hygiene. Because now you don't have to have that additional step of, sure, Spot is gathering my data, but I have to go clean it up afterwards. Because now you're doing the front end work to make sure that the data that Spot is capturing and you're logging and trending is already clean and ready to review. One thing I heard the other day that was really interesting is, you know, these sites, some of these sites that we go on are seasonal. Maybe they're not making their product in the summer, but Spot's still going out and performing inspections, and it goes to a motor that's not on. And it says, hey. The motor's not on. They go, well, we know that. Right? We're not running anything on this side of our production site of facility right now. So being able to actually say, hey. This equipment isn't on right now. Don't perform this inspection. Or this is this is on. Go and perform something over here. It just gives customers the ability to more dynamically control what's going on on their site, which is really important. Yeah. One thing that Ashley mentioned too is that data hygiene is really important. And one thing we've mentioned time and time again, and you probably hammer home with all of our customers, is having a data strategy is huge. You need to make sure that you are investing in a data strategy if you want to be successful with your current operation with or without Spot, And going into the future of AI, if you are already investing in AI agents, probably know this pain. It could be feeding you a lot of AI slop. You don't know what to believe. And we are trying to make this Spot and Orbit integration easier and not add a due burden onto your existing kind of data strategy operation. So just a PSA, invest in that. It is critical. For a food and beverage manufacturer that I had worked closely with, we talked right off the bat before we even talked about what Spot would be doing in the facility, where that data was going to go. And so what we talked about was this they were transitioning into SAP from an existing work order system that they had been using. And they had new asset tags that were going to be put on all these different pieces of equipment that Spot was going to And so right off the bat, we knew that we were going to end up piping all of the data that Spot is going to capture into SAP. In conjunction with that, they have all these other sensors that are already grabbing data from existing equipment. They use a historian of VivaPi, which a lot of our customers talk about. That was also another piece to their data integration strategy onto where they wanted this data to live and in a place that was going to be useful for them, and tie into all the other existing systems that they use. Yeah. That's one of actually the most cookie cutter kind of examples we've seen, and we're really trying to build Orbit in a way to make this easier to get to your SAP, get to your Aviva. And it doesn't have to be those brand names. It can be a homegrown system, too. A lot of our customers, more innovative, have their own software teams that write these middleware Python scripts to get this data. And we play nice with all of it. And we're going to work more to get more examples published, and work with training, and have more materials out there. But we really want to push to say, if you are more connected, you're going to get more value out of this, and you're going to see that ROI savings we talked about earlier. And one of the things I'm really excited about is in the age of agents and AI, as buzzword y as it may be right now, people are actually using this in their facilities now. So one thing that we did in five point two is we built on top of our API and we integrated something called MCP or Model Context Protocol. What this is, it is just a protocol from Anthropic. And what it does is it creates tool calls around your API calls. And that's just a fancy way of saying, I can interact with Orbit through an LLM. So if your SAP or your internal system of Gemini is already approved through your IT network, all you have to do is connect it to Orbit and you can start asking questions. So there was a water treatment customer that I had. And they just said, Hey, we want to use agents for some stuff. And I'm like, Okay, tell me the problem. And they said, Well, when something happens, we want to get notified immediately and then have our agent take care of that. You don't want to do that today. You want that to be deterministic. And using the API, using the webhooks, making a rigid kind of connection when you need milliseconds of latency to make sure if it's a safety critical or system critical alert, you can already do that with our API webhooks. However, once that data is into the system, that's where your agent can kind of start querying and saying, hey, I saw this trend from this data and I see it over here in this other system. Let me pull from Orbit, from SAP, and do a little analysis and then give you a little summary. And then you can converse with it. You know, you have your chatbot, your agent interface, whatever you want to call it. Now you can ask questions about your site through your agent that you already have approved in your system, and really start to converse with Orbit, converse with Spot. And you can even ask Spot to go on a dynamic mission to say, Hey Spot, I haven't seen inspection of motor pump three in a long time. What's up with that? And then you can just have it go check just from a prompt. So while you don't need that today, I would encourage you to start just looking into it, what that means, and then you'll hear the word physical AI. This is us embodying a third party agent into Orbit and Spot and creating that physical AI loop that the industry is so excited about. And we hope it will provide a lot of ROI in the near future. Absolutely. You know, people see Atlas and they see it performing backflips and they think that we're just building things for fun. And we are real product managers. We're building real things with real customers. So as you see these tools and you interact with these tools and you're building a sophisticated data flywheel, we can actually take that data and that information and those use cases and build real product solutions that you can implement, maybe not today, but certainly tomorrow. So these are the types of things that we want you guys to start engaging with so that we can start to build the use cases that support how you want to, you know, take this data and actually perform actions with Spot and Atlas and Stretch. Awesome. This is probably a great place for us to pause and we're gonna take a break to answer any questions that you guys might have. Feel free to write them in the chat and we'll answer them when we get back. Spot does visual along with that. There's AI related visual inspection. So for example, fire