How Drones And AI Are Mapping Heat Loss Across Neighbourhoods in the UK

1.8 homes per minute.

That's how many homes need retrofitting between now and 2050 if the UK is going to meet its energy efficiency goals.

The challenge isn't just carrying out the work. It's knowing where to start, which homes need attention first, and whether retrofit projects are actually delivering the results they're supposed to.

Lucy Lyons, Co-Founder and CEO of Kestrix and a Forbes 30 Under 30 honouree in 2024, joins us on the UK Construction Podcast to explain how her company is using drones, thermal imaging, and AI to create what she describes as the "Google Maps of heat loss". By turning aerial data into detailed 3D models, the platform helps housing providers identify heat loss, prioritise investment, assess retrofit opportunities, and measure outcomes across thousands of properties.

The conversation explores social housing, EPC assessments, retrofit funding, AI in construction, energy efficiency, and the role technology could play in helping the UK tackle one of its biggest infrastructure challenges.

Connect with Lucy Lyons:
Kestrix Website: https://www.kestrix.io/
LinkedIn: https://www.linkedin.com/in/lucy-lyons/

____________________________________________

🚀 Watch, Like & Subscribe to our YT channel for more episodes: https://www.youtube.com/@UK_Construction_Blog

🔔 Turn on notifications so you never miss an episode

Thank you to our sponsor for this episode, WebMotion Media:

https://web-motion.co.uk/

Special shoutout to our dynamic host Jimmy Webb of Construction Cogs: https://constructioncogs.com

____________________________________________

Follow us for other industry updates, thought leadership, house building/DIY tips and the occasional crossword puzzle!

► Website: https://ukconstructionblog.co.uk

► LinkedIn: https://www.linkedin.com/company/uk-construction-blog

► Instagram: https://www.instagram.com/uk.construction.blog

► X: https://x.com/UKCBlog

► Facebook: https://www.facebook.com/UK.Construction.Blog

► Threads: https://www.threads.net/@uk.construction.blog

► Bluesky: https://bsky.app/profile/ukconstructionblog.bsky.social

Transcript

From groundbreaking projects to game-changing innovations, the UK Construction podcast brings you face-to-face with the industry's brightest minds and boldest thinkers. Each episode features candid conversations with construction leaders, architects, engineers and on-site experts who share their hard-won insights and behind-the-scenes perspectives. 

We cut through the noise to deliver actionable intelligence on market trends, emerging technologies and the forces shaping British building. Hello everyone. We're covering an area that I work in today, lifting operations. 

JIMMY:
Hello all. Today we have Lucy Lyons, co-founder of Kestrix. This is a great example of how campaigns can help startups to be really successful in bringing innovation to a much needed area.

Bit of housekeeping now, for anyone out there who's listening on an audio platform, I will say that if we have time at the end, Lucy will be doing a screen share segment where she demonstrates her platform, so if you want to see that you'll need to switch over to YouTube to watch it. Lucy, thanks so much for joining.

LUCY:
Thank you for having me.

JIMMY:
Yeah, it's a great pleasure, it's a great pleasure. I like startup stories, I like success stories, so I'm really interested to see what you're all about. So let's do that.

Let's start with who you are and what Kestrix is all about.

LUCY:
Wonderful, so I am Lucy Lyons, I'm the co-founder and CEO of a startup like you mentioned. Startup is called Kestrix and basically the best way to think about Kestrix is we're trying to be like the Google Maps of heat loss. So we commission thermal imaging and RGB image drone flights over lots of buildings at once and we have algorithms that map and quantify how heat is escaping from different parts of the building and use that to generate an indicative energy retrofit plan.

So it's kind of like the AI energy surveyor. We're basically trying to build sort of the data layer to power the transition of buildings to sort of more energy resilient, net zero, cheaper to heat. We also do assessments for heat pump readiness and solar PV potential, so we use sort of all the geometry and geometric information that you get from sort of the drone flights to see kind of where could a heat pump go on the side of the house, where could solar PV go.

So really trying to take drone images and turn them into actionable insights for making buildings better.

JIMMY:
Right, so that's basically, yes, I was going to say can you briefly explain what a retrofit is?

LUCY:
Maybe I can take it a step back. So globally about 25% of CO2 emissions come from heating and cooling all these buildings that we live, work, and play in. But the reality is that it's actually relatively simple to sort of cut those emissions way down, not to zero.

If you seal all the heat leaks in these buildings and upgrade the energy systems, you can really whack these emissions out. I guess on past just thinking about emissions, obviously homes that leak a lot of heat are really expensive to heat. And yeah, we don't want to have that, especially in sort of social housing or areas where vulnerable people are living.

It can really cause economic problems and challenges to have very, very high heating bills. So when we say retrofit, technically retrofit can also mean, you know, retrofitting your kitchens and bathrooms. But in the case of energy retrofits, we're talking about putting insulation in, making buildings more energy efficient to run, and then indeed upgrading the energy systems.

To give a sense of sort of the scale of the challenge, if we look at the UK alone, roughly 27, 28 million homes here need to be retrofitted to hit our sort of collective net zero energy efficiency goals. If you break that down, it's something like 1.8 homes per minute that need retrofitting between now and 2050. So it's just a staggering infrastructure challenge.

And like I mentioned, it started as a conversation about net zero, perhaps, but really, it's a conversation about fuel poverty. It's a conversation about damp and mould. You know, a retrofitted home is a decent home to live in.

So when we talk about retrofit, we're really talking about upgrading and making homes livable for everyone.

JIMMY:
Perfect, perfect. So you mentioned you work a lot with social housing. So is this something that those companies are really aiming for now that they're net zero?

Are they going to be doing it across all their properties?

LUCY:
So I think it's better to think of it less as a net zero target and more as an energy efficiency target for all the sort of positive spillover effects that I just mentioned. So there's a push across the sector to reach EPCC. So for anyone not familiar with retrofit and energy efficiency, energy efficiency today in the UK is measured mostly through this mechanism, the Energy Performance Certificate.

