Whether you're a sustainability professional or interested in responsible AI adoption, three questions likely keep coming up:
Where do we start?
How do we use AI effectively?
What's the environmental impact?
Last month, I led a discussion with ~40 OnePointFive sustainability professionals that revealed insights valuable for any team adopting AI. This post explores AI use in sustainability and what the rest of us can learn from broader questions about responsible technology adoption.
Where We Are Today
The current landscape of AI adoption in sustainability reveals three common scenarios:
- Many teams use tools like ChatGPT for basic tasks but feel uncertain about best practices.
- Others recognize the potential of AI but feel overwhelmed by the numerous options and potential risks.
- A significant number are holding back due to concerns about adding to their organization's environmental footprint through AI use.
The good news? There's a practical path forward that addresses all these concerns.
An AI Adoption Framework That Works
The key to successful AI adoption is starting with one high-impact, low-risk task.
Let's look at a practical example: data analysis.
If you work in sustainability, you're probably familiar with emissions data like carbon dioxide released from a manufacturing plant. Using a tool like Microsoft Excel with Copilot or Google Sheets with Workspace AI, you can quickly identify trends and patterns in your data. These tools are familiar and accessible, making them perfect for beginners.
Here's how to structure your workflow:
- Data Preparation:
First, organize your data in a clear, structured format. Remove any sensitive information and ensure data quality. Tools like Excel's Power Query or Google Sheets' data cleanup features can help automate this process.
- Initial Analysis:
Use AI to identify trends, flag unusual patterns, and compare performance across different metrics. For example, you might ask Copilot to "analyze the relationship between energy usage and emissions across different facilities" or "identify months with unusual spikes in emissions."
- Expert Review:
Human experts should then validate the findings against known operations, investigate any anomalies, and provide context for variations. This is where your sustainability expertise becomes crucial.
To take this further, a responsible workflow also includes:
Setting clear boundaries: Document what AI can and cannot do in your processes. For example, AI can help spot patterns in data, but decisions about sustainability targets should always involve human judgment.
Verifying AI outputs: Create a verification process that includes technical accuracy checks, compliance review, and expert validation. Tools like Notion or Asana can help manage this workflow, ensuring nothing gets missed.
Ensuring transparency: Documentation is crucial to keep the workflows clear. Record which AI tools were used, what inputs were provided, and how outputs were verified. A simple spreadsheet or project management tool can track this info effectively.
Managing Environmental Impact
In addition to adoption, understanding and managing AI's environmental footprint requires a systematic approach. Let's break down what drives AI's energy use and how to minimize it.
Choose Efficient Tools
Not all AI is created equal. The environmental impact of AI models is primarily driven by their size (number of parameters) and how they're used. Larger models like GPT-4 (with over 1 trillion parameters) require significantly more energy than smaller models.
Each time you interact with these models, they perform billions of calculations, consuming energy and generating heat that needs to be cooled – much like how your laptop gets hot with heavy use.
For routine tasks, consider Claude-instant, GPT-3.5-turbo, or smaller, task-specific models for more efficient alternatives.
Remember: Not every problem requires AI. Sometimes a well-designed spreadsheet or traditional database query is more efficient and just as effective. Before implementing an AI solution, ask yourself: "Could I solve this with simpler tools?"
Optimize Your Usage
Think of AI processing like running a factory – it's more efficient to run at steady capacity than to constantly start and stop. Instead of processing items one at a time, group similar tasks together.
For example, if you need to analyze multiple sustainability reports, collect them and analyze them as a batch during off-peak hours. This approach not only saves energy but often produces more consistent results.
Track Your Impact
While most AI providers don't yet expose direct energy metrics, several tools can help you understand and manage your AI-related emissions:
- For individuals: AI Wattch Browser Extension to estimate carbon emissions from your ChatGPT conversations.
- For developers and data scientists: EcologITs to track the energy consumption and environmental impacts of using generative AI models through APIs. CodeCarbon to seamlessly integrate into your Python codebase.
- For cloud users: Major providers like Google Cloud and Microsoft's Azure offer their own tracking tools.
Remember: The goal isn't to avoid AI, but to use it thoughtfully and efficiently. Start by establishing your baseline emissions, set realistic reduction targets, and regularly monitor and adjust your approach. Every small optimization adds up to significant environmental savings at scale.
Taking Action Today
Ready to move forward? Here are your next steps:
Start Small
Book a free 30-minute workflow audit to identify your highest-impact, lowest-risk AI opportunities. Contact me on LinkedIn using the button below.
Say hello
Learn the Framework
Download my FREE comprehensive guide to AI in sustainability work. It includes templates, checklists, and real-world examples.
Get your AI guide
