Episode 12: Sustainability, Due Diligence, and Roadmapping with Selfridges

Podcast description:

In this episode of the Discovered Podcast, recorded during our Discovered Fringe event @ Ecommerce Expo, host James Gurd leads a thought-provoking panel discussion on the challenges and opportunities in building sustainable e-commerce systems. Featuring insights from industry leaders like Sezin Cagil (Selfridges), Carole Breetzke (Brave Bison), Andy Eva-Dale (Tangent), and Rachel Tonner (Klevu), the conversation explores topics such as composable commerce, the business case for sustainable tech, and the role of AI in driving efficiency and innovation.

The panel dives into practical examples, including the importance of benchmarking carbon emissions, the power of modular architecture for adaptability, and the risks of AI implementation without proper guardrails. Tune in to discover actionable strategies and forward-thinking perspectives for navigating the fast-evolving digital commerce landscape.

Speakers:

• James Gurd (host) – Owner & Ecommerce Replatforming Consultant at Digital Juggler, Host of Inside Commerce Podcast
• Rachel Tonner – VP of Marketing at Klevu
• Sezin Cagil – Principal Delivery Lead – Digital at Selfridges
• Carole Breetzke – Business Solutions Director at Brave Bison
• Andy Eva-Dale – CTO at Tangent

References:

• www.asklo.ai

Podcast transcript:

[00:00:15] James Gurd: What are the challenges in adopting sustainable system design now that the ecom businesses keep up with the pace of technical change because they can all agree it’s pretty rapid right now? So let’s start with Sezin. Over to you.

[00:00:27] Sezin Cagil: So hello, I am Sezin, Principal Delivery Lead in Selfridges. I come from a development background, and I have been managing delivery teams for over 10 years now. I’ll answer it from a non-tech ecommerce side rather than the tech ecommerce side. I think there are big differences. I think from non-ecommerce side, it is mainly the lack of technical know-how and vision. We jump onto the most exciting idea right away there and then without really realizing the impact of it on the business. There is also, of course, additional burden of legacy systems that we need to get off of, which is very expensive. It doesn’t really have any of those projects that are starting.

Also, I think to be able to get to a sustainable roadmap, you really need in-house expertise. But again, non-tech ecommerce is not very good at building technology teams, and I think that is the main thing that they need to invest in, have their own in-house technology teams who understands their businesses really well and who then continues to collaborate with the wider business so that the decisions that comes from that cycle are sustainable for the long term of the business.

[00:01:32] James Gurd: Excellent. Thanks. Carole, let’s go to you next.

[00:01:34] Carole Breetzke: Yeah, I can go next. I’m Carole Breetzke, I’m a Business Solution Director at Brave Bison. Brave Bison is a marketing, media and technology company, and as part of what we do in commerce, we help businesses select the right solutions for their requirements. So I think one of the main challenges navigating the landscape today, because like before, when you you had to do an e commerce website, you could buy into a suite, the landscape would be very simple, and you would maybe pick a few external partners.

But now we have much more complex systems where there’s not one system at the center of the architecture. You have to select a lot of different vendors and the sustainability impact is much more difficult to assess because you got the cost of infrastructure. But you have also the cost of the integration between all the different systems, which is a bit more complex to identify because it’s very opaque.

So I think for businesses I would say don’t try to keep up with the change, the technical change, try to keep up first with the pace of change in your business because as a retailer, you should be the expert in your business. It’s already super difficult to keep up with the change in social marketing, the change of purchasing habits that are being led by AI or by new technologies. But to keep up with the change of technology, just use the right vendors because your technology partner should be the one that have to keep up with the change of technology, and they have to keep up with AI. So I would say just collaborate with people that can have their eye on this technical landscape that is changing.

[00:03:02] James Gurd: Thanks. Let’s go to you next, Andy.