extinguisher will take a photo of where the fire extinguisher should be and ask a basic AI question. Is there a fire extinguisher visible? The answer is obviously yes or no. The boards as well for Lotto. If there's any missing padlocks, he can set an alert and trigger visor a padlock missing. It can also inspect digital meters level of lubrication tanks through AI. Alright. Welcome back, everyone. We have a ton of questions funneling in. Some of them we're not gonna get to, so bear with us. We have moderators in the chat. So if your questions don't get answered by us, hopefully, will get answered, by one of those moderators. Alright, guys. This one has come up a ton. Ben, earlier, you were talking about how integrations might take a little bit of upfront work and some knowledge on how you get, Orbit integrating data into your systems and out of those systems as well. So can you guys talk about, is there any support that Boston Dynamics offers if maybe I don't have the people on-site to do this or I don't have an engineer who can set these systems? Yeah. Absolutely. There's an option specifically for integrating with work order systems that we have that's a low code, essentially no code option to get that done. And I've talked to variety of different customers about this, but what comes to mind is customers in the cement industry. Generally a very risk averse industry, generally don't have a lot of these robust resources on-site that they can just tap into, right? And so what this enables them to do is have cut out the pen to paper kind of monitoring. And now you have the resources just with the maintenance technicians team to go in there and map what information needs to be sent that's captured from Spot into their existing work order system, whether that be SAP Maximo or whatever they're using. Yeah. So that's great for the work orders, but then we get into the APIs and the MCP layer. And what sort of options do customers have if then they don't have those software teams in house for maybe some of not necessarily innovative, but just, you know, larger kind of established companies, like a large food brand, like you were saying earlier, that you worked with? Yeah. No, definitely. If there's no resources in house to be able to commit engineering resources to put something together, we have a couple of different options that our team can help out with to get things done like that. One of them is our integration services. So we have a whole team of project managers, field application engineers who are able to come on-site and commission Spot. And that's part of our standard package. But they also can sell some extra FAE time is what we call it. So they can go on-site and work with you on some of these projects. Now if it's a specific scoped kind of project where we want to do something that's a little bit more complex, we do have a group within Boss Dynamics that's called Boss Dynamics Consulting. And so that group can come in, project intake the project that needs to get done, and really lay out a specific scope of work and achieve that. And so those are the services that we have to enable some of our customers who may not have in house resources to commit to get to the point where we're really in the connected factory space. Yep. We've seen some really cool projects come out of that as well. So definitely use those services if you can. Absolutely. So we've talked a bit about Gemini, and I've seen a lot of questions about what does it actually do. Andrew, do you want to talk us through some of the Gemini enabled features we have with Orbit and Spot today? Yeah, absolutely. You know, our AI tool, it really is not a specific model, right? It is a suite of AI models. And the reason we do that is we've learned that there's not really one model to rule them all. You guys have tons of different applications, like we said earlier. We're in a variety of different verticals that we've been deployed in. And so we need different models to answer different questions that you guys have. And so we've incorporated Gemini into that suite of models, and we've seen tons of performance improvements. So not just being able to do more with the current inspections you have today, but also, like I mentioned earlier, adding modalities. So Gemini gives us the ability to take in video data. We can take in other modality data and start to do more with that as well. So we're really just adding another sophisticated model and a very sophisticated model into our suite. We have the best roboticists with Boston Dynamics now partnered with the best AI engineers with Gemini, and we're able to combine those and marry those into a very sophisticated product, which is Spot In Orbit. One of the things we talked about earlier is Spot going out and seeing a puddle. This has been a question that few people have asked, is if Spot sees the puddle, can it then go and redirect itself around the puddle? That is currently not something that Spot can do. It's something that Spot needs to collect that information, send it to orbit, and interpret that information to then provide that alert. So Spot might not necessarily be seeing that in real time to then create a decision or determine to do something after it sees that thing. And I think we have a very similar question here, which is asking about that rule decision and integration logic. So Ben, do you want to maybe touch on that? Yeah. So the question is, is Spot actually talking to the machines while it's out on a mission? And the answer today is no. We are going to get there, I promise you. But the reality is a lot of our customer sites do not have robust network coverage right now. So we can't always guarantee if Spot goes up to a conveyor belt, it needs to ask, are you on? Should I take this inspection? We can't guarantee that that network stability is there. But you don't always need that in real time because your systems kind of go on a schedule that you've already set and you know about it. So Spot needs to know before it leaves the dock, it'll say, hey, is conveyor one, two, three on? And it'll check it through Orbit, and then say, okay, before going on this mission, I'm gonna see is this machine on? Yes, it is. Alright. Let's do an inspection. Is it off? Alright. Let's skip that inspection today. So it is super powerful still, but we will also get there to do a little bit more kind of like on mission decision logic, but we really need to get that network stability with the customer there first before we can broach that next kind of wave of encounters and decision rules. Awesome. I think that's all the time we have for questions today. Thank you guys both for joining me. If you have any further questions, please go to boston dynamics dot com, fill out our support page, or there's a QR code at the end of this that you can also use. Thank you so much.