It's basically a rating from A to G, tells you how efficient a building is, more specifically, how much it costs to heat the building, which is, yeah, we can go into sort of the pros and cons of the EPC system. But at the moment, nationally, that's how we measure energy efficiency. And the social housing sector has broadly a target to reach EPC rating C.

Actually, you should check the exact date. I don't think the sector has a date specifically, but there are a number of rules and regulations that, for example, the minimum energy efficiency standards for landlords more holistically saying they have to reach EPCC by a certain point. I have to check.

I probably should have my statistics in front of me. But anyway, it's basically a target that all landlords have to sort of comply with, including social landlords. And I think because social landlords have sort of a social duty of care to their residents, this is something they really prioritize.

So the question is not will they do it? It's how will they do it most efficiently and reaching the right people first and with the least amount of capital, also considering, you know, skill shortages, there's not enough retrofit assessors, not enough installers to sort of do the work. So that's really where we feel we come in is sort of using technology to make that happen faster.

JIMMY:
Yeah, gotcha. Gotcha. Yeah, you mentioned energy assessments and EPC.

So is this an alternative to traditional energy assessments? Or is it something that can sit alongside it?

LUCY:
So at the moment, we see Kestrix as a as a complementary solution to traditional EPC assessment. For Yeah, I mean, to go into a bit more detail, there's something called a retrofit assessment as well, which is even more detailed than an EPC assessment. EPC assessment kind of gives you a letter from A to G, a retrofit assessment will be more sort of diagnostic and tell you what exactly needs doing.

You need quite a bit more training to become a retrofit assessor. The reality is, though, is that with 1.8 homes per minute to retrofit, between now and 2050. We simply don't have enough people to be doing, you know, two three hour surveys for every single home that needs doing.

So what Kestrix really does is we can fly over, say, 4000 homes, you know, in a couple of days, and create a 3d heat map of not just every single home, but the whole neighborhood. And that 3d heat map, especially when you use AI to pull numbers out of the images and prioritize properties based on that can help us basically send the retrofit assessor to the right place. So we believe we can keep, you know, keep innovating and iterating on the product and and get to the point where we're extracting huge amounts of information.

But at the moment, the way that we relate to the sector is we're an enabling tool. We want to help make sure that assessors, the few assessors we do have are going to the right properties. You know, what what you see a lot in the sector now is you think that a home is eligible for funding from the government for retrofit, because it's EPC rated D, an assessor will be booked, they'll go to the house, they'll come in and they'll find really quickly that actually some work's been done on the house and an assessment hasn't been done since and actually the house is a B.

Now that renders that house then ineligible for funding, which means you've just poured, you know, 200, 300, 400 quid down the drain and a retrofit assessor's wasted their time, not to mention the residents had someone traipse into their house and only to be disappointed that they're not eligible. So what Castrix is really about is how can we look at lots and lots of homes really, really quickly to determine which are the ones we want to focus on, where are we going to get the biggest bang for our buck? And that's really important when you have finite time, financial resource, and labour resource really.

JIMMY:
Absolutely, I totally get that. Yeah, so that pinpoints the location and it also helps people decide where a deeper assessment is needed first.

LUCY:
Exactly, yeah. And I think even beyond that, it's about equipping a surveyor with as much information as possible when they do go to the property. So for example, a lot of people, you know, for them, the loft is difficult to access.

Many times a surveyor then won't be able to get into the loft on their first visit. With Castrix, you can see the loft, so the assessor can say, is that something I need to go check out? Or actually, does it look like it's fully, it's insulated and the insulation is working?

You know, if they're able to go in with that information, they can just do their job faster. Not to mention, you know, our platform also can tell you with, to quote, level accuracy, what's the area of all the windows and the walls. You know, wall to window ratio is something that's part of your EPC assessment, you have to calculate it.

That saves them that time. So it's about, number one, how can you make sure the assessor is going to the right property? Number two, how can you equip them with information that makes them go that much faster?

So it's a great example of sort of complementing technology. Now, I won't say what would be, what will be possible in the future. You know, technology is moving really fast.

I would love to get to the point where an assessor could do 50 homes in a day because our data fills in 90% of what you need. But yeah, we're unlikely, I think, to replace any surveyor fully.

JIMMY:
No, okay. Would you say it's a good way to also cross-check an EPC that's already been given?

LUCY:
Definitely, yes. So in fact, you asked how we relate to the EPC. Maybe I'll just take a step back and explain a bit how Kestrix works and how we relate to the EPC from a more technical perspective.

So, you know, we fly over homes, our algorithms take the images, stitch them into 3D models, and then estimate the rate of heat loss through different building material components. So that's the walls, the windows, the doors. We actually come up with hard numbers.

So for anyone with a sort of building physics background, we're estimating U-values or basically the rate of heat transfer through different subsets of materials. And you use that information basically to prioritize where in the building should be attended to first. But yeah, I would say in addition to that, we're turning qualitative data into quantitative data.

And what that means is that if you look at all the different sort of elements that make up an EPC, we can give estimates, our estimates of numbers of many of those elements. So I think it's something like 60%, 70% of the EPC could be influenced by numbers that we're So that doesn't mean we replace the EPC. You know, we'd have to go through quite a long sort of regulatory journey to even get close to that.

But we do give data that helps a surveyor sense check the EPC. So what we're now building is, you know, not only can you look at a 3D model on your screen, if you use Kestrix, you can see how does the Kestrix data match up to the EPC data. And therefore, does Kestrix suggest that this EPC is right and you can trust it, which means that it has a stamp of approval, and you can feel really confident about that data.

Or does it disagree with the EPC data? Now, a disagreement could mean many things. It could mean that the Kestrix data is all correct.

It could mean that the EPC data is outdated. It could mean any number of things. But it just means you need to take a closer look.

So we help sort properties sort of into buckets of we can really trust these EPCs to this one seems shaky, the numbers aren't adding up, we need to take a closer look. So we won't be able to sort of give a new certificate, but we can challenge that data. We can use pictures turned into numbers to sense check an EPC.

