#49. Smart and sustainable buildings with AI

Episode Summary

In this episode I speak with Ysaël Desage about sustainable buildings using AI.

Episode Notes


Historical and Current landscape of building energy usage

Smart buildings and even smarter buildings with current AI

What BrainBox AI does

How swarms of AI enabled buildings can work together on a grid scale

AI modeling with 5Sigma climate events

Future of building energy usage


Episode Transcription

Welcome everyone to another episode of Prompt Pod, an open-ended 
exploration into world-changing technology, and my quest to document 
conversations with bright minds in the space. I'm your host, Danny Kirk, and 
today I'm joined by Ysaël Desage. Ysaël is Brainbox AI's technology manager in the 
field of energy networks, with over six years of experience in energy and 
artificial intelligence.
His academic background combines fundamental physics, electrical 
engineering, computer science, and operations research. Recently featured in the
National Observer Newspaper, s i l and his team are recognized for their 
groundbreaking work in transforming buildings into intelligent actors in the 
energy grid. Ysaël,
Welcome to the show.
Ysael: [00:01:00] Thank you very much, Danny. It is truly a pleasure to be 
Danny: Could you give our listeners a brief overview of the current landscape 
of how buildings, um, kind of interact with the grid as far as energy usage goes?
Ysael: Absolutely, and I think it's very important to take the time to, to set up 
the, the whole landscape. You know, before joining brainbox ai, um, I was not 
aware at all about the, the whole impact that building had. Uh, why, because 
we, we have buildings in our everyday lives from the, the house where you live 
in, when you go do groceries, you go to work into commercial building if you 
don't work, uh, from home, of course.
Um, and, and it's just part of our life that we don't.
We actually were even born and they will probably be there even after, um, our 
children, uh, and the next generation come out. And the, the thing is that, I 
dunno if most people dunno that, but [00:02:00] buildings account and their


home operations, they account for more than 30% of the global energy 
consumption. And that's more than 26% of the energy sector emissions.
So that's over one quarter of the whole energy sector emissions in GHGs. Um, 
so that's a huge competitor. Sorry, not a huge competitor. A huge contributor. 
Um, and typically for these buildings, we're aware that they're a huge 
contributor. Um, so what we're starting to see is what we call retrofits. Um, So 
what happens is a retrofit is when you change the structure of the buildings, you
would typically send an engineer or a specialized, specialized resource.
Um, it can take over a few days or even weeks. And then you try to update your 
systems, try to, to get into new reality and try to improve the, the efficiency of 
the building. I'll be brutally ho uh, honest here. Um, we just don't have the time 
for that. Just to give a number to, to our listeners here, almost every month you 
get a whole Manhattan of [00:03:00] buildings that are getting constructed, 
So this is, those are all the new buildings. And add to that, all the buildings that 
already exist, if we're targeting net zero in realistic times, this method simply 
won't make.
Danny: And, um, as far as how, uh, buildings have interacted with the grid, 
how are they using energy historically? What does that look like as far as, uh, 
their demands go and the grid as far as, uh, keeping it up and whatnot? Could 
you give us an overview of that?
Ysael: Absolutely. Uh, the main consumption for buildings comes from the 
heating, air conditioning and ventilation. and as you may know, the 
temperatures are are getting more extreme, so we're getting colder Climate. 
Warmer climates where we're seeing it right now, right In over Europe you have
this bubble of, of heat.
and so the biggest component is the heat. And we'll see it later when we talk a 
bit more about the grid, but it's everybody heats at the same time, right? And 
that causes huge peaks on the grid and that's what is very harmful to utilities and
how buildings are essentially,[00:04:00] 
Danny: I know you have a background in physics and electrical engineering 
and computer science. What, um, got you so interested in kind of, uh, grid usage
and smart buildings to begin into it.


