I recently talked with a venue sustainability director who has been doing this work for a long time, long enough to have watched their building's systems evolve from paper logs to real-time sensors. Real infrastructure. Real years of utility and waste data already sitting in their systems.

I asked what the hardest part of the job actually is. I expected an answer about budget, or staffing, or vendors. Instead, the answer was about a question.

The honest answer, was that it almost always starts with whether they're even asking the right question. Their team runs data exploration projects regularly, digging into utility consumption or waste patterns, looking for something actionable. Sometimes those projects run for months. Sometimes, at the end of those months, the answer is that the original question was the wrong one, and the work has to start over.

Real data. Real sensors already in place. And the bottleneck still isn't the data. It's knowing where to point it.

This isn't one venue's problem

It would be easy to read that story as one unusually candid person describing an isolated struggle. It isn't. It's the norm.

In our interviews with 20 major venues across the US, Europe, and Canada, 95 percent said they rely on manual processes to manage sustainability data, and 90 percent said staff bandwidth limits how much reporting or analysis they can actually do. One coordinator we spoke with described post-event analytics as tedious work "done in an old Excel sheet." Another put it more bluntly: "Because we don't have a dedicated data person, many things can fall through the cracks."

Notice what these two things have in common, the twenty-year veteran with a fully built-out sensor system, and the 95 percent of venues doing this manually. Neither is describing a data shortage. Both are describing the same thing: no clear, fast way to know which question in that pile of data is actually worth chasing before committing real time to it.

That same director also confirmed something we wrote about earlier this year: that waste bills routinely reflect an assumed fill level rather than anything measured, and there's rarely a way to check that number against what actually happened. Hearing it independently confirmed, by someone with decades of firsthand experience, is exactly the kind of validation research needs and rarely gets in this industry.

What actually helps here isn't a bigger model

The instinct, when a data problem this fundamental comes up, is to reach for the most powerful AI tool available and point it at the pile. We think that instinct is usually wrong.

A large, general-purpose language model is often overkill for this kind of task, and it comes with a real energy cost of its own, which is a strange thing to ignore when the goal is resource efficiency in the first place. What actually helps here is smaller and more specific: lightweight, purpose-built models that can look at a venue's consumption and waste patterns, explain the conclusions (no “black-box”), compare them against similar events or similar buildings, and surface the handful of questions actually worth a team's time, before anyone commits months to finding out the hard way. This all can be done with minimal hallucinations.

That's a different job than reporting, and a different job than a general AI assistant. It's closer to a research partner that has already ruled out the dead ends. Months of exploration become a starting list of a few good questions. The sustainability director doesn't need an AI that answers everything. They need one that stops them from spending a quarter chasing the wrong thing.

What if your venue could actually speak with you

All that infrastructure, and the building still can't tell anyone what it actually knows, not in a way anyone has time to listen to.

What if it could? Not as a novelty, but as an actual working relationship: a building that surfaces the one anomaly worth investigating this week, instead of leaving a team to comb through months of readings to find it themselves. A building that says, plainly, here's where your waste diversion dropped and here's the event that correlates with it, instead of waiting for someone to stumble onto the pattern by hand.

That is the conversation worth having first. Not "give me all the data." Just, what's actually worth asking about today.