Blog

T. Wang, Bard College MBA in Sustainability

We all know a compulsive documentarian. Not a moment can be experienced without a carefully crafted tableau, hand selected filter, or pithy hashtags. In fact, this might be you. I know I started slipping into this mindset some years ago. I itched to capture, frame, and share a diary of stills and fleeting thoughts. I even threw myself into online debates. I meticulously dissected every opposing argument, drafted lengthy dissertations on topics ranging from politics to pop culture. I became addicted to the “reply” button and craved the satisfaction of “the last word”. I waited for the approval of acquaintances and strangers meted out in pixel hearts and thumbs turned up. But at some point, I came to realize that my participation was not yielding what I thought it would. I felt no lasting satisfaction in having proverbially screamed into the void. A view counter aggregating algorithmic attention to a photo of my silhouette in a sunset is ultimately hollow compared to my memory of experiencing the real thing.

I started to resent the digital networks that have become ubiquitous social tools, especially as alarming information came to light regarding the way our information and attention are monetized, collected and controlled. Now apparent, these tools do not exist as an altruistic benefit to humanity, they reflect and support the capitalist system they were born out of. We have come to accept that, as users, we provide the largely unpaid labor of content creation, that our information will be harvested and sold and that the very same information will be used to direct the algorithm that keeps us scrolling.¹ We are convinced that to succeed professionally and socially, we must participate in this system of extraction and control, and we should replicate our own commodification by treating ourselves as living brands.

Our Workspaces

We dove headfirst into the connectivity that was promised by technology without interrogating what kinds of social norms the new technology is encouraging. What happens when the capitalist love for productivity is proliferated and distilled through social media? Staff writer at New University and sociology student, Deanza Jayaputri Andriansyah of U.C. Irvine, described it as the “LinkedIn effect” in her 2025 article describing how the confluence of social media and hustle culture is leading to burnout among Gen Z professionals. Treating ourselves as living brands means we are constantly eroding the boundary between leisure and work. As we consume content about someone else’s “hustle”, we are encouraged to match or outperform what we see. These expectations are clearly unsustainable and likely lead to uneven outcomes for workers from different backgrounds. For example, neurodivergent workers struggle in the work place more so than their peers.² From my personal experience as a neurodivergent human and professional, work life balance is already challenging to maintain even without the added pressure to curate an identity. And does this constant curation lead to authentic representation? Does it lead to better outcomes? Higher quality of outcomes? More creative outcomes? More ethical outcomes?

To maintain sustainable workspaces, we need to be able to draw and feel clear boundaries when work begins and ends. We should also redefine success to include personal time and interests outside of professional life. It may seem counterintuitive, but studies show that organizations that promote better work-life balance actually see better productivity from their workforce.³

Our Relationships

Despite social media actually doing little to alleviate loneliness for most people,⁴ I understand the allure of even simulated connectivity. What if we decided to focus on our micro-communities again? I came across this Axios article about how Americans (especially younger Americans) associate with our neighbors less and less. I understand individuals who may struggle to find community within their immediate geography. I am an immigrant and woman of color who grew up in rural areas with low diversity. I often felt alienated from my peers, but these less than ideal experiences taught me how to search for common ground in unlikely situations.

Last year, I struck up a conversation with a neighbor down the road. We bonded over her homespun apiary and our shared love of gardening. However, it became quickly apparent that we hold wildly different political perspectives. She also expressed opinions that are counter to scientific consensus on climate change and other environmental issues. At this point, it would be common in an online discourse for one party to shame the other, but why would I speak like that to a neighbor? Instead, I appealed to our common interests to build a relationship of understanding. I asked her questions and shared personal stories that introduced another perspective. She seemed grateful for my approach. Later that day she texted me when and where our local city council representative held town halls and encouraged me to join her. She said that we may hold opposing political views, but everyone’s voice matters when building community. In the countless times I have engaged in online discourse (regardless of my approach), I never once felt like I ever came close to changing someone’s point-of-view. But with my neighbor, I have hope that I at least gave her a perspective she would not have otherwise heard.

During my education, I was mostly taught how to communicate through formal and official channels. I was taught how to write, how to organize my ideas, how to make presentations, etc. However, no one taught me how to build relationships with someone who is unlike me—someone who thinks differently or may hold hostile opinions. These were skills I had to build on my own through trial and error. I wonder how things would be different if we did teach these skills. To be clear, no one should feel obligated to engage those with hostile opinions in all situations. Physical and psychological safety come first! But I do think we are missing opportunities for change at the micro-scale by always focusing on the macro.