Ysael: I think it's the impact we can, you're trying to do your piece recycle. 
You're trying little things. I have an opportunity, opportunity with a simple 
software update with a simple, software, just adding intelligence to something 
that already exists, avoiding, uh, using uni resources. just with that you can have
a huge impact, and that's the part that really drives me and that what I love 
about my job and what we're doing in the, in this industry.
Danny: could you give us kind of a brief overview of what you do at Brainbox 
ai, I guess what Brainbox AI does and also what you do there?
Ysael: Absolutely. I joined Brainbox AI in thousand 19. and I think that it's 
worth telling a bit of the story behind Brainbox ai. initially our founders, they, 
they were sitting, uh, at a restaurant and [00:05:00] you know how we hear 
about all those self-driving cars and how the technology evolves so fast. But 
when you think about it, self-driving cars, it's high velocity.
You have so many components. You have computer vision, you have a lot of 
things like that. Um, and I mean, we're getting there, right? If you look at Tesla 
or other products, it's really getting there. And they said, well, how come we 
have that for the automobile industry? But we don't have that for buildings, 
Buildings, like we said, are one of the, the core contributors, uh, there, and that's
how they started. They said, well, this look into adding kind of AI into those 
buildings that start collecting data and can just reduce the energy consumption. 
So then they ran a few tests and out nowhere it worked. It worked.
And that's how the first phase started, right? So I was telling about that software
update. Uh, we don't have time for engineers to go in and, and do all the manual
work, and then if it changes one after one month and you have to redo it again. 
So they said, let's use ai, which allows you to [00:06:00] scale, which allows 
you to learn every specific building, understand the.
Part of the solution where, um, essentially you, you just collect the data every 
one to five minutes, and then you have the AI understand what's, what's the 
impact of the sun. For example, let's say at 12 it'll be hitting in that specific 
room. So maybe I need to call before, or I need to anticipate in advance, or I 
know there, there will be a meeting at 1:00 PM on Thursdays.
And again, my algorithms can go from being reactive. That's how they're, the 
buildings are right now to actually be proactive. So anticipate what's ahead.


First generation of then, what we realized afterwards is that we could go even 
beyond started saying, could the building be more about even more things?
And that's when we started what we call autonomous decarbonation. 
Decarbonization, pardon the French word. I have a French accent that probably 
betrayed me by this [00:07:00] time. and the goal of this whole autonomous 
decarb is, we call it measure reduce. A lot of customers, they wanna have a, 
they want to have a plan.
They go towards net. So the first thing we tell them is, well, we need to measure
for you. So I'll do again, a little parenthesis on the grid side for our, for our 
listeners, there's three big, when we quantify GHG emissions, we, we do it 
using what we call three scopes. So scope one is essentially all your direct 
activities as a business, as an entity, as a collectivity.
scope two is essentially related to. So are you consuming energy that comes 
from a clean power supply or are you actually consuming from, coal plant, for 
example? So that's what we call scope two emissions, and then Scope three 
emissions account for all the upstream and downstream impact. So the first 
thing we said is, is, well, we kind of have a solution to reduce, the energy, be 
more efficient.
Now let's help our customers.[00:08:00] 
And then we said, let's add our solution as a second step. So let's actually reduce
the remissions, both in scope one and scope two. So we'll get back to it on, on 
how we can do that with the Scope two and in scope three. Well, they still 
wanna get to, to net zero, but they can't. Let's open up the whole carbon, uh, 
offsets and markets to them.
So in a nutshell, that's.
Danny: I was speaking with, uh, the c e O of Greenlea the other day, and he 
reminded me that there are only about enough offsets, uh, for 1% of all, uh, 
global emissions, and only about a 10th of those are high durability offsets, 
which actually really do anything. So I assume that's kind of a big piece of the 
puzzle for you all.
It's really all about, uh, reduction of emissions versus like, that should be step 
one. Is that correct?


Ysael: Absolutely. And, and I like to [00:09:00] say to call it energetic sobriety.
so yes, we can improve and build more power station, we can do a lot of things, 
but if we could just consume a bit less, then we're already having a a huge 
impact. And the other thing, like you said, is also that, we have something that 
exists already and we want to first optimize this before adding, other new 
technologies and.
Danny: Could you give us kind of a view of the spectrum of smart buildings? 
You know, I have a Nest thermostat, um, and that, you know, has a little bit of, 
very simple machine learning in there, but it honestly is, doesn't seem that smart
to me. Could you give us kind of the full spectrum of kind of the smart 
buildings and the technology and the AI that you see today?
Ysael: That's a very, very big question. So again, I will take the time to set up 
the landscape. I'll maybe start with the grid, uh, if you will, Danny. I discovered 
this when I was doing my, my eng, my electrical engineering studies, and I 
think it's fundamental. The second you plug, uh, your micro, you use your 
[00:10:00] micro whatever, you, you, you plug that hair dryer, in real time, you, 
you need to have something that provides the electron that we're consuming.
So just like a pump would need to spin up faster to avoid pressure drop when 
you actually consume water. Electricity electric. As soon as we consume 
energy, you, you need to have in real time some things that produces it. and so 
the whole idea about this is that initially when the, the power systems were 
created, we said that they were centralized and unidirectional.
So why? Because we had these big power plants that would essentially be very 
far away from cities and they would push the energy towards the city. People 
would consume it, and we would transfer the energy over huge transmission 
lines. The problem.
Are starting
SMO changes the [00:11:00] paradigm. People are locally on the grid. So that's 
starting to make a huge, headache for the utilities themselves because they still 
need to provide that electron in real time to all the consumers whenever they 
consume it. And you and I, I mean, we can, they can have incentives to change 
the price of electricity, but at the end of the day, when I plug and I start my 
microwave, they need to deliver the electron.
If they don't.


