What Sustainability Leaders Need to Know About AI Right Now

The pressure to adopt AI in sustainability is real. Leaders at every level face it: from the board asking about automation to vendors pitching tools that promise to do everything. But a critical finding from our latest webinar stopped the room: sustainability teams adopt AI at roughly half the rate of the rest of their own organisations. And most of the teams experimenting with it never make it out of the pilot stage. 

The problem is not AI. The problem is the wrong AI, applied to the wrong problem, without the right foundation underneath it. 

Your Data Has to Come First 

Jeff Floyd, VP of Data Operations at SE Advisory Services, put it plainly during the session: "AI can take bad data and make wrong decisions faster." 

He shared a real example. An AI rate engine recommended a hospital to switch electricity tariffs to reduce costs. What the AI failed to flag: the new tariff gave the utility company the right to cut power during peak demand days. For a hospital, that is not a cost-saving decision. That is a catastrophic one. 

The lesson here is not that AI is dangerous. It is that AI without context, without the right data foundation, and without domain expertise behind it produces confident answers to the wrong questions. In sustainability, where you have to defend every number to an auditor, a regulator, or a board, "close enough" is not a standard anyone can afford. 

"AI on the Label" Is Not the Same as AI That Works 

Every software vendor lists AI in their pitch deck now. Jeff Willert, Director of Data Science at SE Advisory Services, gave the audience a practical test to cut through the noise: ask what happens after the AI does its work. 

If the answer involves manually copying an output to the next step, you are not buying intelligence. You are buying a smarter silo. Real AI in sustainability connects the entire workflow: data intake, validation, calculation, reporting, and disclosure. It does not stop at step one. 

Willert demonstrated this live during the session, walking through how Resource Advisor+ (RA+) Schneider Electric's AI-native platform for enterprise energy and sustainability management,  handles anomaly detection for a hospitality client. The platform does not apply one-size-fits-all pattern matching. It understands that a hotel's energy consumption behaves differently from a warehouse, accounts for variables like occupancy and weather, and then resolves flagged issues before they reach reporting. Every decision in that process is auditable and traceable. 

Domain Expertise Is Not a Feature. It Is the Foundation 

When Willert was asked whether it matters who builds the AI, his answer was direct: "It matters enormously." 

General purpose AI tools like ChatGPT or Copilot are trained to give you a helpful answer. They are not trained to understand your portfolio, your baseline emissions, your jurisdiction-specific reporting frameworks, or which energy conservation measures your facilities have already implemented. Willert tested this live: he asked Copilot for his commute emissions. Copilot gave him an answer. It never asked whether he drives an electric vehicle. 

The AI inside RA+ is built by people who have spent decades doing this work. They understand how energy data behaves across seasons, how COVID-level disruptions break model assumptions, and how to train a system to say "I don't know" rather than produce a confident wrong answer. That is a design decision, not a technical feature. 

AI Amplifies Experts. It Doesn't Replace Them 

One of the most honest moments in the session came when Floyd addressed the question many leaders avoid: what happens to the people doing this work? 

His answer was two-sided. For transactional work, AI will change things, and teams that do not upskill will feel that. But for the experienced professionals in the room, the veterans with ten or twenty years of domain knowledge, AI becomes a multiplier. Floyd's teams are now the biggest users of the AI they helped build. They teach it, test it, and direct it toward the problems that matter. AI creates the capacity for those experts to focus on strategy, not data wrangling. 

The session also addressed a concern that often goes unspoken: the energy cost of AI itself. Not all AI carries the same carbon footprint. RA+ applies what Willert calls "frugal AI" principles, using lightweight statistical and machine learning techniques where they fit the problem, and reserving generative AI for the tasks that genuinely need it. Using AI responsibly means applying the right tool at the right scale. 

Three Things to Take Away 

• Sustainability is a domain-specific challenge. AI built for it must come from people who fundamentally understand the field, the data, the regulations, and the outcomes. 

• Connected workflows beat point-in-time tools. The entire journey from data intake to disclosure needs to work as one system, not a collection of separate steps. 

• You do not need perfect data to get started. The right AI builds trust into data progressively, catching errors and filling gaps responsibly as you go rather than waiting for a clean foundation that never arrives. 

Go Deeper on Both 

If this conversation sparked questions about how to evaluate AI tools for your own sustainability program, our resources will help you move from thinking to action. 

Download The Sustainability Leader's AI Evaluation Guide: A practical framework to help you cut through vendor claims, ask the right questions, and identify whether an AI approach will actually hold up for the complexity of enterprise energy and sustainability work.