← Back to “AI for Good” Nonprofit Workshop Resources
<aside> <img src="notion://custom_emoji/4911c20d-2ec8-8108-968e-0003c93a53bb/3071c20d-2ec8-80bd-b78f-007a2824c29c" alt="notion://custom_emoji/4911c20d-2ec8-8108-968e-0003c93a53bb/3071c20d-2ec8-80bd-b78f-007a2824c29c" width="40px" />
Chiv Heng Consulting — www.chivheng.consulting • LinkedIn
</aside>
Last reviewed: July 2026
Effective use of generative AI is not about focusing on any specific tool. The most critical aspect is developing your skills in prompting the AI tool and reviewing and editing what it produces for quality and accuracy.
Work with AI as if you are managing a smart, eager intern on their first day. The full metaphor lives on the Prompt Engineering (Beginner Level) page.
The Prompts below are a STARTING POINT to improving your output on delegated tasks.
Define done before you delegate: Name the decision the output supports, its audience, the required format and length, the facts or columns it must use, and what would make you reject it. This is what good looks like. Add these details to the prompt, then check the result against your acceptance criteria.
SAFETY FIRST: Never include real names, addresses, SSNs, or sensitive donor info in your prompts. Use placeholders like [Donor Name] or anonymize your data before uploading it.
Data Privacy Rules + Anonymization Examples (Workshop Handout)
Attach or paste your anonymized data first, then add the prompt. The sample datasets on the Example Use Cases page work for practice.
Original: "Show me where we're over/under budget by more than 10%. Rank categories by size of variance."
Refined: "Act as a Financial Auditor. Review this budget data and identify all line items with a variance greater than 10% (positive or negative). Present these in a table ranked by the absolute dollar amount of the variance. Add a column with a brief 1-sentence risk assessment for each."
Why it works: Ranking by "absolute dollar amount" ensures you see large over-spends and under-spends at the top, rather than hiding large negative numbers at the bottom. The 10% threshold, ranked table, and one-sentence risk assessment are acceptance criteria for the finished output.