AI is one of the most transformative technologies of our generation. It is already changing how organisations work, how decisions are made, and how scientific discoveries happen. For example, here at the Met Office, we’re exploring how AI can help advance weather and climate science, improve services and support our people.
Yet that simple comment captured something important. Public discussions now often focus on what AI takes, whether that’s resources or human expertise or a myriad of other issues, but less attention is being given to what transformational improvements could be achieved with precise targeted and responsible implementation.
The reality is that AI is neither inherently good nor bad. It is simply a tool. Like any powerful tool, its value depends on how, why, and where it is used. I often describe AI as a magnifying glass or an accelerant. When applied to positive purposes, it can amplify and accelerate beneficial outcomes. Sadly, the converse is also true.
The challenge facing organisations today is not whether to adopt AI. The challenge is navigating the trade-offs that come with it and using it responsibly.
AI offers extraordinary opportunities
The excitement surrounding AI is understandable and tangible.
Artificial Intelligence (AI) is moving from the margins to the mainstream of public service delivery, from healthcare[i] and education[ii] to transport[iii], emergency response[iv] and weather forecasting[v]. Right across society AI is already helping organisations analyse vast quantities of information, automate repetitive tasks and generate new insights. In scientific fields, it is helping researchers work at unprecedented speed and scale.
The opportunities are vast: AI has the potential to help organisations do things faster, cheaper and, in some cases, better than ever before. In our field, the opportunities for weather and climate science are particularly significant.
Recent advances in machine learning have demonstrated that AI models can generate global weather forecasts in seconds, something that would previously have required substantial supercomputing resources[vi],[vii]. Researchers are also using AI to identify patterns in large environmental datasets, improve climate modelling techniques and enhance our understanding of weather-related risks[viii],[ix],[x],[xi],[xii].
The potential extends beyond forecasting – AI could help push forward the frontiers of weather science, accelerating scientific discovery, improving decision-making during severe weather events and supporting more efficient delivery of public services.
These possibilities are exciting, but raise important questions.
The environmental dilemma
One of the most immediate tensions concerns sustainability[xiii].
AI is sometimes presented as a technology that can help tackle climate change[xiv]. It can help optimise renewable energy systems, improve transport efficiency and strengthen resilience to extreme weather. But tensioned against this we should acknowledge that, like most digital technologies, AI itself has an environmental footprint.
Training and operating advanced AI models requires significant computing power. That means energy consumption, water use and infrastructure requirements. For organisations whose mission includes environmental stewardship, this creates a genuine dilemma.
The Met Office’s use of AI in weather forecasting provides a useful example. The physics-based forecasting models we use today rely on powerful supercomputers. Emerging AI-based forecasting approaches, or more precisely machine learning (ML)-based approaches, may reduce computational requirements during forecast generation (known as inference). However, this efficiency is achieved only after substantial investment in developing and training the models. As a result, the environmental balance sheet is far from straightforward.
Conversations about sustainable AI need nuance. The question is not whether AI is “green” or “not green”. Rather, we need to understand where AI creates environmental benefit, where it introduces environmental cost and how those trade-offs should be managed.
AI may be part of the solution to climate change. But it can also contribute to the problem if deployed irresponsibly.
Productivity matters. People matter more
When a new technology emerges, concerns about jobs often follow, and history suggests that we should be neither blindly optimistic nor overly pessimistic.
Technological progress changes the nature of work. It has automated certain tasks while creating entirely new opportunities. At its best, advances in technology remove repetitive tasks while creating entirely new professions. The invention of the printing press in the 15th century is an example. Yes, it changed people’s jobs who were reliant on the production of the written word, but it also triggered a global leap in human progress and knowledge sharing. There are parallels with the AI Revolution that we are now experiencing.
AI can remove repetitive administrative tasks and allow people to focus on higher-value activities requiring creativity, expertise, judgement and empathy. That is an opportunity we should embrace.
Public services depend on highly skilled professionals. In our world, meteorologists, climate scientists, technologists and operational experts bring decades of accumulated experience to their work. Their expertise extends far beyond producing a forecast. It includes interpreting uncertainty, understanding risk, communicating effectively and supporting critical decisions that save lives and livelihoods. It also builds on the knowledge and expertise of those who came before them (as scientist we truly ‘stand on the shoulders of giants’). This accumulated, collective knowledge and expertise helps the Met Office to deliver its mission to help people make better decisions to stay safe and thrive[xv].
This is one of the many reasons we say the future of weather forecasting is unlikely to be a choice between physics and AI, or between humans and machines. Instead, the future will be a blend.
The most successful organisations will be those that find ways to combine human expertise with AI capability to amplify the strengths of each.
Preserving scientific understanding
Another important trade-off often receives far less attention.
What happens if we become so reliant on AI that we stop developing the expertise needed to understand, question and improve it? For the individual, recent studies have shown a negative correlation between frequent AI tool usage and critical thinking abilities, mediated by increased cognitive offloading and surrender[xvi].
This risk does not arise because experts and expertise disappear overnight. Instead, it may emerge gradually at both individual and societal levels through lack of regular use, and a lack of education or training opportunities.
As history attests, science advances through curiosity. It advances because people ask why and how something happened, challenge assumptions and test ideas. AI may support this process, but it is unlikely to replace it.
This is particularly important in weather and climate science where climate change means that our future climate will not always look like the past. Understanding the underlying physics of the atmosphere remains an essential part of our ability to interpret forecasts, identify limitations and push scientific knowledge forward.