Not All Bad

To be clear: I am not a Luddite. Social media has absolutely been a tool for good in many occasions, like giving voice to the oppressed. I follow content creators that make me laugh, those who make me think and those who inspire me to create. But, I have become increasingly careful and aware of how and when I engage, especially with the rise of AI generated content and AI fueled data collection.

1. De, Debasmita, Mazen El Jamal, Eda Aydemir, and Anika Khera. “Social Media Algorithms and Teen Addiction: Neurophysiological Impact and Ethical Considerations.” Cureus, January 8, 2025. https://doi.org/10.7759/cureus.77145.
   ↩︎
2. “Majority of Neurodivergent Employees Experiencing Mental Health Issues, Study Finds.” People Management, September 6, 2022. https://www.peoplemanagement.co.uk/article/1798074/majority-neurodivergent-employees-experiencing-mental-health-issues-study-finds.
   ↩︎
3. Marecki, Łukasz. “Impact of Work-Life Balance on Employee Productivity and Well-Being.” Journal of Management and Financial Sciences, no. 50 (July 2, 2024). https://doi.org/10.33119/jmfs.2023.50.9.
   ↩︎
4. Craine, Kelly. “Social Media’s Double-Edged Sword: Study Links Both Active and Passive Use to Rising Loneliness | Media and Public Relations | Baylor University.” Baylor University, February 6, 2025. https://news.web.baylor.edu/news/story/2025/social-medias-double-edged-sword-study-links-both-active-and-passive-use-rising.
   ↩︎

C. Smith, Bard College MBA in Sustainability

As both a graduate student and a full-time working professional, I have watched enterprise AI adoption accelerate over the past two years with a growing unease I struggled to name. This discomfort wasn’t just concern about job displacement, though that conversation is important, nor was it solely anxiety about model accuracy or hallucination in workplace output. Something quieter, more below-the-surface and more systemically pervasive, was driving discomfort: the realization that millions of people were being handed powerful AI tools with no structural guidance on how to use them well, and that this gap was costing organizations, workers, and the environment in ways that almost no one was measuring.

That realization became foundational to my thinking around the locus of control in an AI-First Workplace and its deployment: that prompt literacy, the ability to communicate clearly and efficiently with AI systems, is not just a technical skill of the modern worker.
Prompt literacy is a people operations imperative. The locus of responsibility for building that literacy at scale is not the AI platform provider, the IT department, or the individual employee, that belongs to HR.

The Invisible Environment Variable

Most conversations about AI and sustainability focus on what happens inside the model: energy-intensive model training runs, water-cooled data centers, carbon-heavy infrastructure buildouts. These are all legitimate and pressing concerns. However, there is a variable in the environmental impact equation that receives little or no attention: the structure of the user’s prompt.

AI inference, which is the computational process triggered every time a user submits a query, accounts for 80 to 90 percent of AI’s total energy consumption.¹ The IEA projects global data center electricity demand could nearly double to 945 TWh by 2030.² A single ChatGPT query already consumes up to ten times the energy of a standard Googlesearch.³ The research is clear: when users submit vague, unstructured, or ambiguous prompts, models are forced into costly iterative clarification cycles, generating more tokens, consuming more compute, and drawing more energy per resolution. Structured prompting, by contrast, has been shown to reduce inference overhead by up to 2.39 times.⁴ This is not a marginal technical finding. At the scale of hundreds of millions of
daily AI interactions across enterprise, the aggregate inefficiency of unstructured prompting is a material environmental variable. In addition, unlike data center siting or model architecture decisions, it is one that HR leaders can directly influence right now.

Why This is a People Operations Problem

Organizational behavior is shaped by training, culture, and the defaults people encounter at the point of work, and AI is no different. The habits workers are forming with AI tools today, how they frame a request, how much context they provide, and how many follow-up iterations they expect to need, will only solidify. Retraining millions of workers who have already internalized inefficient prompting behaviors is exponentially more costly than building good habits at the point of onboarding and integration. The window for this training is open now, and it won’t be open forever.

People Operations sits at exactly the right location in the organizational architecture to act on this. HR functions are the primary institutional throughline for AI deployment across the enterprise. Deloitte’s 2025 analysis of AI in talent acquisition confirms that HR is the first mover in AI adoption, and organizations are moving faster than their governance frameworks can accommodate.⁵ This is not a crisis; it is an opportunity. If CHROs and People Operations leaders choose to embed structured prompting standards into AI onboarding, workforce policy, and ongoing learning and development programs, they become the delivery mechanism for one of the most scalable environmental interventions available to organizations today.

