From overwhelm to optimization: the simple AI shift most nonprofits miss
Juggling grant reports, team check-ins, and a dozen other priorities, and now you're supposed to be an AI expert too?
If you're feeling overwhelmed by the AI landscape, you are not alone. Every week seems to bring new tools and frameworks that promise to transform how nonprofits and mission-driven orgs work.
It's chaotic.
I spent a month exploring the latest agent platforms. Here's the main thing I learned: I was dramatically underutilizing the simple AI assistants I already had.
Instead of chasing the newest shiny object, the most effective step was to customize the tools I already use. This simple shift eliminated repetitive work and improved the quality of the output I was getting.
The problem with generic AI
Most leaders I talk to are using ChatGPT or Claude in their most basic form. They ask one-off questions, get inconsistent results, and have to re-explain who they are and what they need in every conversation.
The result is frustration. You know these tools could be more helpful, but trying to find better solutions often leads down a rabbit hole of complexity. The problem isn't your approach. It's that you're trying to solve the wrong problem first.
Strategic customization is the answer
The most powerful improvement you can make to your AI workflow isn't adding a new tool. It's customizing your existing assistant with strategic instructions.
This means you stop starting from scratch every time. Instead of explaining your context and expectations in every conversation, you embed them directly into the assistant's instructions. You go from getting generic responses to outputs tailored for your specific org's needs.
A well-crafted set of instructions transforms every interaction. You're not just saving time on an individual question; you're changing the quality of every response you get from that assistant.
A systematic approach to customization
Here's how to implement this, starting from wherever you are now.
Start where you are. If you're using a free version of an AI assistant, go to the settings and find "Custom Instructions." Write a simple prompt that includes your role, the type of org you work for, and your preferred communication style. This basic step will immediately improve your interactions (available in ChatGPT and Claude)
Upgrade for focus. The $20/month for a pro version is worth it, mainly for the ability to create separate, focused assistants (called "Projects" in ChatGPT and Claude and Gems in Gemini). This is where the real value is.
Build focused assistants. Create separate assistants for different functions. You might have one for drafting funder communications, another for strategic planning, and a third for creating internal team updates. Each one gets its own instructions tailored to that specific function. Your funder assistant knows your tone, talking points, and key project details. Your strategy assistant understands your organizational context.
The difference in practice
The difference is clear. A generic request like "help me write an update email" produces corporate-speak that doesn't fit your context.
A customized assistant for your board communications knows your leadership style & writing voice, understands the need to link activities back to your mission, and remembers to include a call for feedback. It gives you a starting point that is 80% of the way there.
Your custom assistant builder
To make this easier, I've created a template to help you build these custom instructions. It asks the right questions to help you define the role, context, and output preferences for any use case.
Whether you're creating an assistant for donor communications or operational workflows, this template provides a systematic way to customize it.
Copy the AI Prompt coach from here (no email required).
Try one focused assistant first.
Pick the task you use AI for most often.
Build it using the template and compare the results to your previous generic interactions.
Your AI workflow doesn't need to be cutting-edge to be powerful. It needs to be systematic, reliable and optimized for the work you actually do. Start there, build systematically, and let the tools serve your processes rather than forcing your work to fit their limitations.
Next month, I'll share what else I'm doing to improve this approach and to connect operational systems. But for now, focus on building your first systematically customized assistant and experiencing the difference firsthand.
I'd love to hear your results. What did you build? What worked, what didn't?
I read and reply to all messages - we're all figuring this out together, and your experience helps others avoid the same pitfalls. 🙂
See you next time,
Malcolm