Carl Coken and Sandeep Modhvadia from Acuity Brands discuss new tools such as digital twins, machine learning and open platforms that bridge the gap between legacy solutions and emerging technologies, to reduce emissions and operational costs for these huge contributors to climate change. This episode covers how building managers can integrate automated control, organizational alignment, and sustainability reporting to fully address energy and resource allocation - from lighting, HVAC, refrigeration, sustainability programs, and more.
Carl Coken and Sandeep Modhvadia from Acuity Brands discuss new tools such as digital twins, machine learning and open platforms that bridge the gap between legacy solutions and emerging technologies, to reduce emissions and operational costs for these huge contributors to climate change. This episode covers how building managers can integrate automated control, organizational alignment, and sustainability reporting to fully address energy and resource allocation - from lighting, HVAC, refrigeration, sustainability programs, and more.
Resources Mentioned:
Sandeep: [00:00:00] You know, I think there's a lot of lessons that we can learn from legacy IT that we're seeing as the buildings industry moves from more analog to digital and being from more disconnected to connected again, this is goodness. Like I look at this with eyes open and say, well, actually there's a ton of best practice we can learn from.
And we're all in the same boat, and we're all heading to the same place.
Host: Welcome to Connected FM, a podcast connecting you to the latest insights, tools, and resources to help you succeed in facility management. This podcast is brought to you by IFMA, the leading professional association for facility managers. If you are ready to grow your network and advance in your career, go to IFMA.
org to get started. Today, we're joined by Carl Coken, VP of Atrius Engineering at Acuity Brands and Sandeep Modhvadia, VP of Product at Acuity Brands. Together they discuss new tools such as digital twins, machine learning, and open platforms that bridge the gap between legacy solutions and emerging technologies.
They also delve into how building managers can integrate automated [00:01:00] control. organizational alignment and sustainability reporting to fully address energy and resource allocation. Now, let's get into it.
Grow your career and expand your FM skills with the largest curated library of FM content. If MU's Knowledge Library includes articles, case studies, research videos, podcasts, webinars, templates and more, all created and reviewed by industry experts from all around the world, full members can access the library for free. Start reading today at Knowledge library dot ifma.org.
Carl: So it looks like we're seeing a lot of interest in how customers are using technologies like, uh, digital twins to help manage their buildings. So you're the product guy. Talking to the customers the most. Yeah. So what do you,
Sandeep: you know, the, the, I think the big thing is there is an expectation that everything is connected [00:02:00] and everything is connected and there's data flowing off it.
How do I make that data accessible and usable? And that's like, That's a universal problem. That's not just our industry. It's every industry. It's like everybody has their own idiosyncratic ways of generating data by itself. That data is useful, but when you combine it and blend it with other data, like that's when that data starts to become invaluable, right?
That's when you start to be able to get integrated insights. And so, you know, your point about the digital twin is really the combination of that, which is. I have different streams of data coming in, how do I represent them in such a way that they become a facsimile of the real world? So I can get aggregate understanding and, and that is really what a digital twin is.
It's like bringing all those vectors of signals in, combining them and creating a representation of light.
Carl: Right. That's, you know, being able to understand what the relationship is between all of the disparate pieces of data to know that. What is happening is actually happening in the Northwest corner of the second floor of this building in my portfolio of buildings.[00:03:00]
And that, that particular corner has offices and conference rooms and then being able to understand what I should go do about it. The thing that I thought was always really interesting was during the pandemic. I think you showed me the data one time about how nobody's in the office, but their energy usage was still the same.
And they, they never changed it because they didn't have the intelligence to understand. The relationship between what was going on within the building to how they should relate that to their lighting and their, you think, or cooling.
Sandeep: Yeah, the, the, the schedules were static. There was the assumption that every single day was the same.
You know, people got wise that, Oh, you know, on the weekends we don't have people in that. But again, it was a static schedule. Well. You know, what happens? Next week's July 4th, it happens to be on a Tuesday, and nobody changes the schedule because they didn't think about holidays. The schedules got, I would say, more informed over time, but they never got dynamic, and to your point, um, you know, I want to know what's happening in the [00:04:00] northwest corner of the building on the second floor.
