How to Build a Custom AI Chatbot for Your Nonprofit (And Why You'd Want To)

There's a question I hear from nonprofit teams across Canada: "We know AI is a thing. But can it actually help us, or is it just hype?"

The honest answer? It depends entirely on how you use it. If you hand your work over to an AI tool and copy-paste whatever it gives you, that's not going to serve your mission. But if you stay in charge, if you use AI as a thinking partner that helps you be more thorough, more strategic, and more clear, it can genuinely change how your team works.

I want to show you one specific, practical way to do that: building a custom AI chatbot for your organization.

In this article, you'll learn:

• What a custom nonprofit AI chatbot is and how it differs from regular ChatGPT

• The four ingredients you need to build one in about 10 minutes

• Three real use cases, including a grant writing chatbot

• The safety and privacy basics to get right before you start

  
  

What is a custom AI chatbot for nonprofits?

You've probably used ChatGPT or one of the other AI tools to ask a question or draft an email. That's great. But a custom chatbot takes it a step further. Think of it like having an assistant who already knows your organization. It knows your mission, your programs, your tone of voice, your templates. You don't have to re-explain everything every time you start a conversation.

You can build one in any of the major tools. ChatGPT calls them Custom GPTs, Google Gemini calls them Gems, and Claude calls them Projects. The interface looks a little different in each one, but the thinking is exactly the same.

The four ingredients: how to build a nonprofit AI chatbot

Every custom chatbot, regardless of which tool you build it in, comes down to four things.

Purpose. What is this chatbot for? Who's going to use it? What should it do, and just as importantly, what should it not do?

Instructions. This is the system prompt. It's where you tell the AI how to behave, what tone to use, what rules to follow. Think of it as the job description for your assistant.

Knowledge. These are the documents you upload so the chatbot actually knows your world. Your grant guidelines, your strategic plan, your reporting templates, past successful applications, whatever it needs to be useful.

Testing. Try it. Break it. Ask it the questions your team would ask. See where it stumbles and refine from there. Nobody gets this perfect on the first try, and that's completely fine.

A quick walkthrough: building your chatbot in 10 minutes

No matter which tool you choose, the steps are the same:

1. Open your tool (Claude Projects, ChatGPT Custom GPTs, or Google Gemini Gems)

2. Create a new project, GPT, or Gem

3. Write your purpose in one or two sentences: who is this for and what does it do?

4. Write your instructions: tone, rules, what to do and what not to do

5. Upload 3 to 5 documents that give it the knowledge it needs

6. Test it with a real question your team would actually ask

7. Read the response carefully. Where did it get it right? Where did it miss?

8. Refine your instructions based on what you learned, and test again

That's it. The rest is iteration, and iteration is where the magic happens.

Three nonprofit use cases for custom AI chatbots

Here's where it gets exciting. Let me walk you through three examples.

A grant writing chatbot. Upload your org's mission statement, past successful applications, and the grant guidelines. Then ask it: is this grant a good fit for us? What should we focus on? Draft me an opening paragraph. It doesn't replace your grant writer. It gives them a running start. And honestly? For a lot of small nonprofits, the grant writer is the ED at 11pm. This helps.

I recently walked a room of about 25 nonprofit professionals through exactly this at a Non-Profit Technology Night at theDock in Victoria. I built a chatbot that had the City of Victoria's My Great Neighbourhood Grant guidelines, past winning examples, the application form, and all the instructions uploaded. Then I described a real community event, a street party in Fernwood Square, and asked the chatbot to tell me if it was a good fit, what to watch out for, and what made past applications strong. In seconds, I had guidance that would have taken me hours to piece together by reading through all those documents myself.

And that's the thing. I don't want to spend seven hours reading grant guidelines when I'm volunteering my time to do something good for my community. I want to maximize the time I have for the work that actually matters.

Annual report and program reporting. Upload your data, your outcomes framework, your funder templates. Ask it to draft impact narratives, summarize program outcomes, format things for different audiences. This can turn a weeks-long process into a days-long process.

Board meeting prep. Upload your past minutes, financials, and strategic plan. Ask it to draft agenda summaries, prep briefing notes, and flag items that need decisions. Your ED shows up to the board meeting more prepared and less stressed.

