The Frustration
A few weeks ago, I was working on a research task in ChatGPT to fill missing contact details for a not-for-profit CRM. I’d given the AI a clear instruction… or so I thought.
The result? Half the info was missing, and what was there looked like it had been guessed.
Rather than starting over, I asked ChatGPT to slow down and explain its thinking step-by-step before giving me the final answer. That’s when the results changed completely, it caught errors, used better sources, and even flagged details that needed human checking.
What Is Chain of Thought Reasoning?
Chain of Thought (CoT) is when you get ChatGPT to “think out loud” before giving the final answer. Instead of jumping straight to a conclusion, it walks through the logic, checks assumptions, and only then provides the result.
In plain terms: CoT is like asking your AI to show its working, so you can check the logic before you act on it.
Real-World Examples
Workshop planning:
While designing a volunteer teamwork workshop, my first draft agenda was too packed and out of order. Instead of simply asking ChatGPT to “fix it,” I told it: “Review the themes, decide which are most important for this audience, reorder them, and explain your reasoning before drafting the final outline.” The step-by-step reasoning revealed that one theme could be folded into the introduction, freeing more time for deeper discussion where it mattered most. The final agenda was not just shorter, it was sharper and better aligned to the group’s needs.
Decision support (tool comparison):
When weighing two AI workflow tools for a project, I could have just asked, “Which one’s better?” Instead, I told ChatGPT: “List the pros and cons of each first, then recommend one based on those factors.”
The reasoning process surfaced trade-offs I hadn’t considered, including integration limits that would have caused headaches down the track. The final recommendation was stronger because it was backed by a clear decision trail.
Why This Matters for NFPs
Not-for-profits work with sensitive data, limited resources, and high accountability. If an AI tool gives a wrong or incomplete answer, the impact can be real, from misinformed funding applications to wasted staff time.
CoT matters because it:
- Reduces risk by making the AI’s logic transparent.
- Improves accuracy by catching errors before they cause problems.
- Builds trust so your board, team, or funders can see how a conclusion was reached.
How to Use CoT in Your Prompts
Try adding instructions like:
- “Explain your reasoning step-by-step before giving the final answer.”
- “List the assumptions you are making before you give your answer.”
- “Work through the pros and cons first, then give your recommendation.”
- “Show your working, then summarise the result in plain language.”
Call to Action
Next time you use ChatGPT, try asking it to think out loud first. You might be surprised at what it catches and how much better your outcomes are.
Add comment
Comments