But so it starts from that aspect. So what we need going ahead and for the 
energy transition is what we callers. So instead of just being passive consumers,
we want become more aware and give flexibility to. Modulate our, the way we 
consume. So maybe for the utilities, just about consuming a bit less for 30 
minutes, and it could be a whole game changer that would solve their 
Um, so all this new generation, and I think it's a great example, the, the 
Department of Energy, the United States called them the grid, interactive 
efficient buildings. Um, so that's something that we very closely adhere to, and 
it's the idea that, that I just mentioned. So take the buildings, which are probably
one of the biggest burdens right now in the, the natural way that they're, and 
turn them into active.
So then make them smart, make them able to be, to change their modulation 
based on the utilities needs. and so,
Smart buildings and small thermostats.
Danny: Do you think that, um, kind of house unit-based, um, energy storage is 
a part of, um, the solution here? You know, as, as you said, when you plug in 
your, uh, microwave, you, no [00:13:00] matter the cost of energy you do want 
it to work. So is do you think kind of unit-based storage is something that's 
critical to kind of, uh, growing this.
Ysael: I think so. I think the more actors we have, the, the more players you 
have in your team, the more opportunity you have. Um, but you also increase 
the. So it was easy before because you just had the big power plants. Well, now 
if the utility
storage unit.
Um, it is starting to get there. Uh, artificial intelligence is transition.
Danny: Before we jump more into kind of AI and whatnot, um, are there any 
storage [00:14:00] mechanisms, energy storage mechanisms that you're 
particularly excited about seeing developed these days?
Ysael: That's an interesting question. Um, it very difficult to store energy at a 
large scale.


If we can reach a level where it becomes sustainable and sufficiently efficient, I 
think it would be very promising. But I believe a lot more, and like you 
mentioned, smaller units, uh, more volume. So having a lot of smaller units 
distribute the, the intelligence. And to give you an example of storage type that 
we consider in grid interactive buildings, it's the thermal inertia. Heating.
The material of the building. And so if I stop the HVAC for a certain period of 
time, you, you can do an analogy with a thermal battery or a, and then that 
means that the [00:15:00] building will contain parts of it, of it, of its heat and 
then will slowly lose it towards the exterior before it reaches a, a critical 
temperature that you need to restart.
So that's like if your battery was empty. Um, so that's one of. My, the, the things
I played with daily,
Danny: Tell us a little bit about swarms of AI buildings and how they can work 
together on a grid scale.
Ysael: We start from the idea that for each of these assets or for each of these 
buildings, um, we set it, the AI allows us to, uh, be aware, so we know exactly 
how the building will behave, how the building will react. Um, we're aware of 
what's the consumption, what's, what is the unit actually producing the 
electricity right now, wind blowing. Each of these agents is aware of its own 
instance, so it, it can ensure the, let's, let's call it, make sure that the building 
that it is responsible of behaves [00:16:00] properly in its own interest. But then
it also opens up the way to distributed intelligence that we're talking about. So 
let's say on the grid, you can offer a product or offer something that essentially 
you could.
Of buildings in a huge city, then you get very powerful presumes because they 
understand how the asset works. They make sure to never violate things like 
comfort. They make sure to keep, um, the, the power peaks, the local power, 
peak flow. They make sure to consume, to still try to consume less energy for 
that specific building.
Yet they act collective collectively, intelligently, hence the the swarm, concept 
showing off.
Danny: Is there any place in the world that is currently, um, have swarms of 
buildings like this?


