To AI or not to AI? A sustainability guide

Similar to Hamlet himself, we're living in a moment of many messy contradictions. Across organizations, the message is clear: lean into AI with curiosity, experiment boldly, and integrate it into your workflows. AI tools promise to make us faster, more innovative, and more productive. Yet beneath this enthusiasm lies an uncomfortable truth: that every query, every generated image, every AI-assisted task comes with an environmental cost. Data centers consume enormous amounts of energy and water, contributing to carbon emissions and straining local resources. The very tools designed to help us work more efficiently are also accelerating the climate crisis we're trying to solve.

This tension isn't going away. But it also doesn't mean we should abandon AI entirely or use it recklessly. Instead, we need to exist thoughtfully in this area of trade-offs, making deliberate choices about when and how we engage with these powerful tools. This article focuses on what you, as an individual practitioner, can do to use AI more sustainably. But it's crucial to recognize this is just one level of intervention. Systems thinker Donella Meadows taught us that individual behaviour changes rank relatively low as leverage points compared to structural interventions like changing system rules or shifting collective paradigms. The practices outlined here become exponentially more powerful when adopted collectively – across your team, your organization, and the broader communities you're part of. If you're in a position to influence policy, culture, or infrastructure decisions, that's where the real leverage lies. Individual sustainable AI practices are an essential starting point, not the endpoint.

When to use AI (and when not to)

Before we can talk about using AI sustainably, we need to address a more fundamental question: when should we use AI at all? Not every task benefits from automation or prompting, and sometimes the most sustainable choice is not to use AI at all.

When AI adds genuine value

AI excels in specific scenarios where its capabilities align with actual needs:

• Generating quantity for quality. When you need to rapidly produce a large number of ideas during brainstorming sessions or creative exploration, AI can increase your chances of finding outcomes you can use. The value isn't in any single output, but in the volume that allows you to identify gems and excellent “jumping off” points.

• Amplifying expert judgment. AI works best alongside human expertise, not as a replacement for it. When experts can quickly assess and correct outputs, AI becomes a powerful multiplier, particularly in complex domains where nuanced judgment is crucial.

• Summarizing and synthesizing information. For condensing lengthy materials where the cost of errors is relatively low, such as creating overviews from reports or articles, AI genuinely saves time without incurring significant risk.

• Translation across formats and audiences. AI shows real power in reworking content across different styles or audiences, or quickly transforming raw notes, lists, or tables into more usable formats.

• Handling routine, low-value tasks. For time-consuming but low-impact work, such as specific standardized status reports (after verifying that the organizational value is truly low), automation makes sense.

• Coding and initial feedback. Research consistently shows AI is beneficial for various programming tasks and for providing first-pass feedback on documents or strategies, accelerating iteration cycles.

• Exploring multiple perspectives and filling knowledge gaps. AI can generate diverse viewpoints or role-play different personas for feedback and creative brainstorming, expanding your thinking beyond your own mental models. You can also train AI to identify any blind spots or assumptions you might be missing. AI can also provide mentorship, help create drafts and prototypes, and fill gaps where human expertise isn't readily available.

When to step back from AI

Just as important as knowing when to use AI is recognizing when not to:

• When effort is the point. For tasks where intellectual struggle and pursuit are essential, such as deep learning, writing, academic work, and complex problem-solving, using AI can risk missing out on genuine understanding or vital growth opportunities. Sometimes the process matters more than the output. 

• High-accuracy requirements. Where precision is vital, AI's tendency toward errors and so-called hallucinations creates unacceptable risk. While you may notice some mistakes AI is making, some can be subtle and difficult to spot unless you are highly familiar with the subject matter.

• Unfamiliar failure modes. If your team doesn't understand AI's specific risks and error patterns (such as persuasiveness, being overly complimentary, and missing contextual information), it is best to avoid relying on it for critical decision-making.

• AI's evolving weaknesses. Be cautious about assuming AI's capabilities. Some seemingly "simple" tasks may fall outside AI's strengths, and its abilities evolve unpredictably. Ongoing experimentation and knowledge-sharing within your community help everyone stay informed.

