DX

In the digital age, AI has become a familiar tool for businesses in planning, customer care, market research, and content creation.However, behind every AI command is a data center that consumes electricity, water, and emits CO₂. Without mindful usage, the environmental cost can quickly exceed expectations.

KisStartup – with experience supporting thousands of businesses on their innovation and digital transformation journeys – has compiled this guide to help companies use AI responsibly, efficiently, and in an environmentally friendly way.We call it Green Standard Prompting: boosting productivity while reducing emissions.

Why do we need "Green Standard Prompting"?
Every AI command consumes energy and water:
- Gemini (Google): approx. 0.24 Wh, emits 0.03 gCO₂, and uses 0.26 ml of water per average text prompt.
- ChatGPT (GPT-4o): approx. 0.3 Wh per prompt.

1 million prompts can consume around 300 KWh, equal to a household’s electricity use in one month.
So, every time you revise a prompt repeatedly, you're multiplying the power and water usage. That’s why carefully crafting your prompt is not only time- and cost-effective, but also a clear ESG (Environmental, Social, Governance) action.
Principles of Green Standard Prompting:

1. Be clear about your goal: what you want, for whom, and in what format.
2. Provide enough context: product, data, constraints.
3. Limit output length: specify word count or number of bullets.
4. Choose the right model: simple tasks → lightweight models.
5. Ask AI to request more info if needed, instead of guessing.​
6. Save and reuse good prompts to avoid repetition.

Example Green Standard Prompts
1. Planning Content Marketing
System (Role): You are a Sustainable Content Marketing expert.
User Prompt:

• Goal: Plan 2 weeks of content for a {industry} fanpage.
• Audience: {target customers}
• Context: product {...}, USP {...}, budget {...}
• Output (≤200 words):
  - 14-day content calendar
  - Captions ≤30 words
  - Hashtags ≤5 per post​
• Constraints: prioritize repurposing existing content, ask up to 3 follow-up questions if data is missing.

2. Developing a Green Export Plan
System (Role): You are a Green Export & ESG expert.
User Prompt:

• Goal: Create a 6-month export plan for {product} to {market}.
• Context: certifications, production capacity, current partners
• Output (≤250 words):
  1. 5 green requirements/VSS (Voluntary Sustainability Standards) for the marke.
  2. 3 current capability gaps
  3. 3 priority actions for the first 90 days
  4. 2 long-term opportunities​
• Constraints: include a checklist for executives, ask up to 5 follow-up questions if data is missing, cite sources.

​Steps to Build a “Green Standard” AI Assistant for Your Business

1. Define the assistant's role (e.g., Content Coach, Export Advisor).
2. Standardize the system prompt (role + green principles).
3. Create a sample prompt library (like examples above).
4. Train with real data (products, certifications, customer info).
5. Test & refine to minimize prompt iterations.
6. Integrate into workflows (chatbot, CRM, Notion/Slack).
7. Monitor & report on resource savings (tokens, kWh, CO₂, water).

Checklist Green AI Prompting
Before you type a command:

• Is the goal, target audience, and output format clear?
• Is the data sufficient so AI doesn’t have to guess?
• Have you set an output length limit?

When choosing a model:

• Do you really need a large model?
• Are you asking for images/slides unnecessarily when text suffices?

During execution:

• Does the AI ask follow-up questions when data is missing?
• Can this prompt be reused?

After completion:

• Is the output immediately usable, or does it require re-running?
• ​Can the prompt be shared with teammates?

• For an average text command:
  • Gemini (Google): approximately 0.24 Wh of electricity, emits 0.03 gCO₂, uses 0.26 ml of water.
  • ChatGPT (GPT-4o): estimated at around 0.3 Wh of electricity.
• These numbers may seem small, but for 1 million commands → approximately 300 kWh, which is the electricity consumption of a household in one month.
• Additionally, each 0.3 Wh of electricity could be equivalent to 0.03–0.21 gCO₂ depending on the "cleanliness" of the energy source.

• Vague prompt → AI gives incorrect responses → need to run again.
• No length limit → AI generates unnecessarily long text → consumes tokens, uses more electricity.
• Choosing an overly powerful model for a simple task → like using a truck to carry a bag of vegetables.

1. Clear goal: Specify exactly what you need, for whom, and in what format.
2. Provide sufficient context: Give data, conditions, and constraints upfront.
3. Limit output: Request specific word count or number of bullet points.
4. Choose the right model: Simple tasks → small models. Complex tasks → large models.
5. Avoid multimedia waste: Only ask for images/slides when absolutely necessary.
6. Save good prompts: Reuse them, don’t "reinvent the wheel."