Well, can you imagine if it's because, you know, on the third floor, I've got a bunch of furnaces heating up, and all of a sudden that's impacting the second floor. It's that fidelity of information gives you more insight to be able to take action. And then you blend that with the real world understanding and you start to say, okay, Hey, I haven't just got a facsimile.
I've got a real living, breathing digital twin that represents what's happening in the real world.
Carl: Yeah. I've got all that information. And then if you take it even a step further to air quality, being out on the West coast, like where you are now, if there are forest fires, you end up getting pretty poor air quality in the area.
And you want to make sure that the air quality inside the building is safe for folks. And so if you know how, what the air flow is, you know where people are going to be in the building, you know how they're going to use the building, then you might be able to make better judgment calls on it. And so by understanding that relationship between [00:05:00] where the, that air quality sensor is within the building, what the building looks like, and then maybe air quality outside the building and temperatures that you have a whole lot better view and a better ability to make decisions.
Because that's what you're talking about, right? Is it's all about how you, how do you make better decisions?
Sandeep: Yeah. And you know, the, the, the interesting about decision making is, you know, the obvious ones are the ones we've spoken about right now is, Hey, I want to operate my building, I want to make my building more efficient from an energy consumption standpoint, you know, we hear constantly about hybrid work and how that's changing, how buildings are being used.
So. Do you know how the building is being used? You can organize the building so you can make your people more productive and you get a better yield of out the people that are using the building as well. Or if it's a manufacturing facility and you get a sense of, okay, well, you know, one of the hot zones versus the cold zones.
And how do you use that to fine tune output or yield, or if it's a hospital and you're going to lead to better patient outcomes because different [00:06:00] temperatures support different types of. Recovery or conditioning or procedure. Like these are all different ways that a digital twin and having control over the building can help you run your organization more efficiently for better outcomes.
Carl: Well, and even if you add in things like just providing the information to folks, being able to report on energy usage in the building and where in the building that energy is occurring. And then also being able to feed that into their sustainability reporting. So now they get a better feel for what's going on so they can take the right action.
So they can provide the information to the building manager or if it's regulatory requirements, be able to provide that information up to the regulators as to how they're meeting the sustainability goals as well.
Sandeep: Yeah. And I, I think sustainability is just a great best step. Like obviously there's. Um, outside requirements that, that drive it.
There may be, you know, government requirements or just regulatory requirements, but from a very basic standpoint, you can't [00:07:00] manage anything unless you can measure it and account for it. And sustainability is a great first step, right? Every customer says, well, I want to reduce my energy bill. I want to reduce my, you know, energy consumption.
Step one is just measuring what you're doing today. And I think that's a great starting point for anybody to be able to build their own sustainability or ESG roadmap is, okay, we'll just don't worry about the complex stuff. Don't worry about SBTI goals or anything else. Just start with understanding what are you doing today?
And that's a great first step.
And that's a very easy, easy step that most organizations can take with very little effort, but can. You know, really start your journey off on the right foots.
Carl: Yeah, and part of what's required though is also being an open platform, right? You know, the great thing about the Digital Twins, the Digital Twins standards for the ontologies out there, it enables you to go and build an open platform where you can bring in data from anybody's device.
So, you know, whatever the protocol happens to be, most likely it's back to that. But then also if you're trying to use other [00:08:00] types of communication mechanisms for being able to interact with either IoT gateways. And then, you know, HVAC controllers, lighting controllers, being able to gather all of that data in one place.
That's the big part of it is, you know, it is being able to pull the information in, make sense of it like we've been talking about, and then being able to report it out.
Sandeep: Absolutely. I think the great thing is like, you know, standards are becoming more robust and they're becoming more interoperable. You know, you mentioned, you know, whether it's, you know, BACnet or, you know, the protocols that are being used or whether it's industry standards and the ontologies like Brick or T Stack, like, you know, whichever one you choose, you know, you're ultimately not going to go wrong.
It's important just to choose one, because as soon as you choose one, at least that forces your data to start to become organized. And usable and accessible. And I think that again, like that's a great first step. You can't manage what you don't measure. And you know, measurement may just be collecting the telemetry, making it usable. And you know, those insights start to come very quickly once you take those first few steps. [00:09:00]
Carl: Absolutely. And we talked about the ontologies and leveraging those ontologies. What's great about also leveraging open standards, like those ontologies or the protocols is that you get to take advantage of the rest of the industry.
