This Earth Day 2025, our Head of Marketing, Catherine King, sat down with Dr. Karine Serfaty to explore how AI and data are shaping the future of our planet.
Karine Serfaty is a seasoned strategy and data leader with expertise in AI transformation, data strategy, and sustainable innovation. She has held senior roles at The New York Times, ITV, and The Economist Group, where she drove major shifts in digital maturity, data strategy, marketing performance, and business model innovation.
Today, she partners with organisations to future-proof their business models by integrating AI, data, and sustainability into strategy and execution. She teaches media economics at the École Normale Supérieure (ENS) and serves on the board of its executive institute. Karine holds a PhD in Economics from Harvard and certifications in Generative AI and Sustainability.
How do you see AI enhancing sustainability efforts across industries today?
There are several ways AI is already contributing to sustainability. One of the most immediate is workflow augmentation, for example, automating emissions estimates to help companies build a carbon baseline and reduction roadmap. That’s one of the more straightforward GenAI use cases, where it can extract and organise relevant data from invoices, utility bills, or contracts. Gen AI can also help draft reports or client sustainability questionnaires.
But AI is also pushing sustainability beyond the compliance checklist, toward actual action. It supports everything from real-time route optimisation to selecting crops better suited to changing climates, or modelling climate risks and potential business responses. So it plays a role in both mitigation (reducing emissions) and adaptation, since climate change is already a lived reality.
What are the most promising use cases where AI is actually supporting climate action or environmental monitoring?
Beyond carbon accounting and process optimisation, I find three categories particularly promising, and they have something in common: they’re deeply tied to the physical world, where change actually happens.
First, AI can act as a bridge between the physical world and digital capabilities. Think of how image or signal recognition helps detect emissions at the source (like Climate Trace), map energy leaks from homes to accelerate retrofitting, track deforestation parcel by parcel, or construct biodiversity indicators. These uses create new structured datasets that can be exploited for many applications. They lay the groundwork for measurable, targeted action.
Second, AI enables materials and infrastructure innovation. In materials science, for instance, it’s being used to design substances with specific energy properties like better battery storage, more efficient solar panels, or even plant-based proteins. BASF and DeepMind are among the players here.
Third, AI is helping develop new business models based on service, not volume. Think of building management systems, predictive rail maintenance, or “tyres as a service.” The less input you use, the more valuable the system. Companies as different as Siemens, Schneider Electric, Michelin, and Rolls-Royce are all players in this shift. These aren’t moonshots. They’re live and scaling.
What role does a data strategy play in shaping a company’s sustainability roadmap?
A central one. Sustainability reporting is, at its core, a data challenge. But more than that, if you want to move from reporting to action, you need the right data infrastructure to support it, whether you’re optimising processes, piloting emissions reductions, or building more sustainable products.
As with any data strategy, the key is to identify the highest-value use cases and build from there. And with sustainability, there’s often a hidden multiplier: for instance, the supply chain data you collect for emissions might also help you streamline operations or improve customer experience. That can help make the business case stronger internally.
Should companies be measuring the carbon footprint of their data infrastructure?
Yes, and many already do, especially those running large-scale AI or data platforms. But measurement is just the start. What also matters is how and when your infrastructure runs, and where it’s located. I’ve worked with clients who time their compute workloads to align with renewable energy availability.
Importantly, you have to distinguish between cloud operators and cloud users. The latter don’t control the data centres themselves, but they can monitor their impact and make informed choices about providers and efficiency.
The good news is, carbon efficiency often goes hand in hand with cost savings. So it’s both a climate and a business imperative.
How concerned should we be about the carbon footprint of generative AI?
It’s a valid concern. We’re seeing a sharp rise in electricity demand, and the AI boom clearly is one driver. And in some cases, it’s leading to short-term decisions that favour fossil fuel power over renewables.
That said, we should stay pragmatic. Some of the worst-case projections assume no improvement in efficiency, which is unlikely. New models like DeepSeek are already delivering better performance with fewer resources.
Back in the early 2000s, there were predictions that the digital sector would soon account for 50% of global energy use. We’re still in the 2–4% range, depending on the methodology. So let’s keep perspective.
Personally, I think awareness and sobriety have their place, i.e. not using AI for everything, and making smart trade-offs. But at the company level, the real focus should be governance: Where and how is AI being deployed? Are we prioritising net-positive use cases? Are we aligning AI with broader decarbonization goals? That’s where the real leverage is.
How do you evaluate the trade-offs between AI’s potential for sustainability and its resource consumption?
This is the core question: Can a given application be – or become – net beneficial? That includes emissions and compute cost, but also the downstream value, e.g. reducing waste, avoiding travel, improving logistics, or unlocking entirely new ways to allocate resources.
We need more sophisticated frameworks for evaluating these trade-offs. And let’s be honest, AI projects are moving forward either way. Our job, as sustainability and data leaders, is to make sure the right ones rise to the top of the list.