And that's our EPC comparator tool that's coming out in, yeah, about three weeks from recording. So by the time anyone listens to this, that will be available for testing.

JIMMY:
Great stuff. So are you the only company that's doing this?

LUCY:
So as far as turning images into quantitative data, and plugging that quantitative data back into a SAP-esque calculator to sense check an EPC, I could say pretty confidently, yes, that we're the only company doing this. There are other companies that are capturing thermal imaging at scale. But I think this is really Kestrix's differentiators is we're not as focused on the data capture.

You know, we're not capturing millions and millions of homes, we've done 10,000 surveys for very specific clients with very specific needs. But what we are focused on is the quantitative interpretation of that data. So it's not data for data's sake.

You know, lots and lots and lots of data is great for training algorithms. But you know, you have to deliver insights. And for us, the insights that our customers want, have to do with how does this relate to the EPC?

And also, what does this mean for my retrofit and asset management plan? It depends on the stakeholder you're talking to. But many of our customers are social housing providers.

And that's what they're, that's what they're wondering. But yeah, there's a few other, there's great companies working on sort of data capture at scale. XRI is one that we, we know quite well.

And over in Canada, there's a company called Keytech, one called MyHeat, that's working, I think, more on resident engagement. Yeah, there are a few players, but it's, it's hard stuff. You know, if anyone has a background in thermal imaging, they will know that thermal imaging is more an art than a science.

It's, it can be very finicky. It's very sort of dependent on having lots of other data. So if you don't have internal temperature data, famously, it's difficult to glean much from thermal imaging.

You know, different weather conditions can distort outputs. But this is, this is kind of all the stuff that we're aware of and sort of correcting for at Kestrix and trying to, trying to wrangle. So when we say we use AI to do things, it's a bit different from a lot of companies who kind of use large language models to recommend things.

You know, for us, it's much more about 3D reconstruction and, and physics, really, understanding what does a temperature reading mean? And what can we glean from that? And how sure can we be?

JIMMY:
So I hope something just, yeah, yeah, that's, yeah, that really does answer the question. But something, something totally random just popped into my head. So what, so obviously, you get people, you get these people, like, grow weed and stuff in the loft.

Yeah, and that, that will generate a lot of heat, won't it, in the loft? I've heard of cases like when a row of houses are in the snow, and then that one house hasn't got snow on it. So do you pick anything like that up?

And what do you do with that? Do you have to report it?

LUCY:
So it's really interesting. A question that we get a lot at Kestrix, and I promise I'm going to tie this back to what you just asked, is, isn't this illegal? You know, how can you fly drones over people's houses?

Surely this is a GDPR violation. And it's definitely a hazy area. But we've done quite a lot of work on the topic, because obviously, we have to, to do what we do.

Part of that was, we worked with the Information Commissioner's Office, the ICO, in their sandbox program to basically stress test our approach to data privacy. And part of what came out of that was an understanding of, you know, a mechanism by which Kestrix collecting its data can be legal and compliant. And yeah, without going into too much detail, I guess there could be some GDPR experts listening.

But basically, we capture the data we capture on the basis of what's called legitimate interest. And legitimate interest means like, there is a societal benefit, basically, to the work that you're doing. If you capture your data, data on the basis of legitimate interest, the rule is that that data has to only be used for the legitimate interests that you've defined.

So for us, it's retrofit, prioritization, planning, pricing, and engagement, right? It's all about making buildings warmer, cheaper to heat, and lower carbon. That means that if anyone uses our data for something outside that purpose, or for something nefarious, they are, we technically are in violation of GDPR.

So when we worked with customers, we make them sign a contract that says, you will not use the data for anything other than x, y, z, which means that if one of our I mean, first of all, we if we see something like that, we do not report it, we remove that house and we move on. But if our customers were to see that and use that as a basis to send the Met police to someone's house, they would be in direct violation of our agreement. So it is a funny, it's a funny thing where at the beginning, we thought, do we have an obligation to report this, but actually, we couldn't be doing what we were doing, if we weren't really strict about the use of the data.

And that's a really big thing for us. Like we, you know, there are ways you could use this data for an insurance sort of use case. But it's for us, again, we started this company because we want to help accelerate the transition to lower carbon, cheaper to heat, net zero homes, not because we just want to make money in whatever way we can.

So it's, and that's what's enabled us to sort of do this, we get basically an it's not an exception, but like not every company is processing on the basis of legitimate interest. So it's a very interesting question.

JIMMY:
Yeah, that's good. That's good. So do people get notified before you fly over that they send out a letter to all the residents?

LUCY:
So we leave that at the sort of discretion of our customers. Like a lot of customers that we work with use drone inspections to do sort of infrastructure inspection. And they, some of them have a policy on whether they notify their tenants or not.

I'd say, you know, most of the time, our clients do choose to notify. So yeah, I mean, we, we have a draft sort of resident engagement letter, which basically states, Hey, this is what's going on. There's planned surveys, drone pilots will be easily recognizable and have yellow vests, you can go and talk to them if you want to know more about the project.

And then we provide a link to our website, where you can understand our sort of privacy policy and go through a flow to sort of opt out, take your house out of the address list. Um, for context, we've done 10,000 surveys, we've had about five opt outs in the whole sort of history of the company. We've even had phone conversations with people where they call us, they say I want out, we explain them what they're doing what we're doing.

And then they actually are pretty happy and want to be included.

JIMMY:
Yeah.

LUCY:
But yeah, I think our our sort of approach to this is be transparent and give people agency. So if people really don't want to participate, they don't have to. But yeah, the the sort of, I guess the question of what does it mean to notify and give someone the ability to opt out, like I think is, it's ambiguous, right?

Like, I think we make it super clear on our website how to do it, we give our drone pilots always all the necessary information, ways to contact us. So even if you don't drop a letter, like it's a resident can find a way to sort of drop out. But again, because of this legitimate interest sort of point, we're not just capturing all this data and selling it onto the dark web or anything crazy like that.

Very, very careful about where it goes, who uses it, what they're using it for. We had to come up with some very complicated service data licensing agreements to make sure this works, which is a, it's a pain, but like it's it's part of being I guess, an ethical company.