[00:03:04] Andy Eva-Dale: Hi, I’m Andy. I’m an engineer by heart. I’m currently the CTO at Tangent. We make lots of digital products for big enterprise organisations like SAP and Reed. I would say from a sustainable point of view, awareness around impact of our systems is probably one of the biggest challenges. Like, we need to start benchmarking our systems. I don’t know if anyone saw recently, but Google announced they had a 48 percent increase in carbon emissions and like the web itself and ITC has long outstripped the aircraft industry for carbon emissions. I was hoping a plane might go by as I said that, but not so much luck. So I think first of all, it’s about benchmarking your current systems, realizing how dirty they are, start reporting on carbon, then have that as a key metric for success. So if you’re looking at your releases, benchmark carbon, try and see trends going down. Greenfield projects have it as a KPI for success, and I think that’s one of the biggest challenges is awareness and then using that metric to drive your future thinking.

[00:04:02] James Gurd: Thanks. Rachel, back to you.

[00:04:05] Rachel Tonner: What was the question?

[00:04:07] James Gurd: It was about what are the challenges?

[00:04:09] Rachel Tonner: Yeah. My name is Rachel Tonner. I’m working with Klevu, AI search and merchandising platform. I think one of the challenges with building a sustainable ecommerce platform, I’m going to build on what you said before, actually, about like it’s good for the technologies for you to have, like the technologies in there, the third party systems who keep up with the trends for you, because I think as an ecommerce manager you bring with you your experience, but also because technology moves so quickly, sometimes, your experience is going to be outdated and actually being brave enough and self aware enough to maybe know that as you’re coming into a new business and making sure that you’ve got the right technology partners with you to have the best technology stack that you can for what you can afford at the time. But yeah, I think we all come, our experience is bias, isn’t it? So yeah, being brave to make sure that you’re checking yourself and checking your decisions.

[00:05:03] James Gurd: I think that’s an important thing, because bias is in every business, no matter what size, and it’s important to get that. So we’re going to go through different topics, and you each got roughly two minutes, so there’s electric shocks if you go over time, so we can keep to the end. Don’t worry, it’s all legitimate. The first one is the business case for sustainable tech. So the context of this is that it’s obviously been a tough few years trading, a few people alluded to earlier, macro, micro pressures, job market’s hard as well, and people have been fighting for budget. So being sustainable, whether that’s in like future proofing tech or the environmental side, is the right objective. How do you persuade leadership teams that investment is merited? So the question then, the business case for greener systems design, how does it foster future readiness in the digital landscape? Andy, let’s start with you.

[00:05:45] Andy Eva-Dale: Sure. I think the business case is simple. If you adopt progressive technologies and architectural approaches to building out your systems, be it composable systems, design, event-driven systems like caching on the edge or static site generation. You can see three intrinsic benefits.

The first one is performance. Your end user will have a much quicker performance from a global perspective. Secondly is cost savings. You’ll see cost savings as well from your cloud hosting perspective. And lastly is carbon. So those three things are intrinsically linked. And if you think about it, how like modern cloud systems work, They’re all costed on data in, CPU usage, or GPU usage of its AI and memory.

And all of these things take energy, they take time to process and they cost money. They cost money for you and your cloud providers. So if you take the right approaches, you can reduce the time it takes for users to use the systems and have responses back from the systems and the energy and cost of the back end systems in respect of themselves.

A good example of this is a project we did for UK Power Networks. We took them from a monolithic application to a composable application. We saw hosting costs drop by 66%. We saw the average performance for a user drop from four seconds to sub two seconds globally, and we also saw carbon emissions on the home page drop from 1.44 grams to 0.22 grams. That’s hard to quantify, but UK power networks look after the power infrastructure for the southeast. When Storm Eunice happened, they had lots of people request making power cut requests. That’s when the top came off the Millennium Dome, if you remember. They had 3.2 million users hit the site on one weekend.

The carbon savings on just the homepage alone were the same as nine flights from London to Copenhagen. And most users were actually going on three or four pages. It would be four times as much as that. Another intrinsic benefit as well was developer experience. Developers like using these progressive technologies. You’ll have quicker development times, better retention in your development teams. So like the business case for me that does really stack up.

