On September 23, at Green IO London* 2025, 🌵 Aguaro co-led a workshop alongside Capgemini, represented by Sunil Bharadwaj, Senior Director of Sustainability, and Matthieu Poulard, CEO of Aguaro. Titled “How to mitigate the environmental and social costs of AI projects in large organizations?”, the session brought together Green IT professionals to explore Artificial Intelligence’s real impacts — from energy consumption to social effects — and the concrete levers that can make projects more sustainable.

The discussion focused on moving from theory to practice: turning sustainability into something operational, where every project manager can act directly on the environmental footprint of their initiatives.

Artificial Intelligence is being seen today as the next big thing — the ultimate productivity tool. It comes with the promise of optimization, faster decision-making, automation, and a host of streamlined benefits.

Yet these promises come with a cost the naked eye can’t see at first glance. A cost that won’t remain invisible for long — and one that raises urgent concerns.

The hidden cost of AI

The AI models we rely on today require extensive training, and that training demands massive computing power. But training isn’t the only driver. Once deployed, every time an AI model generates a response or processes data — especially during inference** — it also consumes energy. And because models are used millions of times a day, this phase now represents most of AI’s total energy use.

As AI adoption accelerates, this translates into soaring energy consumption and growing pressure on critical resources already under strain. At the same time, AI reshapes roles, raises questions of transparency, and influences how people perceive and trust decisions.

This raises a fundamental question: does AI end up causing more harm to the environment and the society than the benefits it delivers? How to limit the negative outcomes and restore the balance?

From hype to responsibility

For large organisations, the real challenge is no longer whether to adopt AI — it’s already here, widely promoted and implemented. But in many executive discussions, there is still a strong “Fear of Missing Out”: a rush to embrace AI before the competitors do, even when the environmental and social implications aren’t yet clear.

The challenge now is how to do it responsibly, which means asking the hard questions before rolling out new systems:

• “Are we applying AI to real problems with real organisational or societal value, or just chasing hype?”

• “How do we limit unnecessary model complexity and favor leaner, frugal approaches?”

• “Can our infrastructure align with sustainable & renewable energy rather than fossil fuels?”

• “What governance is in place to manage risks — from bias to workforce impacts?”

From awareness to action

The environmental footprint of AI is not an abstract issue. Training a single large generative model can emit as much CO₂ as 24 round trips from London to New York — every day.

If current trends continue, ICT could represent 20% of global energy consumption by 2030.

Beyond energy, there are also social costs: data worker exploitation, local pollution in vulnerable communities, and growing reputational risks for companies deploying AI at scale — from biased or discriminatory outputs to data protection breaches and public backlash over transparency or working conditions.

For now, most of these environmental and social impacts remain hidden. In order to take action, organizations first need to make them visible and measurable. That is precisely what drives us at Aguaro.

Aguaro was created with one goal: to operationalize sustainability and make it part of everyday work.

Because the solution runs natively on ServiceNow, operational teams don’t just see numbers — project and product managers can act on them directly in the tools they already use.

With Aguaro, organizations can measure, model, plan, and act on the environmental costs of their IT operations — from infrastructure and devices to cloud, licenses, energy, and consulting.

The Aguaro for Projects module goes one step further by giving project managers real autonomy. They can assess their project’s footprint at any stage, compare low- and high-carbon scenarios, test design options, and check alignment with eco-design, frugal AI, and frameworks such as RIA31**. They can also apply concrete recommendations — from action cards to design or procurement choices — and collaborate with stakeholders through integrated workflows that ensure these decisions are tracked and implemented, not just reported.

Aguaro’s calculation engine draws on 6,000+ up-to-date emission factors (including 400 specific to AI, and extensible with external databases) and supports multi-criteria modeling across energy, water, and the depletion of rare resources — increasingly important as GPU-intensive workloads grow.

This approach turns sustainability from a constraint into a lever: a source of trust, resilience, and often savings .

As Matthieu Poulard and Sunil Bharadwaj highlighted during the workshop, uncertainty is not an excuse to wait — it’s a variable to manage. Like in finance, you don’t stop acting just because you don’t have perfect data; you improve precision while moving forward.

Embedding environmental metrics into existing tools — rather than isolating them in a separate platform — helps organisations make early trade-offs when they still have room to adapt.

Turning insight into strategy

Once the social and environmental impacts are visible, organizations can move from awareness to real mitigation. That means making deliberate technical choices — using

• appropriately sized models instead of always chasing the biggest

• deploying in renewable-powered data centers

• optimizing code, and integrating carbon accounting into the AI lifecycle

But it also means building the right structures around it:

• multidisciplinary teams that mix tech, sustainability, and social expertise

• transparent reporting

• and risk assessment frameworks that make accountability part of everyday operations

As Sunil emphasized during the workshop, this isn’t just about ethics — it’s also smart business. Reducing energy and resource use saves money. Designing for efficiency lowers risks. And being able to prove responsible AI builds trust with customers, regulators, and investors.

That’s exactly where Aguaro makes the difference. Our solution turns these principles into day-to-day practice — embedding environmental metrics, automated data collection, and scenario modeling directly into existing management tools. It allows teams to connect environmental performance with business performance and translate sustainability into concrete results.

To go further

If you’d like to dig deeper into these challenges, download our latest white paper — “Artificial Intelligence: Origins, Challenges, and Environmental Impact.”

It explores the history of AI, the forces behind its exponential rise, and the environmental footprint that often goes unseen — while outlining clear pathways for more sustainable adoption.

*Green IO London is a conference on digital sustainability, bringing together experts to share concrete solutions for a more sustainable digital future.
**Inference is when a trained model applies what it’s learned to new data to produce an answer.
**RIA31 is the Ethical and Sustainable AI framework developed by the French Institute for Sustainable IT (INR)