JIMMY:
So yeah, it is important. It really is important. I mean, but it's not as though you're out there filming people and, you know, watching what they're doing and where they're doing it, you know, it's purely for their benefit.

And I'm sure people can understand that when they do learn what you know, you can't see through a window.

LUCY:
And for example, you know, we don't pass on the raw images to our customers, you know, it's a it's a 3d reconstruction of four walls. Yeah, sometimes you'll see the outline of a person that you can't, you can't recognize faces anyway. So it's because it's from 50 meters up.

And especially in the thermal image. I mean, really, the I'd say the most sort of risky situation is when there is evidence of a marijuana farm. But again, we have policies for that.

JIMMY:
So yeah. Okay, well, we mentioned you a startup. So let's, let's learn a little bit about your journey.

Where did it all start?

LUCY:
So it really, I mean, just in terms of background, any listeners can probably tell that I'm not from the UK, originally from the US was born and raised in Boston. But I've spent most actually all of my career this side of the Atlantic working on early stage, I'd say climate tech and climate tech adjacent startup propositions. So my previous role was at a carbon accounting, SAS company.

So software as a service, carbon, carbon accounting service, basically, where we worked with all sorts of businesses to understand their supply chain and calculate their carbon footprint, and then tell them basically what to do. And we had lots of customers who own buildings. And what I found is that buildings were really, really complicated.

So we'd, they'd come to us, they'd say, here's our energy bill. And our recommendation would be okay, reduce your energy bill. And they'd say, well, how what about, you know, insulations?

What about heat pump? What about solar, and we'd sort of say, sorry, you got to go talk to a building physics consultant, which are very expensive. So that was sort of always, you know, kind of, you know, something I was thinking about.

I'd also studied urban design and architecture as a sort of a secondary in university. So always been really interested in cities. I think cities are a fascinating place to think about impact in all respects.

I think something like 70% of all people by 2050 will be living in cities. So you can take a relatively small sort of surface area of the globe, which is urban centers, make your impact and then really affect a lot of change on a global level. So cities and in that, you know, the buildings in them, I think, always have been just an interest of mine.

So when I met my co founder, who is a sort of his background is a high tech product leader, he spent, you know, 20 years at different in different sorts of parts of big tech and in Vodafone, he most recently spent 10 years at Google as a product manager, but he yeah, grew up with his parents renovating homes around him. So he's had sort of a keen knowledge of building physics and how sort of structures work. And yeah, it was really when he's got two small children, and he wanted to do something about climate change.

He thought, I'll start I'll start with my own home. And he went to retrofit his own house and couldn't figure out for the life of him. Despite, you know, having this fancy engineering degree and growing up with his parents renovating homes.

Yeah, I couldn't figure out what to do to make it more energy efficient. So he was a as a hobby was a drone pilot. So he took a he took matters into his own hands, and he flew a thermal drone over house and thought, wow, there's quite a lot that you can see from this thermal image that my assessor couldn't tell me.

Now, I don't know who his assessor was, maybe he got not as good of an energy surveyor that day as he as he could have. But I think the point is that he realized you could see quite a lot from from the sky. And he had a background in computer vision.

So his first thought was, wow, I could take this image and train AI to see things that the human eye might not be able to or to spot patterns across lots and lots of different pictures. So when he came to me with that, sort of his, his idea was, gosh, I could sell this to everyone on my street. And we could have a direct to consumer surveying company.

But I'd spent so much time with these big enterprises who had lots and lots of buildings that I said, no, the trick here is that this is really scalable. We could fly this over lots of buildings, we could fly this over whole portfolios, we could fly this over whole cities and then sell the data to installers for lead generation. And that's really where it where it sort of started.

So the vision was always from the beginning, Google Maps of heat loss, you know, map lots and lots of homes and bring sort of this data set to bear on this on the sector. I think it's morphed much more into, we are a really scalable way for a housing provider to get lots and lots of data about a portfolio, which is still, you know, thousands of homes, but I would say our our goal eventually would be to move on from from doing thousands of homes to 10s of 1000s or hundreds of 1000s. We've just got to obviously start with one customer because it's hard, quite expensive to get the drones in the sky.

Not relatively speaking, but yeah, we have to become sort of the de facto data provider before we do that.

JIMMY:
Sure. Yeah. So speaking of the housing providers, who are the key users then?

LUCY:
So the key users at the moment are UK social housing providers who are looking for a faster, more reliable way to plan, prioritize price, and verify retrofit projects at scale. So UK social housing makes up about 20% of homes in the UK. It's between four and a half and 5 million homes.

And they, yeah, they have really aggressive targets to sort of become more energy efficient, reach that EPCC, different housing associations and councils will have different targets. And they've got some money, you know, there's the warm homes, social housing fund, wave three pot, which is just over 1.2 billion pounds over the next three years for social housing retrofit. You know, it's match funding, so they have to also fund their piece.

You have the warm homes, local grant scheme used to have eco the energy company obligation, but that was scrapped following some, yeah, some, some controversy involving audits of past installations of retrofit, which I'll talk about in a minute. But the point is, is there is there is funding available. It's just a question of where do you put it to get the biggest bang for your buck.

And that's really what we help what we help the sector answer. So say you're, you know, Clarion housing group, they're one of our first customers largest in the UK. They, you know, are retrofitting at a volume of hundreds of homes per year, moving up into sort of the thousands this year and the next three years.

And for them, the question is, where do we start? What do we do? And then once the installs have actually happened, are they really working?

And this is with respect to insulation, I think heat pumps and solar PV are slightly different. But yeah, that's, that's really the question. I'd say if we could distill it down, it's, how do you triage so that you pick the right properties?

And then how do you verify so that you know, the work's been done correctly? Those are really the two main value propositions.

JIMMY:
Yeah, so what Yeah, so what you're saying is that it helps with their budgets and helps them to keep under budget rather than just piling out and doing a load of houses at once you need, they need to pinpoint exactly where needs doing first and you help them do that?