[00:07:52] James Gurd: I love that last point about retention, because it’s such a competitive market to find really good engineers and keep them for long periods of times. So Carol, let’s go to you and maybe you can build on about how you work with merchants to help validate the business benefits of this investment.

[00:08:06] Carole Breetzke: Yeah, I’d say the business benefits unfortunately, it’s still a little bit weak, like sustainability is not top of mind for retailer. And I would bounce back on Andy saying that performance is actually like what translates sustainability in an ecommerce experience because just being more sober with the experience creates, of course, sustainability, but creates performance, which is what retailers are looking at primarily. And I think Google did a great job at penalizing companies that have bad core web vitals because it pushed websites to really be more careful.

It’s like video content, size of image, and this has a side effect, really like improving also the carbon footprint of websites. So I would say this is key and I think education is getting there, but through performance rather than through trying to achieve green credentials.

[00:08:57] James Gurd: Let’s get into some terminology because our history loves terminology. Composability. Thankfully, there’s more of a focus. This sort of strategic lens now by the obsession, which is talking about headless has seemed to have dropped off, but the link to composability and sustainability and agility, just throw some more buzzwords out there. But let’s talk about what that really means.

How can composable commerce drive and support sustainable growth and innovation in ecom? And then we’ll come to you about part of the choice on this. So Carol, let’s go back to you actually because you work with lots of merchants. What is your take on like composability and its impact on growth, innovation, technological change?

[00:09:30] Carole Breetzke: So composable commerce is basically using like a modular architecture for your business. So you can achieve, basically the same thing that you would have with a monolithic application or with a composable system. The only difference is that now you have modules that you can easily replace. You have less vendor lock. You can make different systems compete to one another. You can easily swap. So because of that, there’s just more competition in the ecosystem, which is just healthier for the ecosystem. It thrives more innovation. And I would say it also allows you to have an architecture that can sustain the change, the path of change because we discussed this morning that’s now with AI the ecosystem is going to change even faster than in the past 10 decades. So how can you know that the vendor you’re selecting today are going to be actually catching the right train with AI and everything. So just having an architecture where you can swap elements, so that you don’t have one system that is the heart of your business. You can more easily choose your bets and you don’t have to select one vendor that is gonna be critical to your business.

[00:10:31] James Gurd: Yeah, I think most of the business I work with, everyone talks about agility, but actually agility when you break it down, ends up being speed to market most of the time. Obviously there’s elements of cost, et cetera in there. And most of that is down to the development cycles to then rapidly launch new capabilities. And if you have a like, massive backend application that has really started to develop upon. Lots of teams end up their roadmap gets shifted more or more backwards, which is why I’m seeing a lot of people make these changes.

So let’s come to you, obviously, Klevu, working with lots of different merchants. Can you build on this? Because part of the choice is important, but let’s split this on that. What questions should people be asking potential vendors about this in terms of like, how much they are able to support the sustainability of it, but also that speed of innovation.

[00:11:10] Rachel Tonner: Yeah, you mentioned like speed to market. It’s also like speed to value, and I actually think we’re all talking about the same thing. And I was loving what you were saying before about like the website having lots of people on it. Maybe if they’re viewing four or five pages, it’s actually worse for the environment. From ecom performance point of view, you and the customer want to get to what they want as quickly as possible and like with, as little friction as possible. It sounds like once you get that it can actually help from a sustainability, but also a speed to market, time to value.

So I think choosing the right partners to be able to achieve those goals is obviously key. But the questions you need to be asking them, especially if they’re like an AI vendor, it’s what is your algorithm? How long have you had this? Making sure that you’re really like sense checking the technology that they have. And if they’re claiming that there is an AI innovation in there, making sure that you’re really digging into that. What is the USP on that? How could I build this myself if I wanted to? They shouldn’t be able to answer that question for you. That kind of thing. But yeah, also checking the conversion stats, how will this bring me value? Things like, AI search, merchandising, personalization, those things are relatively well mature and well used. And that’s a really safe implementation of AI and also to get to that conversion faster. But there might be other things that are slightly less proven. So making sure that you get that do that due diligence.

