Like all data staff, sustainability professionals are attempting to navigate the accelerating deployment of synthetic intelligence together with its ravenous urge for food for power, water and different crucial inputs.
One other vital query is much less thought-about: What can we not afford to show over to AI? The place ought to a human-in-the-loop turn out to be a human calling the photographs? Listed below are three cases:
Don’t let AI dictate your degree of sustainability ambition
AI optimizes in opposition to the worth mannequin it inherits. For many firms, that mannequin is the financials as at present structured — the place carbon depth doesn’t present up in sourcing selections, water danger isn’t booked as a price and provide chain fragility is invisible till it turns into a disaster.
AI brokers educated on that mannequin received’t simply overlook sustainability worth; they’ll optimize in opposition to the recognized enterprise goal to the exclusion of different variables and time horizons. Ask AI to cut back procurement prices, and it’ll present an answer for right this moment, with out contemplating provider efficiency, influence on Scope 3 emissions or how that provider reveals up in its neighborhood. AI is now being utilized by groups throughout actual property, capital allocation, manufacturing and product improvement in precisely this manner.
Telling AI to “additionally take into account sustainability” is just not the answer as a result of directions get retrained out within the face of optimization. We’ve been speaking in regards to the repair for 30 years: embed the worth of what’s at present externalized instantly into the enterprise mannequin.
Somebody has to determine what goes into the mannequin earlier than it runs. Embedding sustainability issues into the corporate’s enterprise mannequin — translating danger and alternative into income — is the elemental process of sustainability professionals.
Don’t let AI substitute for forward-looking strategic enterprise perception
AI is a pattern-recognition engine. It identifies what has occurred up to now and tasks it ahead. It can not learn throughout techniques with no shared language, no frequent information construction and no historic report of intersection — which is strictly the work sustainability professionals do each day. We join ecological and social techniques that function on totally different timescales and metrics than the monetary system, and determine the place they’re going to collide with the enterprise mannequin earlier than the influence reveals up on the underside line.
Carbon emissions now present up as stranded property on oil and gasoline steadiness sheets and a tariff value on the EU border. Water shortage has rerouted semiconductor provide chains and delayed information heart deployment. Pressured labor publicity has triggered import detentions and bans that stalled stock for attire and auto firms at U.S. ports.
None of those had been surprises. They had been exterior indicators on a trajectory to intersect enterprise technique, they usually had been seen lengthy earlier than they hit the revenue assertion, at the very least to anybody educated to learn them.
Don’t deploy AI with no sustainability skilled within the room
Externalized prices are inclined to turn out to be company crises. Carbon emissions that firms did not e book as a price twenty years in the past at the moment are exhibiting up as stranded property and tariff publicity. Water traditionally handled as free is now a provide chain limitation. AI is already producing the subsequent spherical of crises.
Klarna changed roughly 700 customer support roles with AI between 2022 and 2024 and publicly celebrated the labor value financial savings. By 2025, the CEO was rehiring — not as a result of the AI failed at its process, however as a result of the optimization perform didn’t embody the very actual worth of human buyer expertise.
When xAI chosen Memphis for its Colossus supercomputer, the 100,000-GPU cluster got here on-line in 122 days inside a retrofitted manufacturing facility. The general public solely discovered of the challenge the day the Larger Memphis Chamber of Commerce introduced it in June 2024. To make up for power-grid shortfalls, the corporate put in dozens of unpermitted methane gasoline generators on-site — in a majority-Black neighborhood in South Memphis.
The ability is now dealing with Clear Air Act lawsuits and injunctions. This can be a textbook risk-management failure {that a} sustainability skilled within the deployment dialog would have flagged earlier than the primary turbine was put in.
These aren’t summary ethical issues. They’re enterprise design selections with materials penalties — for resilience, for regulatory publicity, for model, for the provision chains and communities the enterprise will depend on.
The sustainability skilled’s job has all the time all the time been to embed sustainability into the corporate’s working system — the mannequin the enterprise makes use of to allocate capital, consider danger and determine what’s value constructing. We obtained sidelined measuring influence and setting aspirational targets, as an alternative of shaping the choices that create these impacts. AI is now executing in opposition to these unchanged enterprise working fashions at machine pace.
It’s time for sustainability professionals to double down on the work we’re educated for: surfacing what isn’t but priced earlier than the market forces the correction.
On the NYU Stern Middle for Sustainable Enterprise, we’re constructing that functionality into the subsequent era of enterprise leaders. Sustainability isn’t a standalone topic; it’s aggressive intelligence: the self-discipline of studying what the monetary system can’t but see and translating it into strategic motion earlier than the market forces the difficulty.
College students who come by way of our program go away figuring out to ask, What is that this firm depending on? How is that altering? and What does it imply for monetary efficiency? That’s the techniques view that companies want, and solely people can present it.