A practical AI roadmap for Ottawa non-profits

You are not trying to "do more AI." You are trying to get more impact out of the same staff and budget, without breaking the trust you have with funders, members, and the people you serve. A useful AI roadmap is just a clear, modest plan for that.

This is where most non-profit AI efforts in Ottawa go sideways. Someone gets a grant, buys a flashy tool, runs a one-off pilot, then 6 months later nobody is using it and everyone is more skeptical than before. The problem is almost always the lack of a simple, realistic roadmap that fits how your organization actually works.

What an AI roadmap for a non-profit actually looks like

Forget the 40-page "digital transformation" slide deck. For a local non-profit, a good plan fits on 2 or 3 pages and answers four questions:

• Which specific pains are we tackling first (and which are we ignoring for now)?
• What AI tools or workflows will we test for those pains, in what order?
• How will we run each test, measure it, and decide whether to keep it?
• What data, policies, and skills do we need to support all this without creating new headaches?

If your plan does that in plain language, you are ahead of most organizations, including many bigger ones downtown.

For Ottawa non-profits, I usually see the same 5 "pains" show up early in an AI roadmap:

• Grant writing and reporting cycles swallowing staff time
• Repetitive email and phone responses to the same questions
• Messy membership or donor data spread across spreadsheets and tools
• Manual intake and eligibility screening
• Basic communications: newsletters, social posts, web copy updates

Notice what is not on that list: "build our own chatbot from scratch" or "fully automate fundraising." You start with bite-sized, human-in-the-loop workflows you can test in weeks, not strategic moonshots that need a huge budget.

Step 1: pick 2 or 3 high-friction workflows

Start by drawing a line between two categories of work in your organization:

• High-friction, low-judgment work: repetitive writing, copy/paste tasks, boilerplate reports, manual sorting
• High-judgment, high-trust work: case work, clinical decisions, conflict resolution, board strategy, political advocacy

Your first AI roadmap lives almost entirely in the first category. The second category might use AI for background research or drafting, but decisions stay fully with staff.

A practical way to do this is to sit down with your team (or at least program leads) and ask a very specific question: "What do you do in a typical month that is boring, repetitive, and would not upset you if a robot helper did half of it?" Collect 10 or 15 answers, then rank them using three quick scores from 1 to 5:

• Time: how much staff time this eats in an average month
• Annoyance: how painful it feels to do
• Risk: how bad it would be if an AI draft was wrong and a staff member had to fix it (lower is better)

Add Time + Annoyance, then subtract Risk. The top 2 or 3 items become your first AI experiments. Typical winners in Ottawa non-profits are