[00:12:34] James Gurd: Yeah, I think the credible evidence is really important and understanding methodology. And I think we can all be guilty at times of in haste wanting put something in place and not necessarily think about the methodology, which those case studies are built on. So I think you’re right. Pushing on that is really important, but sometimes you do need a leap of faith as well, because there is amazing tech coming on the market and it doesn’t always have existing case studies of big merchants proving it. Sometimes you’ve just got to, you’ve got to take that leap of faith, but with the right level of due diligence to make sure you’ve understood the credibility of the tech.

So let this lead on nicely to the next question, which is about balancing innovation with scrutiny of vendors. Because you don’t want to be stuck down in insane RFPs that take months, and you can never get into market on things, but equally, you don’t want to just stick money into anything and hope it works. As we’re evolving into low code, no-code ecosystems, it’s quicker and easier to pilot, therefore, we’ve got to make sure we make sensible decisions, and speed isn’t always about success, doesn’t always lead to success. Sezin, I’ll come to you first on this. What steps could brands take to improve due diligence around the tech selection?

[00:13:33] Sezin Cagil: I think it is really important to understand the business problem. Get your business problem that will give you the constraints that you need to build on. And those constraints you can then take to a partner or the available tools and solutions and see if it actually fits to your business problem.

I think it is also really important, we have been talking about partners, to really run POCs. Most of the modern partners are very willing to work with development teams. Do run small POC, I can’t emphasize the importance of this enough. It will also give you the insights to the partner that you will be working for years to come and choosing the wrong partner is a killer for an ecommerce who is not technology durable.

[00:14:12] James Gurd: I really like the proof of concept approach and and I also set expectations as well. And I think, some partners are really good at that. You guys are about talking to people about to run an effective POC often involves cost to pay for it. Sure. You can’t just get everything for free, but making sure you budgeted for that is critical. Rather than going to somebody directly, let’s say we’re running a POC and then actually we don’t have the budget to do this. And also there are hacks and workarounds. I’ve worked in the business where they had no bandwidths in their day and their development team to get something through the developers to say, no, we’re not doing this. So they use their A/B testing platform to basically build entirely out of the front end code, proof of concept and run it as a test through the platform rather than going through the development rate. So sometimes you’ve got to be cute around how you get POCs over the line, even though developers wouldn’t like me to say it more.

[00:14:55] Sezin Cagil: I was about to say that.

[00:14:57] James Gurd: Many developers wouldn’t like me saying it, but sometimes needs must. So Carol, let’s come to you on this. What do you advise about, what does good due diligence look like for a merchant looking to select tech?

[00:15:07] Carole Breetzke: Yeah, I think it’s very important to select vendors that are very open on the ecosystem because now you no longer have your vendors that just integrate to your ecommerce platform. Like a tool like Klevu integrates to the CMS, the marketing tool. It needs to be able to ingest different type of data and not just your product. It could be like various different types because commerce are becoming more complex expanding. They’ve got omni channel use cases. I’ve got a different type of use cases.

So it’s very important to have a vendor that is very much open, has different ways of integrating, whether it is building apps, SDKs, libraries. And I would also say having a partner that also embark on technologies that are widely spread so that if you do change storefront or technology, you can also take what you build and just use it with another system. So that’s quite key for me.

[00:15:57] James Gurd: Excellent. So let’s get on to governance piece. Now, this to me is where the industry needs to evolve to closer governance around AI selection and the role of AI in the business, because it’s not just about one team using it. It’s about the broader strategy of the business and what AI role is and how you select tech. And we know the media is obsessed with it. It is a bit bandwagon, but there are lots of practical use cases. But it brings into play compliance, data, privacy stuff. Sezin, let’s come back to you. What challenges do you face in a merchant environment aligning AI driven initiatives with broader product roadmap and strategy?

