Artificial intelligence has rapidly reshaped how organisations operate, analyse information and make decisions. Its ability to process vast amounts of data and generate insights offers significant advantages in efficiency, innovation and strategic clarity. However, this progress comes with increasing complexity and cost. As models grow in scale and capability, so too do their energy demands, environmental impact and ethical risks.
Sustainable AI emerges as a response to this tension. It recognises that AI is not only a tool for solving problems but also a system that must be designed, deployed and governed responsibly. This includes understanding its environmental footprint, ensuring transparency in decision-making and aligning outputs with human values and societal expectations.
At its core, sustainable AI is about balance. It requires organisations to move beyond performance alone and consider efficiency, accountability and long-term impact. When applied effectively, it enables better decisions, stronger trust and more resilient outcomes in an increasingly data-driven world.
Artificial intelligence offers significant opportunities to improve decision-making, but it also introduces significant sustainability challenges. The rapid growth of AI systems has led to increasing computational demands, with training models requiring “vast amounts of computational resources, energy, and water” (Bolón‐Canedo et al., 2024). This has contributed to an environmental impact that is “growing almost exponentially,” raising concerns about long-term sustainability.
This impact is not only conceptual but increasingly measurable in practice. Research shows that training large deep learning models can result in substantial carbon emissions, with some estimates suggesting that the footprint of training a single model can be comparable to the lifetime emissions of multiple vehicles (Strubell et al., 2019). This highlights the environmental cost associated with model development and repeated experimentation. In addition, the infrastructure supporting AI systems, particularly data centres, requires significant cooling, contributing to increased water usage and environmental strain. One commonly used measure of efficiency in this context is Power Usage Effectiveness (PUE), which evaluates how efficiently a data centre uses energy to support computing processes. Together, these factors highlight the growing need to balance performance with efficiency, reinforcing the importance of more sustainable approaches to AI development and use.
At the same time, AI is not inherently negative. It is described as a “double-edged sword” (Zhao & Fariñas, 2022), capable of both advancing sustainability and creating new risks. On one hand, AI can enhance efficiency, improve decision-making, and support complex problem-solving. On the other, it can introduce issues such as bias, lack of transparency, and increased energy consumption.
Sustainable AI addresses this tension by focusing on how systems are designed, implemented, and used. It is defined as incorporating “sustainable practices and techniques… to reduce the associated environmental cost and carbon footprint” (Bolón‐Canedo et al., 2024). This includes optimising algorithms, improving hardware efficiency, and adopting better data management practices.
A critical shift within this approach is moving from performance-only thinking to efficiency-aware design. Research shows that many AI developments have prioritised accuracy over efficiency, with “the vast majority of papers… prioritizing accuracy over efficiency” (Bolón‐Canedo et al., 2024). Sustainable AI challenges this by encouraging solutions that deliver value without excessive resource consumption.
Data plays a central role in this process. AI systems rely on data to function, and “data is the fuel of AI systems and determines its outcomes” (Zhao & Fariñas, 2022). However, poor data quality can undermine both performance and ethical outcomes, as it is considered “enemy number one” in machine learning. High-quality, reliable data is therefore essential for both effective and responsible AI.
Beyond technical design, transparency is a key requirement for sustainable AI. Decision-making processes must be understandable, as organisations need “the ability to know how and why a model performed the way it did” (Zhao & Fariñas, 2022). Without this clarity, trust is reduced and the ability to challenge or validate decisions is limited.
This is where simplification and visualisation become critical. AI outputs are often complex, but their value lies in how effectively they are communicated. The use of “synthetic data visualisations” can help translate complex analytics into accessible insights, supporting better understanding and decision-making (Zhao & Fariñas, 2022). By simplifying outputs and providing context, organisations can turn technical analysis into actionable recommendations.
Context is equally important. AI does not operate in isolation; its outputs must be interpreted within organisational priorities, stakeholder expectations and broader societal considerations. When information is transformed into meaningful knowledge, it enables individuals “to make the correct movement at the right time” (Zhao & Fariñas, 2022). This transformation from data to insight to action is where sustainable AI delivers its greatest value.
Regulation and governance also play a crucial role. Sustainable AI cannot rely solely on voluntary action. It requires structured oversight, as “accountable and sustainable AI can be achieved through a proactive regulatory framework” (Zhao & Fariñas, 2022). Without this, unregulated AI “would be a threat” due to the inability to monitor its impact effectively (Zhao & Fariñas, 2022).
Ultimately, sustainable AI is not about limiting innovation but guiding it. It ensures that systems are efficient, transparent, and aligned with human values. By combining responsible design, high-quality data, clear communication, and effective governance, organisations can harness AI in a way that supports both performance and sustainability.
Review one current use of data or AI within your organisation. Identify where efficiency, transparency, or clarity could be improved. Focus on one practical change, such as simplifying outputs, improving data quality, or reducing unnecessary complexity. Define how this change would improve both decision-making and sustainability and commit to testing it within a real scenario.
Sustainable AI: Know It, Check It, Use It
Sustainable AI is not just about technology. It is about how decisions are shaped, communicated, and acted upon. This checklist helps assess whether AI or data-driven insights are being used responsibly, efficiently and effectively. It focuses on clarity, impact, and alignment with organisational values.
Select one recent decision supported by data or AI. Use the checklist to evaluate how effectively the insight was generated, communicated, and applied. Identify one gap and take action to improve how future insights are translated into clear, sustainable recommendations.
|
Know It |
Check It |
Use It |
|
AI has environmental impact |
Are resource demands justified? |
Reduce unnecessary complexity |
|
Data quality drives outcomes |
Is the data reliable and relevant? |
Improve data inputs |
|
Transparency builds trust |
Can decisions be explained clearly? |
Simplify explanations |
|
Visuals aid understanding |
Are insights easy to interpret? |
Use clear visual summaries |
|
Context shapes meaning |
Is the insight linked to strategy? |
Align with priorities |
|
Governance ensures responsibility |
Are risks and ethics considered? |
Apply structured oversight |