The benefits of a sustainability-first approach to AI adoption

Hanna Barakat & Cambridge Diversity Fund / Better Images of AI / Analog Lecture on Computing / CC-BY 4.0

By Eliot Gillings, Policy Adviser, Royal Academy of Engineering

As AI adoption is fast-tracked, SMEs that prioritise sustainability in their approaches to procurement and use may avail benefits beyond a smaller environmental footprint.

Eliot Gillings (Policy Adviser, Royal Academy of Engineering) leads the AI sustainability strand of the National Engineering Policy Centre (NEPC) Engineering Responsible AI project. The NEPC is a partnership of 42 professional engineering institutions, led by the Royal Academy of Engineering, which provides insights, advice, and practical policy recommendations on complex national and global challenges.

In January of this year, the UK Government gave its endorsement to Matt Clifford’s AI Opportunities Action Plan. Written with the intention of laying “the foundations for AI to flourish in the UK,” both the original report and the Government’s response placed a significant emphasis on “boosting adoption across the public and private sector.” Affirming this commitment, the Prime Minister has since written to his Cabinet and asked them to drive “AI adoption and growth in their sectors” – and has assigned DSIT to the task of supporting devolved and local government to identify AI adoption opportunities.

For many SMEs, promises to fast-track AI adoption will be welcome news. As a recent survey from the Business Growth Fund (BGF) showed, while a minority of SMEs in the UK and Ireland are currently using AI (1 in 4), the vast majority of senior leaders within those organisations (89%) believe AI “presents an opportunity for their business overall.”

For AI adoption to be trustworthy, ethical and inclusive, AI adopters need to have confidence that systems will function as intended, that biases can be addressed and that there are ways to monitor and manage environmental impacts.

At the NEPC, we have been looking especially closely at the monitoring and management of the environmental impacts associated with AI. Our recently published report proposes five foundational steps that can be taken by governments to begin progress towards environmentally sustainable AI. These steps are:

1.       Expanding environmental reporting mandates

2.       Addressing information asymmetries across the value chain

3.       Setting environmental sustainability requirements for data centres

4.       Reconsidering data collection, transmission, storage, and management practices

5.       Leading the way with government investment

Within the report, we supplement these foundational steps with key actions that governments can take to minimise risks to the environment, people, and the economy. While the foundational steps (and supplementary actions) are framed for a policymaking audience – the report outlines a number of potential co-benefits that could be availed by SMEs (and businesses more broadly) who prioritise environmental sustainability in their adoption strategies, such as lower taxes and greater investment, easier to maintain systems, and reduced overhead costs.

Benefit 1: Lower taxes and greater investment

In recent years, the development and deployment of AI systems has steadily become increasingly environmentally impactful – and there are no strong indicators that this will change in the near future. It is therefore not unlikely that the companies who prioritise scale or capability over environmental sustainability in their procurement and use strategies could fail to meet their sustainability targets. This is already happening for some AI adopters. The Capgemini Research Institute, for instance, recently found that almost half (48%) of executives believe their use of generative AI has driven a rise in emissions, with 42% now needing to reassess climate goals.

A failure to reach sustainability targets can have negative financial consequences for a business. Financial institutions and regulators have increasingly embedded ESG considerations into their operations, largely to reduce volatility and risk. This means that businesses who allow their use of AI to undermine their ESG credentials may find themselves decreasing their attractiveness to investors.

2025 is expected to be the year the UK formally adopts Sustainability Reporting Standards, based on the International Sustainability Standards Board’s IFRS S1 and S2 standards. These standards, once adopted, are expected to improve investor understandings of business’ ESG-related risks and opportunities – and align UK sustainability reporting with international standards. While the effect of this move has yet to be determined, it is unlikely to reduce investor interest in businesses with strong ESG credentials.

Additionally, the UK continues to offer numerous taxes and reliefs that reward businesses for environmentally sustainable practices (particularly reducing energy demand).

Benefit 2: Easier to maintain systems

Generally speaking, the level of investment required to effectively use an AI system scales with its size and complexity. As such, businesses who adopt smaller and more environmentally sustainable systems may see benefits in the form of fewer maintenance and oversight requirements, as well as a smaller environmental footprint. 

Large AI systems generally require data to be more diverse, plentiful and of higher quality than leaner systems – meaning higher levels of investment are required by adopters to meet the system’s data quality and cybersecurity requirements.

Businesses seeking to reduce the cost of adopting AI may, therefore, come to view frugality as a helpful guiding principle. This is especially true for smaller businesses looking to improve cybersecurity and create efficiencies, as these are tasks that typically do not require complex data analysis for smaller enterprises, and for which small models are often suitable.

Benefit 3: Reduced overhead costs

In recent years, costs for businesses using AI have steadily risen. These costs are predicted to continue rising – a 2024 report from IBM predicts that 2025 will see “the average cost of computing” climb to 89% above 2023 levels.

This trend generally disadvantages smaller firms – who may start looking towards smaller and alternative AI models to meet their needs. Small and alternative models, such as task-specific models, generally use smaller datasets and less compute than larger multi-purpose models – often without a significant drop-off in performance. Smaller datasets mean less requirements to invest in data collection, storage, and preparation (all of which have environmental impacts) – and less compute means fewer cloud costs. In fact, some models can be deployed on edge or end-user devices, significantly reducing cloud, internet, and security costs.

In conclusion

For many SMEs, AI offers an exciting opportunity to improve cybersecurity, customer support, content creation and even their approaches to managing environmental impacts. Accessing those benefits without creating undue exposure to risk, however, requires businesses to carefully consider the costs and benefits associated with the procurement and use of AI systems and services. As I’ve argued over the course of this blog, businesses that prioritise environmental sustainability can avail financial benefits – but there is a strong argument that any effective adoption strategy must emphasise environmental sustainability.

As we continue the Engineering Responsible AI project, we welcome opportunities to collaborate with likeminded organisations who are passionate about engineering responsible AI, minimising the environmental risks, and collectively driving change. If you’re interested in connecting – please reach out to: infrastructurespolicy@raeng.org.uk