Can AI be Regenerative?

AI is no longer experimental. It’s embedded in marketing, operations, customer service, analytics and product development. But here’s the uncomfortable truth: every AI model you train, fine-tune or prompt runs on energy-hungry infrastructure. The more we automate, the more electricity we consume. If businesses are serious about net zero commitments, AI can’t sit outside the sustainability conversation.

So how do organisations make AI use more sustainable while still capturing its value?

1. Measure the carbon cost of AI

You can’t manage what you don’t measure. Most businesses track travel emissions and office energy use, but very few track the footprint of cloud computing and model training.

Work with your cloud providers to understand the energy intensity of your workloads. Platforms like Microsoft Azure, Amazon Amazon Web Services, and Google Google Cloud now provide carbon reporting tools. Use them. Track model training cycles, storage, inference calls and idle compute. Make AI part of your Scope 3 emissions reporting.

If AI is driving value, it should also be accountable.

2. Right-size the model

Bigger is not always better. Large language models are powerful, but they are computationally expensive. Many business use cases do not require frontier-scale models.

Before defaulting to the largest model available:

• Ask whether a smaller fine-tuned model will do.

• Consider task-specific models instead of general-purpose ones.

• Reduce unnecessary inference calls by improving prompt design and workflow logic.

Optimisation is a sustainability strategy. Efficient models reduce energy consumption and cost at the same time.

3. Optimise training practices

Training models is far more energy-intensive than running them. Sustainable AI teams:

• Reuse pre-trained models instead of training from scratch.

• Limit experimental retraining cycles.

• Use transfer learning rather than building new architectures.

• Schedule compute-heavy tasks during times when renewable energy supply is higher (where available).

Some hyperscale providers already shift workloads across regions to optimise for renewable energy. Ask where your models are physically hosted. Data centre location matters.

4. Clean up data pipelines

Messy data equals wasted compute. If you’re feeding duplicated, low-quality or irrelevant data into models, you’re increasing energy use for marginal gain.

Good data governance reduces emissions indirectly. Clean datasets mean fewer training iterations and leaner storage requirements. Archive or delete redundant datasets rather than storing everything “just in case.”

Digital clutter has a carbon cost.

5. Design AI use cases that replace high-emission processes

Sustainable AI isn’t just about reducing its footprint. It’s also about what it replaces.

For example:

• AI-enabled logistics optimisation can reduce fuel consumption.

• Predictive maintenance can extend asset life and reduce material waste.

• Smart energy systems can lower building energy use.

• Virtual collaboration tools can reduce travel.

When AI displaces high-carbon activities, the net impact can be positive — but only if you calculate both sides of the equation.

6. Embed sustainability into AI governance

Most organisations now have AI governance frameworks focused on risk, bias and privacy. Sustainability should sit alongside these considerations.

Build sustainability checkpoints into:

• Procurement decisions

• Model selection

• Architecture design

• Vendor evaluation

• Reporting dashboards

Treat carbon intensity as a design constraint, not an afterthought.

7. Align AI strategy with climate targets

If your business has committed to net zero by 2030 or 2050, your AI roadmap must align with that trajectory. Otherwise, you’re creating internal contradiction.

Forward-looking organisations are integrating AI strategy into broader ESG frameworks. They are asking:

• Does this AI initiative reduce overall emissions?

• Is it energy-efficient relative to alternatives?

• Can we prove its environmental return on investment?

AI can accelerate sustainability transformation — but unmanaged, it can just as easily increase emissions.

The strategic shift

The next phase of digital transformation isn’t just intelligent — it must be responsible.

Sustainable AI requires systems thinking. It means understanding the upstream energy impact of compute infrastructure, the downstream behavioural effects of automation, and the organisational incentives driving adoption.

The businesses that win won’t simply use more AI. They’ll use smarter, leaner, accountable AI — designed not just for performance, but for planetary constraints.

That’s not a trade-off. It’s a competitive advantage.