[00:16:31] Sezin Cagil: It is really the lack of understanding. I think it is very difficult within today’s world to have in-house people who have been in business for many years to really understand what you can really deliver using AI. So we need to find ways of innovating internally and educating our workforce and aligning the mindset that is happening in the tech world in most of the enterprises, you won’t be able to use AI you have, they won’t allow you to use AI. So the sooner they allow you to use AI, the quicker you will be educated. There are so many versions, tools around the later you get into the game. the longer it will take for you to learn. The learning curve is exponential. So I think that is the main thing. Get your workforce, touch it, play with it. Who said it in the earlier panel that they talk to the ChatGPT, every day? That’s a brilliant way of starting. Just see how you feel with the tool. I think also there was a point about the data earlier. Again, data is the both the problem and the solution. You need to have the right data, the right amount of data to be able to play and bring the right solutions for your business that will be sustainable.

[00:17:42] James Gurd: And Andy, let’s go to you to build this. What pitfalls do you see that they impede like ecommerce businesses and not just ecommerce businesses, but the bussinesses getting the most of AI investment?

[00:17:52] Andy Eva-Dale: Yeah, sure. I think whenever you’re considering AI products or AI product builds, there’s probably four key pitfalls for me. One is value. The second one will be around cost of running in production or influence, brands protection and then lastly, performance. When we say performance in AI, we don’t mean how quickly it responds, it’s how much it understands the question. So if you look, first of all, at value, lots of brands are coming up saying, I don’t know what the problem is, but AI is the answer. And that’s not how you should be approaching it. You need to make sure that you have solid business cases around implementing it. That you see either reduction in costs or increased profits. They’re the core drivers behind implementing it. And like by taking the POC approach and something we do as well, you can test and iterate this quickly without having huge investment. Trust partners, get excitement in your business, get stakeholders on board. And that’s what a key part of starting an AI project.

From a brand perspective, did anyone see the DPD case study? DPD released a chatbot on their website and you could ask it anything. You could ask it to write an embarrassing poem about DPD and it’ll be like, DPD is awful at delivery, blah, blah, blah. I’m not a poet, you get the point. The problem is lots of brands are like trying to get AI out quickly without thinking about the proper due diligence and protecting their brands. And jailbreaking of AI is a really big concern for your brand. So you need to make sure that you have safeguards in place.The proper guardrails are in place around the implementation and you don’t just try and get something out because that could be disastrous for your brand.

From a performance perspective, as well you need to make sure that you’re not just implementing AI quickly and you might have heard of a technique called RAG. It’s where you get a foundational model in the cloud and you can inject your content on top, it’s very quick, and it suddenly understands your domain. The problem with that is it does understand your domain to a point that the understanding isn’t amazing out of the box. You need to fine tune it. You need to prompt engineer it. There’s these development techniques you need to do to make sure you get the most out of the products. And lastly is cost. AI costs a fortune to run in production. Be warned, when you start running in production and putting volumes of people through it, it’s expensive. It’s really expensive and there’s a big carbon impact around that back to the previous point. So you need to start thinking about how you optimize the systems you’re building.

You don’t need to use ChatGPT 3.5 or 4.0 for absolutely every use case. You should start to look at the same as composable systems design, bringing them down to small individual chunks or agents that do one job and start to think about using small language models. They don’t need to run necessarily on GPUs. You can get really good performance out of CPUs, meaning your cost in production is much less, but you don’t need to use a sledgehammer to crack every nut is the point.

[00:20:44] James Gurd: Yeah. I think that guardrail point is really worth emphasizing because when I look at the examples where chatbots specifically have gone rogue, it’s because the focus was on training them around content. For outcomes, but not on making sure that they don’t then drop swear in everywhere and abuse people. So the guardrails weren’t in place. And if you don’t have governance, you end up with compromised outcomes. So I think that’s such an important point to take away.