Ysael: This is, so it's the grid. Interactive buildings are quite a new concept, I 
would say. Um, to give you an idea, um, [00:17:00] at COP 26, we won the, the 
Tech for Our Planet Challenge with this idea that we introduced. Um, so it's 
something very new and even in the, the test and the, the things that we. It's 
very, very new in the industry.
So we're at the very early beginnings, and I think the, the next five, 10 to 15 
years, we'll likely see an explosion, um, of these kinds of technologies and the, 
the, what they can actually offer because it's something entirely new that we're 
not used to, uh, to have available.
Danny: does Quebec and Canada in general support this sort of thing? Is there a
lot of funding going toward, um, kind of more efficient energy usage in 
Ysael: So that's a good question. Again, I'm not maybe the expert in funding on,
uh, on Canadian and Quebec sustainability. Um, I know that global funding 
worldwide is definitely increasing according to the Energy, international Energy
Association. I would say that in our North America in general, um, there is, 
there are [00:18:00] major.
Danny: Do you think one is more important than the other as far as using ai and
the first being, um, individual buildings using ai and then the second being, uh, 
the energy providers, essentially the grid. Is there one that is AI would be more 
useful for, I suppose?
Ysael: So AI goes into a process, so. In the context of buildings, it is the 
process, like we said, of understanding the building, learning, the building, 
learning the asset. utilities are starting to use AI as well. Uh, maybe it's more on 
the other side of the optimization of which assets should I call, and then the 
control decision that I do.
At the kind of eventually those.
Danny: When we have swarms of AI enabled buildings that are all working 
together, [00:19:00] does this change the way we interact with our health, uh, 
homes or are we just kind of business as usual and we won't even notice it 
happening? I.
Ysael: So that's a very important point to bring. do you want to have the user 
actively in the loop? so when you do things like that, then you need to fall back 
on statistics. So let's say take a, we have a smart thermostat and then we create


some sort of program where we tell you we, we'll send you a notification and 
then if you want, you'll it.
Then as the utility, what happens is I need to go and say, okay, statistically I 
think 70% of people will answer, but I'm not sure I have, uh, uncertainty on this.
so then it's a lot less reliable as a tool. But then if you as a user agree that. You 
have your ai, of course, you always have the final word if you want, but you 
accept kind of a more of an autonomous solution that can communicates with 
the grid.
Then the grid and the utility can know a lot more precisely that when it 
[00:20:00] actually asks or uh, exchanges information, it'll be accurate and it'll 
actually be properly delivered. And that is a huge, uh, game changer for the 
utilities themselves.
Danny: I think the autonomy there is a big piece. You know, I kind of, uh, 
strike the metaphor of. home security systems. You know, when you get a 
security system installed, the first thing they tell you to do is turn off all the 
notifications 'cause they're annoying and it doesn't work that well. you know, I 
have my smart thermostat and I had to teach it quite a bit.
but almost at the end of the day I've just kind of set it on manual just because it 
took so much of my time to learn. Honestly, if, um, if I could give the utility. 
Carte blanche as far as keeping it at the right temperature at the right time of 
day and do whatever you want to do that. Um, it seems fine with me.
Do you have an opinion on kind of that user experience as far as like those 
smart homes go?
Ysael: I think you said it, the key.
I [00:21:00] personally share, personally, I share your opinion. I'd much rather 
have a carte blanche algorithm that essentially give authorization. I define.
of course we always want the human to have the last word, so at any time I can 
remove my, uh, agreement, I can deactivate the system. I can control at the core 
of the design. but I think there are people that would prefer having notifications 
themselves, and I think we, we can have the, the both. The best of worlds.
Um, so yeah, I would.


Danny: I know when we first spoke, we mentioned, AI modeling and five 
Sigma climate events. How are you view, uh, viewing those events, uh, these 
days that are literally changing? Models of weather patterns and whatnot as it is 
related to the ais that you're trying to build and train right now.[00:22:00] 
Ysael: So it is very interesting because if you look, and I'll go outside of energy 
outside of hvac and I'll focus purely on AI right now. the holy grail of artificial 
intelligence is, is what we call generalization. So what is generalization? It's the 
idea that, did you really understand what you learned on?
So with the experience that you acquired, are you able to perform properly on 
something that you never saw? That means if I never touch something warm or 
hot, can I in advance know that I'll get burned without being burned? And, and 
that's at the very core of AI because yes, it's very nice to be able to learn and to 
have something that sells to that.
We've seen that since the two thousands, but now we need to go a step beyond. 
So we need to say, is the AI able to understand?
Memories and connect patterns that it actually can project and understand that if
I do this, this will happen. Um, and that's the key in of AI in general. And 
[00:23:00] even for HVAC and for buildings. It's also the key because you 
mentioned it. things change and even I'll go to the, the Five Sigma events that 
you mentioned.
but first of all, just think about it. When you start in a building, you're either in 
what we call heating
hvac. So how can the AI know what's the impact of cooling? That's where you 
need to step up your game and, and you need the most recent researchers in AI 
so that even without seeing the cooling season, you already have a good 
estimate or a good understanding of what would be the impact. And then as the 
new data comes in, then it continuously adapts and improves again.
And the same, uh, reasoning applies to those five Sigma events. So just for, for 
your listeners, five Sigma events are things that happen very, very rarely on a 
statistical, uh, basis. and again, we're seeing it right now with the, the heat 
waves. So when these heat waves actually happen, even if we have not seen 
them, uh, it's very important for.[00:24:00] 
That's part the whole key of the development and the research of these 
algorithms and even of the validation. So before deploying algorithms wanna