How to

Once you've determined that AI genuinely adds value to a task, the next question is how to implement it sustainably.

Actions you can take individually

• Track your AI usage. Pay attention to how often and why you're using AI. Notice patterns in your work where AI becomes a default rather than a deliberate choice.

• Make sustainability personal. Set your own goals for mindful AI use, like "only use AI for tasks that save more than 15 minutes" or "batch AI requests once daily instead of continuous queries."

• Question before you prompt. Before using AI, ask: Is this necessary? Could I do this myself in a reasonable time? Will the output actually be used? Making this a habit reduces waste at the source.

• Choose efficient tools, models, and approaches. Minimize the data you send to AI tools. Paste only relevant sections, not entire documents. Don't use the most powerful model for simple tasks. Basic formatting or simple questions don't need advanced AI; match the tool to the task's complexity.

• Batch similar tasks. Group AI requests to reduce redundant context-loading and minimize repeated prompts for similar work. It’s actually often better to start a new chat for something that only needs a quick prompt, as it won’t need to load all the history and context.

• Clean up digital waste. Regularly review and delete old AI workflows, unused saved prompts, obsolete tool integrations, and unnecessary data. Digital waste is physical.

• Choose sustainable tools. Opt for AI tools with transparent sustainability practices. Prefer local-first tools when they meet your needs, and use browser extensions or desktop apps strategically rather than always-on web services.

• Building better agents: One of the most effective ways to use AI more sustainably is to reduce the back-and-forth clarifications. Creating well-configured AI agents with specific rules and criteria helps you get what you need faster, ultimately reducing the number of prompts required.

• Share progress and collaborate. Foster a culture of awareness by sharing environmental learnings, promoting training, and working with colleagues and stakeholders for greener practices.

Actions you can take as a team

• Validate AI outputs as a team. Build a "second opinion" culture, don't rely on solo validation. Group reviews help catch hallucinations and assess risk.

• Continuously re-evaluate AI use. As AI evolves, revisit guidelines, share use cases, and learn from mistakes to stay ahead of change.

• Remove non-essential features. Reduce scope creep by focusing only on essential elements, lowering complexity and risk.

• Anticipate consequences and align values. Use AI thoughtfully, keeping a close eye on its societal impact and aligning practices with sustainability goals.

The question remains

So, revisiting the question posed in the beginning: to AI or not to AI? Unlike Hamlet's terribly tragic choices, ours need not end in catastrophe. To avoid such a disastrous finale, it requires deep reflection, critical thinking and intentionality.

AI is a powerful tool when used wisely, particularly for tasks that are high-volume, routine, or require broad but not deep understanding. For critical, creative, or high-stakes work, human effort and judgment remain irreplaceable. The question isn't whether to use AI, but when and how. Taking a critical lens means asking not just "can AI do this?" but "should AI do this?" It means thinking strategically about what you're trying to accomplish and whether AI genuinely serves that goal. Sometimes the most sustainable choice is the simplest one: doing it yourself, or not doing it at all.

This raises more profound questions about skill development and human capability – our own modern version of "to be or not to be." What skills do you want to maintain and nurture yourself, and what are you comfortable offloading to AI? If you never struggle or invest in your writing skills, do you lose the ability to think through complex arguments? If you always rely on AI for first drafts, do you develop the ability to think independently?

These aren't easy questions, and the answers will differ for each person and each context. However, by embedding sustainable practices, conducting ongoing evaluations, and keeping these critical questions at the forefront of our work, we can ensure that AI adoption drives real value while minimizing environmental impact and preserving the human capabilities that matter most. 

The goal isn't to reject AI or to use it thoughtlessly. It's about existing thoughtfully in the tension, making intentional choices that serve both our immediate needs and our long-term responsibility to the planet and to ourselves as thinking, creative humans. The question "to AI or not to AI" doesn't demand a single answer – it demands that we keep asking.