AI for Sustainability: Facing the Ethical and Responsible Conundrum

AI is permeating our lives at an unprecedented pace. From banking and healthcare to manufacturing and logistics, AI and agentic systems are being rapidly embraced and tested.

Talk about AI for sustainability, the conversation suddenly becomes more cautious. Almost hesitant. Why?

Because sustainability invites direct scrutiny. The moment AI is applied to climate, carbon, or ESG challenges, the questions get sharper: How sustainable is your AI itself? What about its carbon footprint? Are you solving one problem by creating another?

These are valid questions—but they may also be holding us back.

The Fear of Being Judged Shouldn’t Stall Progress

If we pause and reflect, we don’t apply this level of hesitation elsewhere. We don’t reject AI in healthcare because data centers consume energy, nor do we question AI in finance because algorithms require compute. Instead, we focus on responsible use, governance, and outcomes.

So why should sustainability be any different?

Avoiding AI for sustainability out of fear of judgment is not a responsible choice—it’s a missed opportunity. If anything, the complexity and urgency of climate and resource challenges demand the most powerful tools we have. The challenges we face—carbon accounting at scale, whole-life assessments, biodiversity impacts, supply chain transparency, transition risk—are deeply data-intensive and interconnected. Solving them manually, or with fragmented tools, will be slow and insufficient.

Using AI irresponsibly is never justified—whether in healthcare, finance, or climate. But not using AI at all, when it could accelerate solutions for a sustainable future, is equally problematic.

What Responsible AI for Sustainability Actually Looks Like

The ethical question isn’t whether we use AI for sustainability—but how. Responsible use demands intention, discipline, and design:

Carbon-aware model usage: Train large models thoughtfully—once where possible, not repeatedly—and prioritize reuse and fine-tuning over brute-force retraining.

Sector-specific intelligence: Use targeted, domain-aware prompts and agents rather than generic, compute-heavy approaches.

Expert-led deployment: In enterprise AI adoption, allow experts to guide nuanced analysis and research, reducing unnecessary compute and noise.

Outcome-driven design: Ensure AI applications directly enable measurable sustainability outcomes—not just dashboards or disclosures.

Awareness and governance: Build literacy around both the benefits and the footprint of AI-enabled sustainability operations.

Yes, AI will have a carbon footprint. But so does every other technology we deploy at scale. The real question is this: What is the net impact?

The Bigger Picture: Net Positive at Scale

For hard to abate sectors like the Built environment, if a relatively small AI footprint enables large-scale emissions reduction, resource efficiency, transparency, and faster climate action, the trade-off is not only justified—it’s essential. The built environment industry contributes to 39% of global GHG emissions and will continue to emit—whether we use AI or not. Decisions made today will define sustainability outcomes long after current debates fade. When applied responsibly, AI becomes more than a technology—it becomes a decision partner across the lifecycle of the built environment.

The difference is whether we accelerate transformation to achieve :

• Lower embodied carbon at design stage

• Smarter material choices at scale

• Reduced retrofit costs and emissions over decades

• Better-performing, more resilient assets

Then the net outcome is overwhelmingly positive. Not because it is perfect.

But because it allows us to move faster, think systemically, and act earlier—when it matters most.