7. The Small Charity AI Playbook: Starting from Zero with No Budget

Whilst we've developed lived experience embedding AI in a small charity. We'll start with an honest disclosure. We are an AI consultancy. We sell an AI toolkit. It would be very convenient for me to tell you that getting started with AI requires professional support.

It doesn't.

If your charity has no budget for AI, no dedicated tech person, and no idea where to begin, everything in this post is free. You can implement all of it this week with nothing more than a smartphone or laptop and an internet connection. No subscriptions. No courses. No consultants.

This post is the playbook for starting from zero.

What you need before you open any AI tool

Two things, both of which take less than an hour.

First: your one-page AI policy. We covered this in Episode 1, but here's the short version. Write a paragraph that names the tools staff can use, states that no personal or beneficiary data goes into them, requires human review of all AI outputs, and names a person to report concerns to. Print it. Email it. Pin it above the kettle. This transforms AI from "something people do secretly" into "something the organisation supports."

Second: your first-task decision. Pick one task. Not the most complex or ambitious one — the one that eats the most time every week. For most small charities, this is one of three things: meeting notes, first-draft communications (newsletters, social media, emails), or report formatting. That's your starting point.

The free tools: what's actually available

There are four generative AI tools worth knowing about. All have free tiers. You don't need all four — pick one and learn it well.

ChatGPT (chat.openai.com) — The most widely used AI tool in the charity sector, with 55% of charities who use AI choosing it. The free tier gives you access to GPT-4o with usage limits. It handles conversation, drafting, summarising, brainstorming, and basic analysis well. You can use it without creating an account via the mobile app, though an account gives you conversation history. The free tier uses your inputs to improve its models — so remember, no personal data.

Microsoft Copilot (copilot.microsoft.com) — If your charity uses Microsoft 365 (many do through the nonprofit programme), Copilot Chat is included free with your subscription. It searches the web in real time, which ChatGPT's free tier doesn't always do, making it useful for research tasks and checking current information. It's also integrated into Bing, so you can use it directly from your browser.

Google Gemini (gemini.google.com) — Google's AI tool. If your charity uses Google Workspace (available free to nonprofits through Google for Nonprofits), Gemini integrates with Docs, Sheets, and Gmail. The free tier is capable and, like Copilot, has web access built in. Google's nonprofit programme includes enterprise-grade data protections for Workspace users.

Claude (claude.ai) — Made by Anthropic. Strong at longer documents, detailed analysis, and nuanced writing. The free tier has usage limits but is generous enough for daily use. Particularly good for working with long texts like policy documents, reports, or grant applications. Anthropic offers nonprofit discounts on paid plans — up to 75% off Team plans through Goodstack verification.

For a charity starting from zero, my recommendation: start with whichever one you already have easiest access to. If you have Microsoft 365, try Copilot. If you use Google Workspace, try Gemini. If neither, ChatGPT is the most documented and has the largest community of charity users sharing tips.

Week 1: The first three tasks

Don't try to transform your organisation. Try three things.

Task 1: Summarise something. Take the longest document you've read recently — a policy update, a funder's strategy document, a government consultation — and paste it into your AI tool with the instruction: "Summarise this in five bullet points, each under 20 words, for a charity CEO with limited time." Time yourself. Compare the result to what you'd have produced manually.

This task is useful because it demonstrates the core capability (AI processes text faster than you can) with zero risk (public documents, no personal data, no external use).

Task 2: Draft something. Take a communication task that's on your to-do list — a volunteer newsletter, a social media post, an email to a partner organisation. Give the AI tool the context: who it's for, what tone you want, what information to include, and how long it should be. Review the draft. Edit it into your voice. Send it.

This demonstrates the drafting capability that saves the most time day-to-day. The output won't be perfect — it never is — but it gives you something to react to, which is faster than starting from blank.

Task 3: Improve something. Take a piece of your own writing — a grant report section, a board paper, a case for support — and ask the AI tool: "Review this text. Flag any jargon, suggest simpler alternatives for complex sentences, and check whether the argument flows logically." Read the suggestions. Apply the ones you agree with.

This demonstrates AI as an editor, which many people find less threatening than AI as a writer. You're still the author. The tool is making your work clearer.

