If it feels like AI usage has become common practice in your company's day to day operations or AI integrations are ubiquitous across the software and service platforms you use to run your business, you may be wondering if you should be accounting for the extra emissions from your team's AI usage in your annual carbon inventory.
Accounting for AI use emissions is not yet a standardized practice and the necessary data to precisely measure your AI use emissions (like how many kWh of energy you consumed in your 10 minute convo with an AI chat bot) is largely inaccessible from the providers of common Large Language Model (LLM) AI tools. Though Google published a blog summarizing the environmental impact of Google Gemini citing median Gemini Apps text prompt uses 0.24 Wh of energy, 0.03 gCO2e, 0.26 mL of water.
However, there are lots of folks who want to quantify their AI use emissions to understand how they compare relative to other business activities. We've seen a number of free web based tools emerge that aim to give users of common LLM chatbots like, ChatGPT and Google Gemini, the ability to estimate their usage emissions. See the table below for a list of options that may be helpful in estimating your company AI use emissions.
When deciding whether to account for your team's AI elated emissions in your carbon inventory, you should assess if AI related emissions are material to your company's footprint. Greenhouse Gas Protocol does not set mandatory materiality thresholds, but emission sources contributing less than 1-5% of total emissions may be deemed immaterial, depending on the context.
For consumer goods brands, AI use emissions tend to be tiny relative to the emissions from all other business activities that are included in a corporate inventory (like energy use or making and shipping your products). It's unlikely that day to day AI usage by your company employees will contribute to your footprint in a material way.
As for AI integrations that are built into other software platforms you may purchase or subscribe to already to run your business, if you are accounting for the platform fees in your inventory it's probably not necessary nor practical to estimate the additional emissions from the AI functions that are integrated into those platforms, and may even result in double counting. Reporting the total service fees in your inventory is a conservative and practical way to account for the emissions from use of your software platforms.
Lastly, if you are a service based company that is developing and running LLMs as a core part of your business, AI related emissions are more likely to represent a material portion of your footprint. You should assess if activities such as running servers are a significant part of your company energy consumption and account for the emissions from your energy use in your corporate inventory.
Tool Name | Notes |
Methodology cited, developed by non-profit Code Carbon
| |
| UK based Methodology cited |
Methodology not cited. Reach out to Greenly for more info | |
Methodology not cited |