← Back to “AI for Good” Nonprofit Workshop Resources
← Back to “AI for Good” Nonprofit Workshop Resources
<aside> <img src="notion://custom_emoji/4911c20d-2ec8-8108-968e-0003c93a53bb/3071c20d-2ec8-80c9-9383-007a1c296af6" alt="notion://custom_emoji/4911c20d-2ec8-8108-968e-0003c93a53bb/3071c20d-2ec8-80c9-9383-007a1c296af6" width="40px" />
Chiv Heng Consulting — www.chivheng.consulting • LinkedIn
</aside>
Last reviewed: July 2026
This page is a short, current snapshot of what is changing in AI. It focuses on the big, durable shifts, not model version numbers (those change weekly, and fixating on them is a trap).
AI has moved from answering questions to doing multi-step work: browsing the web, working across your files, and returning finished drafts. When people say "AI agents," this is what they mean. You have probably already felt it in your own tools, even if you have not called it that.
The leading tools (Claude, ChatGPT, Gemini) ship new versions constantly, and each briefly claims the top spot. The skill that transfers is prompting well and verifying output, not memorizing which model is "best" this month.
Tools like NotebookLM can now find their own sources and run "deep research," not just read what you upload. Powerful, but it means you verify more, not less.
Error rates have dropped sharply, but made-up citations and confident wrong answers still happen, especially with numbers and sources. Verify anything that informs a decision.
Expect growing expectations to disclose AI use, label AI-generated content, and handle data responsibly. You do not need to track every law. Assume transparency and provenance are becoming defaults.
Tools like Claude's Cowork let non-technical users hand off multi-step tasks (organize a folder, draft from your files, run a weekly report). It is included with paid Claude plans (Pro and up), so for nonprofits the main thing to weigh is cost.