How to make sustainable AI a reality

Is “sustainable AI” a contradiction in terms? This was the question at the heart of our first “AI in Motion” event, a series tackling the big, human questions around AI. The session started with a debate chaired by TLT’s Tom Sharpe, with expert contributions from Natasha McCarthy (Royal Academy of Engineering) and Robert Keus (GreenPT), followed by a panel of expert speakers in Emily Farrimond (Baringa Partners), William Quan (Fleete Group Limited) and Andrew Burgess (Greenhouse Intelligence), chaired by TLT’s Emma Erskine-Fox.  

The consensus: sustainable AI is not only possible, but essential. Here’s how organisations can bridge the gap between ambitious AI goals and net-zero commitments.

The scale of the challenge: Be realistic

AI’s rapid proliferation brings undeniable challenges for decarbonisation. The energy and water demands of AI models are significant, and reliable sustainability metrics are still hard to come by. As Andrew Burgess noted, “One of the biggest challenges is the lack of information and measurement. We talked about some of the metrics that are available to people, but does anybody really believe those metrics? The people that are generating them are the people that are selling the models as well. So it’s a tricky one to balance until there’s much more transparency over the data that’s available on the sustainability.” 

Without trustworthy data, meaningful comparisons and informed decisions are difficult.

Key takeaway:

Acknowledge the scale and complexity of aligning AI adoption with net-zero ambitions. Honest conversations about the challenges are the first step towards solutions.

Creative solutions: Be creative

Sustainability in AI isn’t just about technology - it’s about mindset. The panel emphasised the need for creativity in how we develop, procure, and embed AI. This includes exploring smaller, more efficient models, asking tough questions of vendors, and building sustainability into policy, governance, and training. As Emma Erskine-Fox put it, “There are lots of ways that organisations can be looking to reduce the environmental impact of their AI usage through creativity in the way that they develop AI solutions, the way that they procure them, through embedding sustainability in the training, policy, guidance and governance frameworks around the use of AI internally.”

Key takeaway:

Think outside the box - innovative approaches to AI development and deployment can significantly reduce environmental impact.

Transparency and responsibility: Be accountable

Transparency is a recurring theme. Developers must be honest about the resource and energy use of AI systems. Sustainability considerations span the entire AI value chain, from hardware and infrastructure to data centres and use cases. Responsibility is shared: developers, organisations, and end users all have a role to play. As Robert Keus said, “As human beings, as the single users of AI, we have our own responsibility of how we should use it.” And Natasha McCarthy added, “It just shows just how distributed the kind of responsibilities and decisions are around making AI and its use more sustainable.”

Key takeaway:

Sustainable AI requires end-to-end responsibility and greater transparency at every stage, from development to individual use. 

Empowering users: Be informed

Accelerating responsible AI adoption means empowering users. Training is essential, not just in how to prompt AI, but in understanding its true costs and capabilities. This builds confidence and ensures AI is used where it adds real value. William Quan highlighted, “How do you accelerate adoption in the right way, but also how do you train the end users to be able to prompt, but also understand the technology in a way that maximises efficiency and productivity?”

Key takeaway:

Invest in user education to maximise AI’s benefits and minimise unnecessary energy consumption.

Hope for the future: Be hopeful

Ongoing dialogue, improving data quality, and progress towards common standards for emissions reporting all offer genuine grounds for optimism. As Tom Sharpe noted, these developments will enable "adoption of the right tech in the right way."

Despite the challenges, there are real reasons for hope. As Emma Erskine-Fox aptly summarised: "Be realistic, be creative, be hopeful."

Key takeaway:

Progress is possible - by working together, we can harness AI’s potential for good while safeguarding our planet.

Final thoughts

Sustainable AI isn’t a contradiction. It’s a collective challenge that demands realism, creativity, transparency, and hope. By embracing these principles, organisations can ensure their AI ambitions support, rather than undermine, their sustainability goals.

Key takeaways:

• Be realistic about the scale of the challenge.
• Be creative in finding solutions.
• Be transparent and accountable at every stage.
• Empower users through education.
• Stay hopeful - progress is within reach. 

This publication is intended for general guidance and represents our understanding of the relevant law and practice as at October 2025. Specific advice should be sought for specific cases. For more information see our terms & conditions.