LUCY:
Well, yeah, it's really an efficiency question. You know, if you need to retrofit 100 homes, you know, you may need to send retrofit assessors to

to find the right ones. Wouldn't it be better to just put a drone over 200 or even 140 and be sure about the ones that you want to do and then only send the retrofit assessors to the ones you're sure you're going to do. You just end up saving quite a bit of money even though you have to make an upfront investment in the drone flights.

It's much cheaper to scan a home with Kestrix than it is to do a retrofit assessment. So it's basically precision retrofit and not to mention you save a lot of time and resident disturbance if you have a plan before you start.

JIMMY:
Yeah, exactly that, exactly that. Can you, are you able to give us like an example of a typical case study? Let's say for something like Peabody.

I know of Peabody and I have a friend who's a surfer. Have you got like a typical case study that you can give an example of?

LUCY:
Yeah, sure. So we actually, we put out a case study with Peabody in the beginning of the year and this was the first time Kestrix verified the outcomes of retrofit. So idea here, maybe just to go back with a bit of context, why verify retrofit?

You know, it should be, you put insulation in, you assume it works. Unfortunately that's not the case. There are a whole range of different types of contractors and installers.

There are some really good ones that do a really good job all the time. And there are some that are not so good. And the reality is, is that in the sector, there's not really a good way to check if the work's been done well.

You know, you can do a blower door test, you know, after the fact, you can do a visual inspection, but in reality, there's just too much work going on to do these inspections for every home. So the sector unfortunately has been plagued by a sort of a lack of accountability. So yeah, this really came to bear in sort of the middle end of last year.

The National Audit Office put out a pretty scathing report showing that 98% in a survey of exterior wall installations that had happened under the energy company obligation funding stream had some sort of failure mode. So we're talking kind of thermal bridging, water ingress, problematic insulation installed, which leads to lots of knock-on effects for the resident, you know, higher bills because it's not working, not to mention damp and mold risk. You know, lots of problems can come from this.

Now the survey kind of, there's lots of perspectives in the industry on whether it was a totally unbiased assessment. You know, they sent a mail out as offering surveys. And of course the people who responded were probably the ones who were upset with the quality of their install.

So there might've been a bit of distortion, but over under the point is that the work's not being done correctly in a lot of instances. And that's because there's no accountability in the sector. So Peabody sort of approached Kestrix and said, hey, can Kestrix be used as a cheap, scalable, uniform way to assess that the job's been done right?

And to not only assess the job's been done right, but to actually understand quantitatively what was the value, what was the uplift of an energy performance measure? So we worked with Peabody to scan a subset of the properties they retrofitted as a part of 2.2. So we scanned the properties before they were retrofitted. And then Peabody did not tell us what measures were installed.

We flew the drones over after. And basically we identified that loft insulation had been successfully installed and that solar PV had been put on the roof. And we did this using our rapid thermal performance assessment algorithm.

So algorithm that understands basically what is the heat loss quantitatively that's coming from a building material component. We also correctly identified that nothing had been done to the walls. So yeah, the outcome was that Peabody can trust Kestrix for verifying loft retrofits.

We also could, by understanding the U values before and after, we could estimate the predicted energy usage throughout the year and therefore the energy bill savings. So combining solar and insulation, there was an average savings that we estimated of about 520 pounds per year per home. So that's the equivalent of buying a smart TV, which is what we compared it to.

And yeah, an estimated average savings of CO2, assuming the sort of grid mix stays the same, of about 1.3 tons per home per year, which is the equivalent of driving about 5,300 miles. So a lot of that was solar, a little over half coming from just installing a solar panel, which is not something we had to calculate. We just detected a solar panel and then calculate savings from that.

But it did also come from our detection of insulation in the walls, or sorry, in the loft. But I think what it shows, what it goes to show is that you could do this across thousands of homes and track year on year, how are outcomes actually changing for residents as we invest more and more money in retrofitting homes. And I think that's really important for resident engagement.

It's really important for sort of communication back to policymakers. If Desnes is investing this money in energy efficiency improvements, what is actually the impact for real people? I think in a world where people don't really understand the connection between net zero and their own lives, like this is the kind of stuff that can really help illustrate that and bring that to life for people.

So, yeah, I mean, I think that was, I'd call it more of a business white paper, to be honest, than a case study, because at the end of the day, there's still a question of, you know, what's the economic benefit for Peabody of validating retrofit outcomes? Yeah, I mean, I think if we think about contractors, there's the possibility of pay for performance. So basically only get paid a bonus if your work is done correctly.

We can, we really feel that we can be the data layer for that decision making, which can just get a lot more sort of value flowing into the sector for good work. Yeah, we hope that we can play a role in sort of playing like the evidence base basically for net zero around insulation.

JIMMY:
I imagine it can go further than retrofit as well. So I imagine that people can use the data that you've given now for pre-built purposes, like for planning delivery to make sure things are done properly, built properly to begin with and going forward. So you get that energy saving right at the start.

LUCY:
Yeah, and I mean, I think that's another thing is like we do tout ourselves as the Google Maps of heat loss because we're on this mission to get insulation into more homes. But the data that Kestrix shows is definitely rich asset data. And I know that the audience is probably mostly construction focused.

We're talking high resolution 3D models of a house where you have a quote level sort of accuracy measurement of all the windows, walls and doors. So like going to Kestrix and looking at a building on Kestrix before you go to site has to be, I mean, I would imagine helpful for more than just understanding heat loss. And similarly, thermal imaging, while we don't claim to spot instances of damp and mold, we're not inside the property.

You can see water ingress in a thermal image, depending on where you're talking about and how severe it is. And yeah, I mean, I think as we map more and more homes and have a higher penetration of sort of data available, I definitely see us working with anyone doing construction on the outside of the house, because you can see the roof, you can do a roof inspection, you know. Some data is available, like if any listeners are interested in checking this out, I think we're also keen to explore whether we can help other sides of the sector with this high resolution data.

JIMMY:
Yeah, sure, sure. Yeah. I recently interviewed someone who works in the passive house area, which is obviously like the standards for this type of area.