The World Economic Forum has articulated this imperative clearly: AI agents should be onboarded with the same rigor applied to human employees, with well-defined roles, safeguards, and structured oversight.⁶ I would extend this logic: if we onboard AI agents with human-employee rigor, we must also train human employees with AI-partnership rigor. That means teaching people how to interact with AI systems clearly, efficiently, and intentionally.

What Human-in-the-Loop Means, Operationally

The phrase “human-in-the-loop” has become a compliance concept, something invoked to satisfy auditors and ethics committees and then, imposed prescriptively across the enterprise. I believe this term should be embraced as an operational standard. In an AI-first workplace, keeping humans meaningfully in the loop requires that humans know what they are doing when they engage AI. An employee who cannot construct a well-formed prompt is not in the loop, they are simply and passively along for the ride.

Structured prompting frameworks such as CIDI (Context, Instructions, Detail, Input) are learnable, teachable, and when embedded into organizational workflows, become adopted as second nature quickly.⁷ The research on choice architecture confirms this: smart defaults shape behavior at scale without removing individual agency.⁸ If organizations build structured prompting into their AI governance policies, and HR functions deliver that training at the point of workforce integration, the environmental and operational benefits compound.

This is what a genuinely human-in-the-loop, AI-first program looks like. It doesn’t look like AI tools handed to employees with a disclaimer, but a workforce equipped to direct AI clearly, evaluate its outputs critically, and iterate efficiently. The prompt is the point of contact between human judgment and machine capability. Treating it carelessly is not just an individual inefficiency, it is an organizational and environmental choice.

The Equity Dimension

There is an equity argument embedded in this thinking that should be named explicitly. Access to prompt literacy should not be a function of which organizations have invested in training budgets. Organizations that have built internal prompt engineering capability produce more efficient, lower-impact AI outputs. Those that have not generated excessive compute overhead through disjointed, iterative exchanges. This creates a two-tier system in which resources determine environmental impact. That asymmetry is neither just nor sustainable at scale. When People Operations leaders embed structured prompting into standard AI onboarding (as a baseline workforce training, not premium professional development) they close that gap. Prompt literacy becomes infrastructure, not advantage.

A Call to the People Operations Community

I am writing this for practitioners and leaders in People Operations because they have both the organizational access and the professional responsibility to act. The sustainability conversation in this field has rightly centered on equitable hiring, inclusive culture, and workforce well-being. Now it is important to expand that frame. How organizations deploy AI is a sustainability question, and how employees engage with AI is also a sustainability question. Ensuring that deployment is thoughtful and governed well is a People Operations question.

The prompt interface is not a neutral input field. It is a design choice with environmental, organizational, and human consequences. HR has the tools, the relationships, and the institutional position to make structured, sustainable AI interaction the organizational default. That work begins in onboarding, continues in policy, and compounds quietly and materially every time a well-trained employee sits down to direct an AI system with clarity and intention.

1. James O’Donnell and Casey Crownhart, “AI Energy Usage and Climate Footprint,” MIT Technology
   Review, May 20, 2025, https://www.technologyreview.com/2025/05/20/1116327/ai-energy-usage-climate-
   footprint-big-tech/. ↩︎
2. International Energy Agency, Energy and AI: Energy Demand from AI (IEA, 2024),
   https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai. ↩︎
3. Lekha Naik, “The Carbon Cost of Your AI Prompts,” CNN, June 22, 2025,
   https://www.cnn.com/2025/06/22/climate/ai-prompt-carbon-emissions-environment-wellness. ↩︎
4. Drishti Shah, “Optimizing Token Efficiency in Prompts,” Portkey.ai, March 21, 2025, https://portkey.ai/blog/optimize-token-efficiency-in-prompts/. ↩︎
5. Deloitte, “AI in Talent Acquisition,” Deloitte Insights, 2025, https://www.deloitte.com/us/en/services/consulting/blogs/human-capital/ai-in-talent-acquisition.html. ↩︎
6. World Economic Forum, “AI Agents Onboarding and Governance,” World Economic Forum, 2025, https://www.weforum.org/stories/2025/12/ai-agents-onboarding-governance/. ↩︎
7. AI Academy, “CIDI Prompting Technique: A Beginner’s Guide to Mastering ChatGPT Prompts,” AI Academy Blog, https://www.ai-academy.com/blog/cidi-prompting-technique-a-beginners-guide-to-mastering-chatgpt-prompts. ↩︎
8. Michael Schrage and David Kiron, “The Great Power Shift: How Intelligent Choice Architectures Rewrite Decision Rights,” MIT Sloan Management Review, January 28, 2025, https://sloanreview.mit.edu/article/the-great-power-shift-how-intelligent-choice-architectures-rewrite-decision-rights/. ↩︎