You don't have to be the one trying to drive it. You don't have to be the one trying to manage it. You can work with the rest of the industry to be able to pull that information in, to be able to set the standards that it is accessible for everybody. And then the other side of it is making sure that you can take that once you've normalized the data.
Making it openly available to everybody else too.
Sandeep: And the great thing is to your point, like somebody else has done the heavy lifting, like the standards are out there. They're really robust. They're very usable. There's choice in standards as well as interoperability between the stands. So there's no bad first step.
Like if you just. Right. Making a decision to go out there and start collecting that data, you'd be surprised at, you know, A, how easy it is to do it now, given that infrastructure that's out there and available, uh, but B, how quickly you figure out, [00:10:00] you know, what data works, what data doesn't work and, you know, where the gaps are, you know, I kind of think of it like painting by numbers, you know, somebody who's already built it, somebody who's already put the outline there, somebody who's already put the numbers in there.
You just go through those first couple of gaps and the picture will come to light pretty quickly. And where there's gaps, like, again, there's lots of help out there to help complete it.
Carl: And that's, you're right. Cause it's, it's really important, particularly for the folks that are managing a diverse portfolio of buildings is each of those buildings.
Probably were built by different people, designed by different folks. They implemented a different HVAC system in each other. They might have different controllers than the other. Depending upon when it was built, it might still be using analog controllers as opposed to digital controllers. And so being able to pull all that information in and being able to access all of it becomes much, much easier once you decided to leverage an open standard.
Sandeep: Absolutely. And it's, that's, that, that [00:11:00] first step of like buildings are different. Sometimes scares people. My buildings aren't connected. You know, it's scary. We're like, Hey, you know, I just have this, you know, smorgasbord of different vendors and equipments and just stuff in there. I don't even know what's in there.
Yeah. That's. Everybody's in the same boat. Like there's nobody that has a complete clean portfolio of buildings that are standardized with the right, the same infrastructure at the same levels and connected, like it's either the most sophisticated organizations. And so it's like, there's, again, like there's, there's a lot of company in them, his read that most organizations feel, and that's good. We're all in the same penny together.
Carl: Yeah. Well, you know, it's funny, I think I told you about this. One of our, one of the folks in my team, they were out doing an implementation with a customer and okay. Now they had a mix of the different devices and the like to be able to pull the information in. But what they also found was that all of those devices happened to be on the IT network, on the corporate network.
[00:12:00] And, and what was funny was that, that the customer didn't even know that. So this building had been around for quite some time. It had been put in there. They started asking my guys, they're like, well, that, that, you, why did you do this? Didn't do anything. We just started accessing the data and it happened to be on that network.
I don't know, but it, it helps to expose here's where all the information's coming from. So now maybe they could take a little bit different approaches to, yeah, in this case, split out, split it onto an OT network as opposed to, you know, having it all combined on a single corporate backbone.
Sandeep: Yeah. And things like this lessons that we can learn from the, from the IT you know, do you segment your networks, do you have different classifications of, you know, different classifications of security or accessibility or responsibility. There are best practices in adjacent workloads or industries, whatever domain, whatever you want to call it, that we can certainly draw from. And we, you know, we certainly do that as we think about buildings, you know, to your point around [00:13:00] seeing a building and we couldn't audit it, you know, we've seen organizations where customers said, Hey, we physically can't connect these devices to the internet because we don't know how or we don't know, you know, we don't know what infrastructure we have and it's better to just keep things blocked from the internet and leave things disconnected and vulnerable. Well, you know, you're still vulnerable and you still have other types of threats that can attack you from the inside.
And so, you know, just shutting off the internet and things better or the other one, which was in thing, which is like, oh, well, we thought these things were not connected to the internet. But turned out they are because somebody went to mobile modems and sell, you know, mobile hotspots and connected them that way because they needed to update them or for whatever reason.