So I suppose in a way you can help clients move closer to passive house style performance, even if they're not aiming for full certification.

LUCY:
Yeah, I mean, passive house is really interesting, right? Like, I think there is, obviously, if all houses could be passive, not only would we all have lower bills, but the topic of energy security would really come up because, you know, if you're using so much less to heat homes, you're not as reliant on gas. The grid is less strained, you know, when we talk about AI and data centers becoming an issue.

You know, if I were in charge, I would make every house passive. But the reality is, is we don't have the funds to make every house passive. So passive house is kind of the ideal.

There are only a handful of them, I believe, in the UK. Really, what defines a passive house is, you know, it's an internationally recognized standard, meaning, you know, the energy is ultra low in a building because it's designed to not require heating or cooling. It's, yeah, I mean, it relies heavily on extremely low U-values.

I think the relation of Kestrix to a passive house is that we estimate U-values without a reading of internal temperature. So if we flew over 100 homes, and one was a passive house, we would be able to tell you and sort of independently verify that. That said, I would hazard a guess that if a house is passive, the owner or retrofitter probably has the funds available to do a full on sort of blower door test or on the boots on the ground thermography assessment.

Which, yeah, I mean, we're pretty transparent about this at Kestrix. You know, a boots on the ground thermography and blower door test is going to cost, you know, 400, 500, 600, 700, maybe more pounds, depending on the size of the house. A Kestrix scan is sort of sub 100 pounds if you're doing them at scale.

So I mean, yeah, even sometimes well below that. So we are not going to get probably as a perfect U-value. I mean, if anyone claims to have a perfect U-value, they're wrong, because even in the lab, there's plus sort of minus 6% accuracy on a U-value.

But yeah, I mean, we can help verify how close a home is to passive grade, for sure.

JIMMY:
Yeah, yeah, yeah, yeah, definitely. I mean, the guy I interviewed reckons that everyone should be working to a passive health standard anyway. And like you said, you know, you feel that that would be great, an ideal word for all houses to be like that.

But I think, yeah, as you say, it's a bit of a luxury at the moment.

LUCY:
Yeah. And I mean, this is this is the thing. It's, if we think about retrofit, and just the sheer scale and number of homes that need to be treated, it becomes a question of, do you treat a very small subset with a sort of very deep retrofit?

So we're talking not just insulation, but heat pumps, energy upgrades. I mean, there's a whole debate going on right now. Can we just put solar on all the roofs and ignore sort of the fabric?

I personally think there's a lot of problems with that, because we're not just trying to pursue net zero, we're trying to pursue warmer, healthier homes. And that is about insulation. But yeah, I mean, doing a retrofit to passive house standards is a major undertaking.

You know, if you do a three bedroom home in in the UK, you're looking at anywhere between sort of 800 to 1500 square meters. So it could cost between 40 grand and 80 grand. You know, if your home is only worth 300,000 pounds, it just doesn't make economic sense to do that.

And with 40 grand or 80 grand, you could do an EPCC retrofit for I don't know, three, five, seven homes, which just when when we're talking about sort of cost versus volume, it just it doesn't really make sense. I'd love to be proven wrong, like if we can find a way to make passive house scalable, then castrix would become very important, because then we could verify all the passive house claims. But yeah, I mean, I think that's, that's the struggle, right, is to do something really, really well, you need a lot more money.

And, and that's just not a luxury we have. So. But yeah, I think castrix is trying to solve that problem just for the survey, right?

Like we have very limited time and limited resources, limited labor, how can we industrialize certain processes? And, you know, we have friends in other parts of the supply chain, Bunda house, for example, are trying to industrialize, you know, EWI, and, and insulation, you know, they're actually making the insulation, we're trying to industrialize the survey. And that's really what you need is like, that's how technology can help this is we have just this crazy scale of work to do.

You've got to use tech to help it happen faster and with less money.

JIMMY:
Yes, gotcha. Gotcha. Let's let's talk about AI, use a lot of AI.

There is a bit of a stigma attached to AI in certain areas. It's really coming of age now. How do you how do you build trust in the data and the conclusions that come from it?

LUCY:
So I think that there's a lot of misunderstanding about what AI is and isn't in every sector, but but also in construction, I think we have a lot of customers who want to use Kestrix just because it has an AI, the word AI in it. And then they expect that we're going to have a chatbot. There are chatbots and chat to bt and Gemini and Claude all exist and are making waves in the construction sector.

But Kestrix is not using, I mean, we're using them internally, but we're not, we're not using, you know, chat to bt to produce outputs. So I think, yes, you can, you can ask the question of can we trust chat to bt to tell us what to do on a construction site? When it comes to Kestrix, though, we are building AI with real data that we're collecting for our clients.

So any AI that we build is trained on thermal images that we've captured of real buildings. There's no kind of chat to bt making things up. It's the way AI works is you put data in and get insights out.

And for us, actually, you don't really see it as much on the customer side, you see it more in our 3d models. So it's very complicated to create a 3d thermal model of a building. Well, we've now done it 1000s of times.

So our AI can create that much faster, which means that it's much cheaper for our end customer to purchase than if they were going to purchase that from a design studio. You know, a 3d thermal reconstruction of a building can put you back hundreds, if not 1000s of pounds. With Kestrix, you know, we do this, we do this at scale, it's taken us a long time to develop that not just a long time, but a lot of data.

Again, though, that's, you know, we're talking about computer vision, our computer vision algorithms, you know, we work with different models, but, you know, ours combined with the off the shelf models we work with, can spot what's a window, what's a door, what's a, you know, what's a roof, this is the sort of thing where, okay, if we are incorrectly classifying a window as a door, obviously, that's going to lead to problems for our, for our retrofit recommendations. But, but that's the sort of challenges we're facing when we talk about AI.

And I guess, yeah, I mean, the the sort of view value estimation component is only as good as our ability to estimate what's a window in a thermal image, and what's a window in a visible spectrum image, and then make sure that they're overlaid on top of each other. So all the information is available. But yeah, I mean, I think, beyond, can you trust Kestrix's AI or not, is the question of can you trust Kestrix's methodology for estimating heat loss?

