The smartest AI will help us use less, not just do more

Speaking alongside Bill Nye at Dreamforce, Salesforce's annual conference, was not something I expected to find on my 2026 bingo card. But there we were last week in San Francisco discussing one of the increasingly important questions behind artificial intelligence: the energy required to power it.

The session was completely oversubscribed with some of my Kiwi colleagues watching the livestream from outside. Amid all the excitement about what AI can do this said to me that the conversation is starting to mature. Increasingly, we are also asking what does AI cost? Who benefits?  And how it should it be deployed?

The energy behind intelligence

AI sustainability was the impetus for our Dreamforce conversation, but the energy powering it was the focus and what Bill Nye talked about most. His message was straightforward: electrify everything and use less. For New Zealand, that presents both an opportunity and a responsibility. We have a highly renewable electricity system, but a significant share of our total energy use still comes from fossil fuels and so we can't stop here. The more we can electrify transport, industry and other parts of the economy, while reducing unnecessary energy use, the better positioned we will be.

At One NZ we've embraced AI across the business, while increasingly scrutinising where we use it and how. That was recognised at the Global Sustainability Awards the week before Dreamforce, where we won bronze in the Tech & AI category, behind NVIDIA with silver and Canon with gold. The recognition reflected not simply embedding AI into our operations but doing so responsibly and with clean energy in mind.

One example is our mobile network. Software and machine learning now allow parts of the network to enter short "microsleeps" during periods of very low demand, waking immediately when required. In our 2025 financial year this enabled a 14% reduction in electricity use. We've also reduced reliance on diesel generators by putting batteries on mobile cell sites which can be centrally optimised using AI when wild weather causes power outages. These applications may be less visually impressive than humanoid robots or generative AI demos, but they matter: intelligence can help physical infrastructure respond more precisely to real demand.

Sustainability is becoming part of the mainstream tech conversation

Sustainable technology is increasingly moving out of its own niche. As I write this, Climate Week NYC is now underway with a programme including AI and data centres, clean technology, electrification, energy efficiency, transport and investment. Environmental sustainability is becoming less of a separate conversation from technology and increasingly one of the conditions by which good technology will be judged.

That was all reflected at Dreamforce. Marc Benioff used his keynote not only for product announcements but to lean into questions around ethics and societal impact. This included conversations with leaders from Anthropic, OpenAI and NVIDIA about the responsibilities AI companies carry as the technology becomes more powerful. For businesses like ours, the takeaway is to consider AI's broader impacts and make conscious choices about where we use it and where we don't.

Deciding what we should keep human

Another session that resonated was Ethan Mollick and Paula Goldman exploring what it means to remain human when AI becomes an everyday collaborator. That technology can deskill us if we are not deliberate about what we outsource. AI can make us faster but faster is not always better. If we want to learn deeply, develop judgement or exercise imagination, automatically delegating work to AI may remove precisely the human capability we need to preserve.

Matthew McConaughey and will.i.am took that idea further, challenging the audience to imagine the world we want in 2046 and work backwards. I liked that framing because it restores a sense of agency. Technology does not simply happen to society: we make choices about where it is applied, whose problems it solves and who benefits.

When AI meets the physical world

Sometimes the most powerful examples are outside the conference hall. One of my favourite experiences in San Francisco was taking a Waymo. Sitting in the back seat while a car navigates pedestrians, intersections and city traffic with nobody behind the wheel gives you a very different appreciation of autonomous driving. I also found myself feeling surprisingly safe.

What interested me was not just the AI, but the convergence of machine learning, automation, electrification and cleaner energy. Sustainable AI should not only mean making AI itself more efficient. It should also mean applying intelligence to make transport, telecommunications, buildings, energy systems and other infrastructure more efficient.

Responsible AI should help us move faster

Responsible AI cannot just be a collection of principles on a corporate website. Teams need practical guardrails: clear decision points around risk, data, transparency, sustainability and customer impact.

Done well, those guardrails should not slow innovation. They should give teams confidence to move faster because people understand the boundaries and know when deeper scrutiny is required.

Trust matters too. Our 2026 AI Trust Report found that frequent AI users experienced the sharpest decline in trust, suggesting that greater familiarity with AI can also bring greater awareness of its limitations.

The choices are ultimately ours

After Dreamforce, the overarching thought I came home with was not whether AI will become more capable. That seems inevitable. The harder questions are ours.

Where should we use AI? Where should we deliberately preserve human judgement? How can we use AI to consume less energy rather than more? How do we ensure its benefits extend beyond the people already closest to the technology? And perhaps most importantly: are we using AI simply to do more, or to make the systems around us genuinely better?

The smartest AI will not just help us produce more, move faster or automate more tasks. It will help us use less, waste less and make better choices about what we do next.