If organisations become dependent on systems they no longer fully understand, they risk losing the very expertise that allows them to innovate and adapt. For example, recent research has shown that embedding AI agents may reduce managerial oversight in AI-mediated work[xvii]. Loss of the curiosity that fuels scientific discovery, and the specialist knowledge that results, would be a heavy price to pay for the efficiency gains when embedding AI.
Trust is built through governance
As AI systems become more powerful, effective governance becomes ever more important, yet governance is sometimes portrayed as a barrier to innovation. But the truth is that good governance should enable innovation.
Clear accountability, robust assurance processes and proportionate oversight help organisations identify risks early and deploy technology with confidence. They provide a foundation for trust.
For public services, trust is fundamental. For example, people rely on weather forecasts to make decisions that affect their safety, businesses and communities. Our most recent research suggests 85% of people in the UK trust the Met Office. That trust has been earned over decades and will not be taken for granted.
The same principle applies to AI. The goal is not to move fast at any cost. Rather to govern well, innovate responsibly, and ensure that benefits are realised while risks are understood and managed.
Choosing the future we want
The debate about AI is often presented as a choice between enthusiastic adoption and cautious resistance. This is a false choice.
The real challenge is finding the balance. In doing so we need to reposition AI as a tool not an outcome, and avoid the risk of it becoming ‘a hammer looking for a nail’.
We need to embrace AI’s extraordinary potential while recognising its environmental impacts. We need to improve productivity while investing in people. We need to accelerate innovation while maintaining scientific rigour. And we need to move quickly while retaining public trust through good governance.
AI will undoubtedly transform weather and climate science and services. In many ways, it already has. The question now is how we shape it in a way that is not only more capable and efficient, but also more sustainable, trustworthy, responsible and ultimately human. How we use AI as a tool to amplify human endeavour.
This echoes the Human-Centered Artificial Intelligence championed by Stanford University where the purpose is ‘advancing humanity through AI and data science’[xviii]. In this approach AI is developed for and with humans, ensuring AI technology ‘amplifies, augments, enhances, and empowers humans rather than harms or replaces humans’[xix].
For me, this sounds like a worthy North Star.
[i] https://www.bbc.co.uk/news/articles/cx2pj502ev6o Accessed 21/08/26.
[ii] https://www.bbc.co.uk/programmes/m002dyx6 Accessed 21/08/26.
[iii] https://www.bbc.co.uk/news/articles/c248myy0jr1o Accessed 21/08/26.
[iv] https://www.bbc.co.uk/news/articles/cvgjyj9292mo Accessed 21/08/26.
[v] https://www.metoffice.gov.uk/research/ai/ai-in-weather-science Accessed 21/08/26.
[vi] Bi, K., Xie, L., Zhang, H., Chen, X., Gu, X. and Tian, Q., 2022. Pangu-weather: A 3d high-resolution model for fast and accurate global weather forecast. arXiv preprint arXiv:2211.02556.
[vii] https://www.turing.ac.uk/blog/fastnet-under-bonnet-our-ai-model-weather-prediction. Accessed 10/08/26
[viii] Bouallègue, Z.B., Clare, M.C., Magnusson, L., Gascon, E., Maier-Gerber, M., Janoušek, M., Rodwell, M., Pinault, F., Dramsch, J.S., Lang, S.T. and Raoult, B., 2024. The rise of data-driven weather forecasting: A first statistical assessment of machine learning–based weather forecasts in an operational-like context. Bulletin of the American Meteorological Society, 105(6), pp.E864-E883.
[ix] Kent, C., Scaife, A.A., Dunstone, N.J., Smith, D., Hardiman, S.C., Dunstan, T. and Watt-Meyer, O., 2025. Skilful global seasonal predictions from a machine learning weather model trained on reanalysis data. npj Climate and Atmospheric Science, 8(1), p.314.
[x] de Burgh-Day, C.O. and Leeuwenburg, T., 2023. Machine learning for numerical weather and climate modelling: a review. Geoscientific Model Development, 16(22), pp.6433-6477.
[xi] Dueben, P., Bauer, P., Fuhrer, O., Koldunov, N. and Kristiansen, J., 2026. Machine learning is revolutionizing weather forecasting–the next step is a change in how we work. arXiv preprint arXiv:2606.25076.
[xii] Nipen, T.N., Haugen, H.H., Ingstad, M.S., Nordhagen, E.M., Salihi, A.F.S., Tedesco, P., Seierstad, I.A., Kristiansen, J., Lang, S., Alexe, M. and Dramsch, J., 2026. Regional data-driven weather modeling with a global stretched grid. Artificial Intelligence for the Earth Systems, 5(2), p.250001.
[xiii] https://www.metoffice.gov.uk/blog/2026/can-ai-developments-really-be-green
[xiv] https://www.metoffice.gov.uk/blog/2026/can-ai-developments-really-be-green Accessed 10/08/26
[xv] https://www.metoffice.gov.uk/api/assets/file/a5-brochure—strategy-launch-low-respdf?prefix=assets Accessed 10/08/26
[xvi] Gerlich, M., 2025. AI tools in society: Impacts on cognitive offloading and the future of critical thinking. Societies, 15(1), p.6.
[xvii] Wiles, E., Hsu, M., Bedard, J. and Kropp, M., 2026. Putting AI on the Org Chart: Evidence on Delegation and Oversight. Working Paper, 2026. URL https://www. emmawiles. com/storage/ai_employee. pdf.
[xviii] https://hai.stanford.edu/about Accessed 07/1/26.
[xix] Xu, W., 2026. Human-Centered Artificial Intelligence (HCAI): Foundations and Approaches. arXiv preprint arXiv:2601.01247.
Leave a comment