[00:21:06] Rachel Tonner: We’ve actually just launched a new product at Klevu called Asklo, which is very much around like what we’re talking about here. It’s a very specific use case. It’s just on the PDP. It’s a chatbot that answers questions just specifically about that product. The ring fencing to open AI is very small and specific about like the uses of that product and and it works really well. We haven’t launched it wider than that because right now it works really well just for that one product. Just, in a very small use case. But retailers are starting to use it and trust it because it actually does work. So if you’re bringing that into your business, I would also recommend, yeah, just find like a small problem and solve it.

[00:21:46] James Gurd: Yeah, I think that’s a great example. I had a client of mine who put in an AI chat. It was an open source AI. It wasn’t one of the big third parties. But to solve a very specific problem, which is they got, they’re a discount food brand. So they buy like end of stock stock was going out of date and they sell it really cheap. So they get constant inquiries about stock availability, about deliveries, et cetera. And about 80 percent of their customer service tickets are all on that alone. So to exactly what you said, the use case, the chatbot was, we’re not opening it up to AI conversation. It’s purely AI, there’s prescriptive flows for delivery information, stock information returns. Anything else is pushed into do you want to speak to somebody that then submits a ticket for a CS person to follow up?

So very narrow I’m sure there’s probably some someone who will find a way to abuse it at some point, but narrows that window of opportunity. Let’s maybe then we go to our last question, which is more of a great discussion. Let’s make it nice and social, shall we? Which is sharing an example of successfully integrating AI into ecom business to enhance performance. I know we’ve had a few, but let’s try and find some extras. I think this is always helpful. The practical takeaways, theory, concepts, frameworks are great, but like what, what works basically. So let’s start with you. Give me that nice sense. But like you work a lot with merchants, so what’s a good example?

[00:23:03] Rachel Tonner: Absolutely. We see a lot of our clients come in, like maybe they’re using a non AI search and merchandising tool. Maybe they’re using a native search from their platform, and it’s really easy and almost immediate to see the uplift from then putting in an AI search and merchant personalization tool like Klevu. Astrid & Miyu saw a 22 percent decrease in bounce rates. Paul Smith saw 73 percent more search led revenue. Cambridge Satchel saw, I think, 64 percent more conversion from their product recommendations. So these are real. This is real data, and it’s really proven. And yeah, I would just say that even if like other technologies might be AI washing, like I said before, really check in to see how mature that algorithm actually is so that you can invest in a wiser, responsible way.

[00:23:49] James Gurd: Yeah, nice. Let’s and this is work away down. That’s much easier.

[00:23:54] Sezin Cagil: Okay. So we tried lots of different, AI tools: attribution, imagery, descriptions. Somebody was talking about translation that found really badly. But the one solution that worked really was actually a chatbot. This was quite a few years ago now, but that was integrated with the backend system. So you can self serve. We didn’t have the power in the customer services team to manage all sorts of international markets. So through this, chatbot that was translated, we could easily integrate multiple markets and they could self serve and create a ticket when needed. The translation at that context worked, but we run a POC.

[00:24:29] Andy Eva-Dale: We are currently doing a large transformational project for a professional services company, a well known one in the UK, I can’t say who unfortunately, but we are implementing conversational interfaces and remote process design for them. The business case is absolutely solid. We are significantly driving down large cost centers for their business and we’re providing a new source of revenue, a new revenue stream for it. So both decreasing costs and increasing profit for them. And it’s a good use back to the point of having a value out of using AI. It’s a solid example of that.

[00:25:00] Carole Breetzke: The most successful use case for us is using AI for like generative AI for SEO optimized content. So we do have a bot like trained on SEO contents. So just optimizing some keywords and also on the tone of voice of the brand just to generate a lot of pages. And that’s the best case we’ve got because you can really achieve like a lot of volume that wouldn’t be achievable manually.