So, you know, I think there's a lot of lessons that we can learn from legacy IT that we're seeing as the buildings industry moves from more analog to digital and being from more disconnected to connected. Again, this is goodness. Like I, I look at this with eyes open and say, well, actually there's a ton of best practice we can learn from and we're all in the same boat and we're all heading to the same place.[00:14:00]
Carl: Yeah, you're absolutely right. I mean, one of the things that I've, I've always equated what we're doing with these digital twin ontologies is we're creating something similar to what the DMTF did with the common information model for managing network devices, so same general concept. So, so you are able to understand what devices are connected to what those devices are made up as.
I made up of and then being able to understand the information that's coming from them. A little, it's certainly very different, but the same kind of concept. So you're right. I mean, you can take the concepts from the, uh, like the network management world or even enterprise applications or some other, um, parts of the industry and be able to leverage that knowledge, that expertise to be able to, to better manage and better provide information back.
Sandeep: I think one of the things that's really interesting that's coming out is, you know, everybody wants to do something with AI, right? Like, I can't read an article without hearing about AI. I can't get a resume without everybody claiming to be a now an AI expert. [00:15:00] You know, AI in my mind really comes down to three things.
You know, one is you need domain experience, like AI without domain experiences help you any, anywhere, you know, you obviously need to build the right algorithms, the right models to be able to do something. Um, but then third piece is, is the data. I think it's one of these really interesting things is we've historically looked at having data to have understanding, but now data can help build knowledge and build wisdom from an IT perspective and ultimately intelligence. And, you know, I'm curious as you think through this, the data that we're collecting, like, how does that data feed into AI ambitions?
Carl: Oh, it, it, it, it's, it's fun because the, you're right. Collecting all of this data allows us to get all these disparate points.
But once you've got that information. It allows us to understand patterns and being able to understand either it's patterns of how people are working inside of a building or it's patterns based on [00:16:00] what's going on with the weather related to the building. So being able to understand patterns is, is how you basically build your model, your inference model.
Now, yeah, using an LLM or using some other technology. Yeah, I, you know, it depends, right? It depends on what we're trying to accomplish. And, and for us, all of that data is really going to help us to, to, to be able to determine which usage is right. So, you know, an LLM might be good for being able to make it easier for people to ask questions of, of the environment to understand, all right, if I went and did something, what would happen or help me understand what's going on with this particular building at this point in time.
Uh, but then you want to be able to take the next step, which is I'd like to automate. I'd like to be able to go from, I understand that it's going to be, you know, 95 degrees outside my building faces West and it's got lots of big windows and a bunch of people want to be, uh, having a meeting on [00:17:00] that side of the building at two in the afternoon when it's going to be the hottest time of the day.
What do you do? Right. Do you, do you go and increase airflow? It's not always about do I decrease temperatures? Do I increase it? Can I increase the airflow over there and be able to do that in an automated fashion without having to have an interaction person. So the, the big thing for us is being able to do that, being able to understand that, you know, something happened it's July 4th or, you know, it's not July 4th, it's July 3rd, but we decided to make that a company holiday, but nobody's in the office.
So if nobody's in the office. I shouldn't bring the lights on and I shouldn't start cooling the building and do that based on what I'm seeing in terms of occupancy of the building, as opposed to just, I've got a very simplistic schedule view of the world. All right. And, and then the next steps are the same, right?
It's all right. You know, what else do I want to go and do? If I've got an understanding of, you know, you come into the building, you, you know, you're in your team come into the building at 10 o'clock in the [00:18:00] morning versus, you know, versus eight o'clock in the morning. I want to do something different with the building.
Uh, I want to have lighting to be different than I would otherwise. So all of those things can help to save a bunch of energy for the customer, help them get to the sustainability goals, but also make decisions a whole lot quicker and make it based off of a broad set of data, as opposed to just a very finite set of data.
Sandeep: And it's interesting you mentioned that because you think about the technology stack, there's obviously intelligent controllers at the edge, living in the building. You've got these like huge cloud services in the sky, data galore.
How should we think about the AI capabilities, where they live? What does the ideal state look like?
Carl: So ideally, what you're doing, you're collecting the data at the summarization and aggregation of the data. In the long term, what you really want to be able to do is have created a set of [00:19:00] models and then push the model down to a device locally. So that local edge device can run some set of models based off of the information that it's getting.