Staying in charge: why the human leads

Here's what I always tell my clients: you are the human, and you are in charge. The AI is working for you. You're not giving your responsibility over to it.

When I use a chatbot to help draft something, I'm never going to just copy-paste what it gives me and call it done. I want it to help me get my ideas out. I love writing, but I don't love a blank page. So I'll tell the chatbot all about my project, my ideas, my concerns. Sometimes I'll use speech-to-text and just talk it out. Then I'll ask it to give me a draft, and from there I polish it. I put it in language I can stand behind, because anything I submit with my name on it needs to represent what I actually mean. (I've written more about what it really takes to write with AI, the ethics, the edits, and the human discernment that goes into it.)

And if you're representing a nonprofit, that matters even more. Your organization has values, a voice, a way it wants to show up on the page. You don't want a grant application that sounds like a chatbot wrote it. You want one that sounds like your organization wrote it, just faster and maybe more clearly than you could have managed at 11pm on a deadline night.

The chatbot can also help you think. Ask it to find weaknesses in what you've written. Ask it to compare your draft against the grant guidelines and pick it apart. Ask it to give you better questions to answer. It can be a genuinely fun, collaborative experience. It can actually add to your creative process, not take away from it.

Before you start: AI safety and privacy for nonprofits

A quick but important note. Before your team starts building chatbots or using any AI tool, please think about what data you're putting into it. Never enter personal information, client data, or anything confidential into a consumer AI tool without understanding where that data goes.

I've written specifically about whether ChatGPT is confidential for nonprofits in Canada, a plain-language guide to what "confidential" really means and what should never be entered.

And if your organization doesn't yet have a policy around AI use, I'd strongly recommend starting there. Here's my board-ready AI policy checklist and a deeper look at why nonprofits need an AI policy in the first place.

These tools are powerful. Using them mindfully and with intention is what makes the difference between a risk and an asset.

Try it yourself

Here's my slide deck from the theDock talk so you can follow along with the full walkthrough: How to Build a Custom AI Chatbot in 10 Minutes, Slide Deck

And here's the custom instruction template I used to build the grant writing chatbot in my demo. You can copy this into whichever tool you choose, whether that's Claude Projects, ChatGPT Custom GPTs, or Google Gemini Gems, and start building your own: Grant Writing Assistant, Custom Instruction Template

If your nonprofit is thinking about adopting AI more broadly and wants help with policy, governance, or safe rollout, that's the work I do with teams.

The real takeaway

The tools will keep changing. Six months from now there might be a new platform, a new interface, a new name for the same idea. That's fine. What won't change is the thinking behind it: purpose, instructions, knowledge, testing. If you understand those four ingredients, you can walk into any tool that comes along and know exactly what to do.

But here's what matters most. You stay in charge. AI can be used to to help you do better work. To be more thorough. To share your excellent ideas more clearly. To think more strategically about the things that matter to your mission. It can help you see angles you might have missed and ask questions you hadn't thought to ask. It can model structured, strategic thinking in a way that strengthens how you think, not just what you produce.

Your voice. Your values. Your direction. AI can help you carry all of that further, with more clarity and more confidence, than you could alone. And that, to me, is worth learning.

Frequently asked questions

Is it safe to upload grant documents to an AI tool?

Grant guidelines and application forms from public funders are generally low-risk since they're publicly available. Avoid uploading anything with personal information, client data, or confidential financials. When in doubt, check my guide on ChatGPT confidentiality for Canadian nonprofits.

Do we need an AI policy before we start?

Ideally, yes. Even a simple one-page set of guidelines gives your team clarity on what's okay to use AI for and what isn't. Here's a board-ready checklist to help you get there quickly.

How long does it actually take to set up?

About 10 minutes for a first version. The real time goes into testing and refining, which you can do gradually over a few days as you use it.

Which tool should we pick?

The one your team already uses. If you're on Google Workspace, try Gemini Gems. If you already pay for ChatGPT, try a Custom GPT. If you want the strongest writing quality for grants and reports, try Claude Projects. The thinking is the same across all three.

---

About Sarah Downey

Sarah Downey is a Canada-based consultant helping nonprofits adopt AI safely, ethically, and confidently through governance clarity and policy development.