[00:25:23] James Gurd: One example that I really like is so one of my clients has got a lot of products that are really crap for data.You may want to experience this when you work with multiple suppliers and the quality of data varies massively. But you don’t have the resource to optimize every single one human written. You just can’t. So you focus on the bestsellers, the category kittens, which means you build up content debt over time. So actually, we just put it through AI content generation stubble off while back with copy then moved on to more modern models to gradually backfill those, but at this time to market of backfilling and reduce removing some of that crap content was so much quicker, needed less time. And then they’ve now got better product copy across the board.

So there are some real tangible quick wins in that. I think that one caveat is always having that human expert oversight so that you send check and don’t get those random things that creep in. That are completely irrelevant. I had it once where basically they’d swap the name. I was trying to get a copy for my own website, and they’d put another consultant’s name in. And I thought, that’s interesting, they all promote themselves. So yeah, I think there’s lots of smart examples out there. Yeah, lots of people really put them together. So that probably gives us a couple of questions from the audience.

[00:26:11] Audience: So you talked about the benchmarking of the carbon footprint or whatever. So how would you actually go about that? Particularly as in a more composable world where you’ve got edge delivery. So you’ve got so many services, all the which will have servers and stuff kick somewhere else. How do you actually go about, say, coming up with a number for a home page, for example?

[00:26:47] Andy Eva-Dale: Absolutely. You need to think about it all the way down the stack. Starting off with the browser itself the Green Web Foundation has released a free open source JavaScript product called co2. js, which will calculate the usage on the browser. You need to think about that because the brightness of the screen, the size of the images, the number of fonts, all this sort of stuff burns energy for the end user. Then if you’re using cloud and you’re using GCP, AWS or Azure as part of your billing data in there, you have carbon metrics as well, and you can aggregate these into single dashboards. If you look if you’re on Azure, for example. There’s a Power BI, free Power BI plug in, although Power BI isn’t free, it’s that allows you to see like all your impacts or mission.

It’s a mission dashboard. Let’s just see what your missions over time. There’s also a free dashboard from the website Carbon Foundation. I think it is which you plugged in all into one point. So that kind of covers the browser. It covers your usage in the cloud. Then the third point you mentioned was your supply chain. Now, that’s trickier. We’re working with a lot of partners now to make sure they’re transparent about their usage, but as part of the EU greenhouse gas protocols that are being implemented, there’s something called Scope 3. Scope 3 means, as part of the supply chain, and in tangent, we’re part of that supply chain, we will be responsible for, in 2025, reporting on our carbon emissions up, up, up the chain. So Scope 1 and 2 are your direct and indirect emissions as an organisation, and Scope 3 is your supply chain. So you’ll find that lots of organizations will start reporting on that sort of, integrations part. Dead easy, yeah. It will do, yeah, it has to as part of EU law, so yes.

[00:28:29] Audience: Nowadays, things are getting lightning fast, usually more GPUs, CPUs all the time. How do bring the sustainability in your coding practice? In fact, you have coding standards and you don’t have sustainable coding standards.

[00:28:41] Andy Eva-Dale: Oh, we do. Yeah, we’ve actually the W3C, I don’t know if you’re aware of them, they do like general web standards, they’ve actually just released sustainable standards out of beta. I’m part of the Beam sustainability community. We publish standards for the marketing the digital agency world, and we’re also on lots of partner advisory boards, making sure we have a sustainable standards within each organization. And this will be things like front end standards around static site generation, optimization of images. There’s even Microsoft themselves have published this sustainable design practices. So there’s things like the circuit breaker pattern. If you’re a backend engineer, you can use which is a sustainable design practice. And from an infrastructure perspective, you look at IaaS to PaaS to cloud native, and all of this migration loops into sustainable systems design. So I think on an AI perspective as well, to the point I made earlier on, By not using large language models for everything and optimizing that and thinking about GPU usage thinking about performance and trying to use small language models in an agentic architecture. All of these things are sustainable systems design. And there’s lots of practices out there, lots of people publishing about it. But you should look at the W3C ones. They’re pretty interesting and new.

[00:29:53] James Gurd: Thank you very much everybody for coming on and for listening.