Week 2: Build your starter prompt library

After a week of experimentation, you'll have a sense of which tasks AI handles well and which prompts produce useful results. Now capture that.

Create a shared document — Google Doc, Word file, whatever your team uses. Title it "AI Prompts That Work." Add the prompts from your first week's experiments, with a one-line note on what each one does and any tips for getting a good result.

This doesn't feel like a significant step, but it's the single most important structural move you'll make. A shared prompt library means knowledge isn't locked inside one person's head. When a colleague needs to write a newsletter, they don't start from scratch — they find the prompt that worked last time. When a new staff member joins, they have a starting point. When someone discovers a better approach, they update the library.

Over time, this document becomes your organisation's AI playbook. It's more valuable than any training course because it's specific to your workflows, your tone of voice, and your reporting requirements.

Week 3: Expand to a second person

If you started as a solo experiment, now involve one more person. Share your prompt library with them. Ask them to try AI on a task from their own workflow — ideally something different from what you've been doing, so you're testing breadth rather than depth.

Debrief together. What worked? What didn't? What prompts need adjusting? Add their discoveries to the shared library.

This is how organisational adoption starts — not through a top-down mandate or a training programme, but through one person showing another person something useful.

Week 4: Review and decide

After a month of experimentation, you have enough evidence to make an informed decision about next steps. Sit down — with your line manager, your CEO, or your board if appropriate — and answer four questions:

Where did AI save time? Quantify it if you can, even roughly. "I spent two hours writing the newsletter last month. This month it took forty-five minutes including AI drafting and my editing."

Where was the quality good enough? Not perfect — good enough to be useful after editing.

Where did it fall short? What tasks produced outputs you couldn't use?

What do we want to try next? Pick one or two new areas to explore in month two.

This review is what turns experimentation into strategy. It gives you evidence for your board (Episode 3), data for a potential business case (coming in Episode 7), and confidence that AI is worth continuing with.

When free stops being enough

At some point — maybe in month two, maybe in month six — you'll hit the limits of free tools. Usage caps will interrupt your workflow. You'll want features that paid tiers offer: longer conversations, document uploads, image generation, or enterprise data protections.

When that happens it's worth knowing about the nonprofit discounts available (these are available at the time of writing):

OpenAI offers 20% off ChatGPT Team plans for verified nonprofits, with deeper discounts for larger organisations through direct sales.

Anthropic offers up to 75% off Claude Team and Enterprise plans for nonprofits verified through Goodstack.

Microsoft's nonprofit programme provides heavily discounted Microsoft 365 licences (up to 75% off), and Copilot can be added for around £20 per user per month with a further 15% available through TechSoup.

Google Workspace is free for eligible nonprofits through Google for Nonprofits, and Google Cloud provides approximately £780 in annual credits that cover Gemini and Vertex AI usage.

Verification for most of these discounts runs through TechSoup (techsoup.org) or Goodstack (goodstack.io). If you haven't registered with TechSoup already, do it — it's the gateway to nonprofit pricing across most major technology platforms.

What this playbook doesn't cover

It doesn't cover data analysis, workflow automation, or AI-assisted service delivery. Those are real and valuable applications, but they require more infrastructure than "start from zero" allows. They're where structured guidance — whether from a consultant, a programme like CAST's, or a toolkit — adds genuine value.

It also doesn't cover governance in depth. If you've followed this series, you've already read the AI policy guide (Episode 1), the trustee briefing (Episode 3), and the GDPR guide (Episode 4). If you haven't, go back to those before expanding your AI use beyond basic experimentation.

And it doesn't replace strategic thinking. AI is a tool, not a strategy. The question isn't "how do we use AI?" — it's "what are the biggest drains on our time, and can AI help with any of them?" Start with the problem, not the technology.

The honest pitch

Everything in this post is free. You can implement all of it without spending a penny, and many charities will get substantial value from doing exactly that.

When you're ready to move beyond experimentation — when you want a structured readiness assessment, a comprehensive use case discovery process, a prompt library with 50+ charity-specific prompts, and a 12-week implementation roadmap — that's what the Third Sector AI Toolkit provides. It costs £400 for all six modules. But it's there when you need it, not before.

Start here. Start free. Start this week.