And yeah, we're just about to start some work with an academic institution on this, because it's as much a sort of thermography and building physics and machine learning question as it is just a general AI question. Yeah, and we're, I mean, we've, we've come out with early results on our accuracy. So if we think about our sort of ability to classify, is there insulation or not, which determines whether we make the right retrofit recommendation.

We've run tests on sort of solid wall properties, understanding is there insulation or not, in a solid wall property, we're, yeah, when we compare to ground truth data, we've got a sort of 92% classification accuracy, that was just on a 200 at home sample. And the instances in which we did not get the right answer, there were trees blocking the drone image. So again, it's not like our AI is gonna pretend a tree's not there and make a guess.

And we don't know, it's more like, can it read the image correctly? And yeah, we're working on verification of that sort of all the time.

JIMMY:
Okay, are there any limits to the technology?

LUCY:
Yeah, there are lots of limits to it. I mean, we are not going inside the house, and we're not making things up about the inside of the house, we could say this is a fully, you know, a full scale solution, and estimate things about inside, but we, we don't, I mean, we estimate internal temperature to estimate our U values. But like, we're not going to be able to get the, you know, the type of light bulb inside the house, which has an albeit small impact on the EPC.

We don't have access to the information on the floor insulation, which can definitely impact energy performance. But that's exactly it. We're not trying to replace a survey, we're trying to get 60 to 90% of it done so that a person can go faster.

And, and we, we can avoid doing surveys, we don't need to do.

JIMMY:
Yeah, sure. So yeah, so that is where you need the people on the ground as a follow up. So like you say, you're not going to lose them anyway, in the future, I'm not going to replace them.

Yeah, okay. So we've got a little bit time. So should we do the screen share now?

LUCY:
Sure, yeah. I mean, I won't get you through the whole platform. But I think just this could help sort of bring it to life.

JIMMY:
Yeah.

LUCY:
Shall I share my screen?

JIMMY:
Go to more.

LUCY:
Yeah.

JIMMY:
And then share.

LUCY:
Yeah, I'm just gonna pull this up in a separate tab. Great. Can you see my screen?

Perfect. So yeah, I mean, on the screen, you can see an example of one of our 3d models. Like I mentioned, we start with a high resolution, what's called RGB, or red, green, blue drone image, this is captured during the day.

Again, there's not really any AI in this, this is just images stitched together. So this isn't a fake house or anything. So real, real place.

You can sort of zoom in and, and see quite a lot, you know, about the roof and different things that you wouldn't be able to see from the ground, which is just useful. Then you can look at sort of quantities. So if you want to see different roof sections, you can understand sort of how big they are, what they are.

So that's a that is AI detecting that this is a roof, which you see sort of apart. We're going to have you values filtering in for these in the interface very soon, as well as material classification, because material impacts sort of emissivity. So you have to know that.

Yeah, and then you can do sort of the same thing for walls, you can do the same thing for windows, parts of the window. So this is just useful for, you know, construction planning, I suppose. And then with the same model, you can flip into thermal.

JIMMY:
Oh, wow.

LUCY:
This is where you really start to see stuff that you don't see on the outside. So yeah, like this roof, for example, a lot going on in here. Yeah, you know, we detect a roof defect here, which is pretty obvious.

We detect instances of thermal bridging. Yeah, we don't pinpoint watering grass, but you can see that a lot, especially if you're a trained eye. But the nice thing is you can flip between different thermal views.

So this one's a bit not as easy on the eyes, but it is, you know, you get more granular about what you're seeing. And then you can go back into the normal sort of view. So you can see things in context, which can be quite helpful.

Yeah, and then per view, you can basically click on different elements. In this case, we've got some poorly performing double glazing. And then we give a retrofit recommendation if applicable.

So in this case, this recommendation to replace window with a high performance double glazing. And then we list basically what the impact would be. So that's based on our estimated energy performance of this component.

How much could you save? So yeah, I mean, I think going back to the RGB version, last thing that's pretty interesting here is we've got a what we call a flash heat pump viability assessment or feasibility assessment. If you click on this, we know not only the size of all the windows, walls, and doors, but we also know the distance of the windows from the neighbor's windows.

And noise, for anyone unfamiliar with heat pumps, is a big part of what determines whether a heat pump can go on the side of the house or not. So it's noise and then size of the heat pump, which is based on sort of the thermal performance of the house. And then there's also a location.

So is there actually a spot on the wall for it? So what we do here, as you can see, it's a bit small, but there's a little red box here, which signifies a heat pump can go here. This is where it would go based on the energy performance of the house that we estimate.

This is sort of the heating capacity and the size of the system that you would need. So we're just updating our database now to include sort of all the different models of heat pumps that are available in different parts of the UK. We also take into account conservation area constraints when we talk about retrofitting the house, including with heat pumps.

And yeah, we're going to have a next step is a solar PV visualization module. So it's kind of just like a full flash energy assessment for a building.

JIMMY:
I love it. I think it's brilliant. I really do.

LUCY:
Nice. So glad. And yeah, if I had more time, I would show you sort of our whole dashboard.

Yeah, no, sort of does our EPC comparator and stuff. But I think for now, we Yeah, if anybody's interested, they can they can come back and talk to me.

JIMMY:
Yeah, yeah, sure. I do. I really love it.

I think, like I said, at the beginning, I love a success story. Actually, I recently wrote an article for a magazine on construction technology and machinery. I wish I would have known about Kestrix when I wrote it, because I definitely would have included that.

LUCY:
Thank you. I hope. Well, a success story is we're not there yet.

We're working really hard to make this technology. I mean, it is working, but to make it really, really robust and something that could really become sort of the data layer underpinning all these decisions, you know, we have to earn the trust of the sector. So we're going at the pace that we need to working with the largest housing associations and the contractors and really getting their feedback.