run some of those validation tests to make sure that it would be robust to such, 
uh, circumstances.
Danny: From working with Brainbox ai, how has this changed how you interact
with your home?
Ysael: So that's, I would say we focus mainly on commercial buildings. I would
say personally, it made me want, just like you to, to acquire smart thermostats 
and even test, uh, a few algorithms on my own. however, I believe in the, I want
to have the biggest impact as fast as possible, and I think commercial buildings 
are a very good starting point.
Why? Because they have what we call centralized systems. The numerical 
signals from and controls, which is a bit [00:25:00] more difficult for houses 
because houses have, so sometimes custom things you need to go directly to the
unit. Uh, in bigger commercial buildings, it's already centralized. So that's 
where with a single install, we can have a bigger impact. Um, just to give you 
an idea with the solution that we have right now.
Bigger buildings consume a lot more energy. And I give you just a few numbers
because I'm pretty proud. As, as one of the developers of this, uh, we've seen 
reductions in total energy cost up to 25%, 40% in carbon footprint, and even 
60% increase in occup while keeping those.
Danny: That's incredible. And kind of along those same lines, what do you 
believe is underrated in the flight fight for climate change, uh, to correct these 
issues? Is there anything in particular that you think needs to be talked about 
Ysael: I would say everybody needs to do its part. of course there are big 
players, like buildings are one of them. Um, we [00:26:00] know who the big 
contributors are. Uh, so I would say all the improvements are good. And I think 
the important is that the sum of all these smaller improvements is the thing that 
will to make the difference and ideas.
Danny: Give us your view of 10 years in the future as far as kind of smart, uh, 
buildings go. What does that look like to you?
Ysael: So I'll go back to that idea. of the, right now the. Reactive. They're doing
their things on their side, consuming whenever they want. if for in 10 years 
you'll have agents for the building, you'll have agents for the electric vehicles.


Everything that consumes energy will essentially have their own ai and they 
will be able to communicate together.
They'll be able to exchange information, they will be able to partake and, and 
work collectively towards essentially, uh, bringing value to.
Structure. [00:27:00] That's a personal, uh, vision of the, the whole energy 
landscape with the introduction of ai. And I think that we're already seeing the 
impact of the lack of flexibility. so if I take a look simply at what happened with
Ercot in, uh, in 2021, um, again, a few numbers, 4.5. $195 billion of property 
damage, 57 death.
Uh, and that was just one instance, right? so if we're able to avoid this and just 
add, like I was saying, as much intelligence as we can and provide flexibility to 
these utilities, which will be one of the, the key points in the energy transition, 
um, I think it's definitely one of the most exciting things.
Danny: Always our final question. What was the hardest? You've laughed 
Ysael: Ooh, that's a good question. What was, when was the hardest that I left? 
I'm trying to remember all the safe forward work events that I, that I can 
mention in, in this podcast. I'm not sure actually, [00:28:00] Danny, maybe I 
need to.
Danny: Yeah. Yeah, no worries. Any funny videos or anything like that?
Ysael: No, not either.
Danny: Nice. Well, yeah, maybe a com comedy show in Montreal soon. Yeah. 
Well, SAEL, thanks very much for coming on our show. If listeners are 
interested in learning more about what you're doing, where can people find you 
Ysael: So I would recommend, uh, brainbox AI's main website. So we're always
accessible. I'm just one of a huge team. We have so many smart people. Always 
happy to exchange with you or professional, um, websites such as LinkedIn also
to go.
Danny: Excellent s Sael. Thanks again.


Ysael: Danny, it was my pleasure. Thank you. And thank you to all the 
Danny: And thanks to you, my dear prompter for tuning in, and I hope you 
enjoyed this conversation as much as I have. If you enjoyed the show, please 
consider subscribing and leaving a good review. Take care and always be