What a small nonprofit should automate first, and what to leave alone

What makes a good first task

Good first candidates tend to share four traits:

• It happens often: daily or weekly, not once a year.
• It follows a pattern: the same kinds of inputs, handled in roughly the same way each time.
• It's mostly reading and writing: emails, forms, notes, spreadsheets and documents.
• A mistake is easy to catch and cheap to fix, because a person reviews the result before it matters.

Common starting points

• Sorting a shared inbox: reading each message, labelling it and routing it to the right person, with anything urgent flagged.
• Drafting funder and board reports from data you already collect, with every figure linked to its source.
• Meeting follow-up: turning notes into action items and draft emails that wait for your approval.
• Checking documents for completeness: is the application form signed, are the attachments there, what's missing?
• Summarizing long documents, such as a policy update or a stack of grantee reports, into a short brief with references.

What to leave with people

• Decisions about eligibility, priority or safety. An agent can prepare the file; a person decides.
• Crisis and distress responses. Speed and judgement matter, and so does a human voice.
• Relationships: donors, partners and the people you serve. Drafting a thank-you is fine; the relationship is yours.
• Anything where a mistake would harm someone and might not be noticed in time.

A quick way to find your first task

1. 01For two weeks, have each person jot down the repetitive tasks that eat their time, with a rough number of hours.
2. 02For each task, ask: if an AI drafted this, how quickly could a person check it? And what's the worst that happens if the draft is wrong?
3. 03Pick the task with the most hours, the fastest check and the lowest harm. That's your pilot.
4. 04Run it alongside your normal process for a few weeks before you rely on it, and keep track of how often it's right.

You'll often find the answer isn't the task people expected. The monthly report that everyone dreads may be less time in total than the inbox someone sorts every morning.

Be honest about what isn't worth it

Some tasks look automatable but aren't worth the effort: they're rare, the data is messy, or checking the AI's work would take as long as doing it. Saying no to those is part of doing this well. A good first project saves real hours, builds the team's trust, and shows you what your data needs before you take on anything bigger.

See five example workflows drawn out step by step