But yeah, I hope we're a success story when we've when we're starting to influence every retrofit that's happening, which I think we're, we're on our way, but we definitely like to stay humble. And we're not saying that this technology is going to replace everything tomorrow. But um, yeah, we want to help the construction industry take advantage of all the efficiency gains really that that technology can can bring.

JIMMY:
I'm sure you will. I'm sure we will. I think you've answered this question, but already, but what is the bigger ambition?

LUCY:
Yeah, I mean, I think we really want to become the data layer underpinning the transition to energy resilient buildings, right? Buildings, like I mentioned, make up 25% of CO2 emissions. If we just had a blueprint for how to make them all more energy efficient, I think it'd be a hell of a lot easier to fix them all.

You know, I think the first step along the way to that is sort of being embedded in retrofit decision making for those who own the buildings. But if we think about the social housing sector, it's only 20% of the UK, which is a relatively small market, you know, I'm from the States, I'd love to see drones flying all over the place and collecting this data and it being used for planning on all fronts. I think that we'd love to understand how once all of our data outputs are sort of recognized by the regulators, or sorry, recognized by institutions, understand how we can work with the regulators and those who sort of write the EPCs.

Again, we don't want to replace the EPC, but if Kestrix flies over, our algorithms are verified, and we indicate that a bunch of the data is out of date, it should be easy to just update it and not have to send somebody. Once we get to that point, it's very, very sort of economical for our customers to use us at real scale, which would unleash the drones, let us sort of fly all over cities and talk about millions of homes instead of thousands. So that's where we're going.

But we again, we build with the partners who need us first. And that's what we've been doing with the sector.

JIMMY:
Well, we definitely need it. We're in a bit of an economic crisis. You know, everyone's fighting in their purse strings at home.

So it's definitely something that we need. So well done.

LUCY:
That's the hope.

JIMMY:
Yeah. But that's all the questions for you. Do you have a question for me?

LUCY:
Oh, I don't know. I guess. I'm curious if you are more scared or excited about AI when it comes to the construction center.

JIMMY:
Construction center. Okay.

LUCY:
Not center construction sector.

JIMMY:
Yeah. So I think I'm more excited. So I did have.

So at heart, I'm a creative writer, right? So I write creative writing, also write publications for magazines. And probably about a year ago, this is all to do with like the open AI stuff.

So I was dead against it all. I think it lost all the human element of it. But I've kind of embraced it now.

And I think in the construction terms, which is a little bit different. I think it's, it's a really, really, really useful tool. And I'm excited.

But I am also a little bit scared. Because I don't think we're going to get to what you call it. What's that thing on Terminator?

Forget what it's called now, what the company was called now. Anyway, I don't think I don't think we're going to get to that stage where it all takes over the world and kills everybody. Because I think now because it's trained, as long as we it's like a child, basically, if you if you, you have to train it and educate it and treat it well.

So if it learns that behavior, then I think we're all going to be alright. But in terms of the construction and the tools it can use in the construction industry, I think it is a really important tool. And it can save a lot of time and money and give a lot better results than what a human brain can.

LUCY:
Hmm. Interesting.

JIMMY:
Yeah. So there you go. What do you think?

LUCY:
I mean, I think for the construction sector, there's a lot of efficiency gains to be had. I think more broadly, I mean, this is me speaking for myself, maybe rather than Kestrix. Kestrix doesn't have any particular defined political stance.

I'm concerned about sort of the state of the world and what's motivating world leaders. I don't think that world leaders are taking enough measures to ensure that it doesn't get out of control. And there's lots of ways it can get out of control.

I mean, I think everybody thinks about Terminator style, you know, AI turning against us. But what we first have to consider is our resource constraints. I mean, you know, the energy grid, the electricity grid is already strained as we transition to EVs and heat pumps.

If we don't invest in the grid, and all these data centers come online, we're going to be seeing blackouts. When there's blackouts, you know, in many places, hospitals have generators, but in many places, they don't. There's just not really any checks going on.

And those running the AI companies are all in a race, and are all obviously very profit driven. So if you don't, yeah, put measures in place, you're not going to end up with the best outcome for everybody. It's an example, I think, of where capitalism needs to be checked.

JIMMY:
Absolutely.

LUCY:
I'm not totally confident that leadership around the world is concerned with that. And I think also countries are competitive, right?

So Trump may not care. You know, the UK may, may be more sensitive, but there's also this worry of falling behind. And you see this all over Europe.

I think there's this tension between we want to do the right thing and regulate, but also there's a lot of concern that Europe will fall behind in doing that. So it's hard. I just, yeah, I think particularly the US needs better leadership.

JIMMY:
Absolutely, absolutely. And I totally agree. And I did do a summit on it a while, a couple of years ago, just telling everybody to slow down with this until we can fully understand it and control it.

And like you say, it is a race, because obviously everyone wants to make life easier and more profitable for each other and themselves as quickly as possible. But the thing is, with these people in power, you know, and, and also, I'm probably going off topic a little bit, but also people with loads and loads and loads of money, they, I think they probably, they tried to find something else to do, like they need an agenda. You know, they made all their money, they've got all their power.

So what can I do now? And I think that can be a little bit dangerous, particularly with all the money and the power that they've got. I just think they just need to be a little bit more careful with what they're doing with it all, personally, when it when it concerns human nature, the world in general.

Yeah. Yes. Good question.

Very good question. Yeah. I wasn't expecting that one.

You're all done. Thanks very much. It's been really insightful.

LUCY:
Yeah, it's been a pleasure. Really interesting conversation.

JIMMY:
Yeah. Where can people find you and Kestrix?

LUCY:
So you can find Kestrix by looking us up online at www.kestrix.io. You can also connect with me on LinkedIn, Lucy Lyons, I think my name should be probably in the title. I'm pretty easy to find. And yeah, really keen to have feedback, perspectives from others in the sector about the technology we're building and the way we're approaching things.

Yeah.

JIMMY:
Nice one. Thanks for your time. Enjoy the rest of your day, whatever you're up to.

Yeah.

LUCY:
Thank you. You too.

JIMMY:
Cheers.

LUCY:
Bye.