Certainly passing the information up, do the learning and changing of the models up in the cloud, to put half the model actually working down at the edge. But also you want to have the flexibility. If you don't have a controller that has the capacity to run the model, and can't do it at a reasonable amount of time, then sure, send the data up to the cloud, and send up what you need.
And then go ahead and take the action and trigger the action down locally. But ideally, you've got a master controller there that's got enough capacity to be able to run those models and run those and make the decisions locally. And then also be able to provide the information to the site manager, as opposed to having to wait for somebody at corporate headquarters to go do something, the site manager is going on a different view or a different set.
And also give them the opportunity to be able to override it [00:20:00] if they need to.
Sandeep: So the override point is interesting because when we think about AI, trust and the reliability like, you know, in system, can I trust it? How do we, how do you help people feel comfortable with the fact that, you know, an AI is now powering your building, you know, as hundreds or thousands of people in the building.
I'm building my safety net, but now I'm going to trust it to a computer. Like, you know, how do we help people feel comfortable with it?
Carl: Yeah. And so I think part of it is starting them with here's the action we recommend you take. Hit this button if you want to take that action so that they still have that control and they can see what it is and they can feel comfortable that, okay, I'm the one still in control of it.
And then over time that will make them feel more comfortable that yes, the decisions that the system is making, the recommendations that it's making are the correct ones. And then I can feel comfortable with just telling it, go ahead and do those on your own. Just notify me that you'd made the [00:21:00] change.
And so I think that's the baby step for, for some folks is set it up with the flexibility to be able to enable them to say, yeah, you gave me this recommendation. I'm, I'm going to make, I'm still going to have the human push the button to actually, to take the action. And then over time, all right. Yep.
Every one of those makes sense. Let me just let it go. And I trust it now to be able to do that on its own. If somebody feels really comfortable up front, then yeah, sure. Just let it run that way on its own automatically up front. But I think most folks will want to have the final say so on taking action themselves first, at least for a little while.
Sandeep: Okay. So you told me I'm getting C3PO before I'm getting the Terminator.
Carl: Hopefully it never gets to Skynet.
Sandeep: All right. So, and so that's, I mean, it's an interesting point. So we're, you know, we're going to feel good about the AI. We feel good about the technology. I've got, you know, C3PO there being like, like personal [00:22:00] AI tuned building. How, how far are we away from this becoming a reality?
Carl: You know, I, I really don't think it's that far off, but I don't think it's more than, you know, if, if we put our minds to it, the next couple of years should be achievable to have something that is. So we're able to make the recommendations for the customers on what they should be doing with their buildings and be able to do that fairly, you know, in a very intelligent manner.
And then if they're comfortable with it, allowing it the system to go ahead and do what it needs to go do.
Sandeep: So we're a lot closer to the autonomous building than we are the autonomous car.
Carl: Yeah. Yes. Yeah, very much so, very much so. I think the autonomous car, you know, what's funny is that if you think about all the electronics and all of the number of processors that are in a car right now, I mean, there's thousands in a car and yeah, that there's probably more in a car than there are in a typical building these days, [00:23:00] but the decisions that you have to make for a building are not as necessarily as complex as they have to do for autonomous car.
You're not having to worry about, you know, somebody just randomly stepping off the curb while they're looking at their phone in front of a car, doing that in front of a building. Don't care. The leg's not going to move. So, so the decision making process doesn't have to be the same. It's not necessarily as complex as that for a car.
Sandeep: It's super interesting to hear the comparison between, you know, vehicles and buildings. A lot of people think about AI, think about self driving cars, but you know, the self maintaining, self healing, self operating building is probably something that's going to impact more people sooner. Just given how tangible it is.
So, you know, really interesting stuff. Well, Colt, thank you so much for taking the time to, to go through this and, you know, talk about this topic with me, if you'd like, love to learn more. I'm Sandeep Modhvadia, find me on LinkedIn, find me at atrius.com, [00:24:00] Carl.
I'm Carl Coken, Sandeep. As always, it's great chatting with you.
Again, find me on linkedin.com or at atrius.com where you can learn more about what we're doing to make this dream a reality.
Host: Thank you so much for listening. I hope you really enjoyed this episode. And as always, please don't forget to rate, review, and subscribe to the podcast for more incredible content.