Reference
Explore
AI has a footprint. Pretending otherwise doesn't help anyone.
Map it, measure what matters, and use AI on purpose, not less, but deliberately.
Estimated organizational AI-related carbon footprint
Scope 3 CO₂e (EPA grid average)
kg CO₂e
Real-world context (rough equivalents)
Illustrative one-way drive distance
Eames Lounge Chair equivalent (embodied carbon)
You can't reduce what you
haven't named.
Most teams can't say what AI they're running, how often, or for what. That's not negligence, it's how tools get adopted. You install them, you use them, you never inventory them.
But AI tools aren't free to run. They consume energy, and the amount varies a lot with the model, the frequency, and where it runs.
So start here: map your AI footprint. It isn't a data science exercise, it's a ten-minute conversation with a piece of paper.
Calculates a relative compute tier from your entries and shows the result below. Nothing leaves this page.
Your rows are saved in this browser (refresh-safe). Use “Copy worksheet as text” or CSV to share.
Tracking our impact from the start.
Token volume.
The most honest proxy for compute, where platforms expose it. More tokens = more compute = more energy. Imprecise. Still better than nothing.
Model selection.
This is where most of the variance lives, and most teams skip it on purpose. The biggest model is rarely the right one. Match the model to the task and you win on both quality and cost.
Large models
Best for: complex reasoning, multi-step synthesis, nuanced judgment calls. High cost per token.
Balanced models
Best for: drafting, summarizing, Q&A, most daily work tasks. Good quality, lower cost.
Efficient models
Best for: classification, extraction, structured data, simple single-turn tasks. Lowest compute cost.
On-device vs. cloud inference.
Matters more as AI moves closer to local models. On-device is generally more efficient for simple tasks. Cloud is necessary for complex ones. Knowing the difference is an emerging skill worth building now.
Avoided emissions.
Underused as a frame. When AI replaces a process that had its own footprint, a business trip, a print run, a redundant approval cycle, that displacement counts. So when you add a workflow to your inventory, ask what it replaced. The answer changes the net math. Name it. Track it.
Three levels.
Start at the one that's yours.
Right-size your model.
GPT-4 for a subject line is like driving a semi truck to pick up a sandwich. Use a smaller model for simple, high-frequency tasks. Most platforms let you choose. Most people don't.
Build a use-case filter.
Before deploying a new AI workflow, name one thing it replaces. That replacement had a footprint too, and the net matters. Then ask: what model does this actually need? What's the frequency? If the answers point toward a high-compute, high-frequency workflow that doesn't replace anything, that's the conversation to have before you build, not after.
Procurement questions that matter.
These are reasonable questions. Vendors who won't answer them are telling you something.
Use AI deliberately, then put it to work.
Sustainability is one lens. Here's where most associates go next.
Review before you send, especially when the output sounds confident.
Footprint mapped. Now practice deliberately.
Getting Started path →MillerKnoll AI · 2026