Automate the work you don't want to do first

4:28

What should a nonprofit automate first with AI?

Start with the work you keep saying you don't want to do, as long as it still needs to get done. Dread is a reliable signal. It usually points at repetitive, rules-based work that eats hours and doesn't need your judgment, which is exactly what AI handles well.

The phrase came from Colleen. She said "I don't want to do that" about a task one day, and I wanted to put it on a sticker. It could sound petulant. What we mean is that a lot of work in 2026 no longer needs a human being to do it.

AI can do a lot of this work and so if you have a thought that says I don't want to do that that's a good place to start for automating some artificial intelligence sequences

Scott Williams · Recorded conversation, Sep 1, 2026

If you want a menu of common starting points, our 5 workflows every nonprofit should automate first is a good list. This article is about choosing from your own week, using the work your own team already dreads.

What is an AI employee?

An AI employee is an AI setup that does a defined job on your behalf, on a schedule or a trigger, without you sitting there. It has instructions, access to the tools it needs and a way to report back. You review its work the way you'd review a staff member's.

The difference from a chat window matters. I describe chat as a hammer. Your hand has to be on it the whole time, and the moment you set it down, the work stops. An employee keeps working when you leave the building. That's the version of AI that actually gives a small team its time back.

How do you find the right first task to automate?

Ask each person on your team one question and write down their first answer. If anything were possible, what's the work you would no longer be doing that still has to get done? People answer fast, because they already know.

Colleen asked this of a staff member at an If Possible customer, a ministry with about a million-dollar budget. Big vision, small team.

if anything is possible, what's the work that you would no longer be doing? But still has to get done.

Colleen Cook · Recorded conversation, Sep 3, 2026

Her week went to three things. Promoting events on social media, answering customer service email about information products, and managing event registrations so everyone landed in the right group. She answered almost instantly. So we built a system to do that work her way. Now she can make a month of social content in about an hour, and people are complimenting her on it.

What did we automate first at If Possible?

One of the first AI employees we built reads our recorded meetings every night and pulls out every promise we made. It started small, with one job and one purpose, and it's still one of the most useful things we run.

We record almost every conversation, so the transcripts already exist. It hit me that I didn't need to write anything down. I could build an AI employee that goes back through every recording on a nightly basis and looks for one thing only.

Colleen took the same idea further with her AI executive assistant. It reads her calendar, her messages, her meeting recordings and her task list. At the end of the day it sends her a Slack message about what got done, what slipped into tomorrow and which new promises still need a home. In September it caught a commitment she'd made out loud to build slides for a workshop.

I didn't write it down anywhere. I went on to the next thing and it was like, hey, just a reminder, there's no home for that. I was like, thank you. You just did your job, AI employee.

Colleen Cook · Recorded conversation, Sep 10, 2026

How do you train an AI employee until it runs on its own?

Train it the way you'd train a new person. Do the work alongside it at first, correct what it gets wrong, then check in less and less often. When the same corrections stop coming up, you can leave it to do the job and review the results.

I picture someone sitting in the chair next to me. At first we work through every step together. Then I back away slowly and keep checking in.

Plenty of teams quit in the frustrating middle where the AI is mostly right and you're still correcting it. That middle is the training. Every correction you write down becomes an instruction it follows next time.

How do you pick your first automation this week?

Spend one week logging every moment you think "I don't want to do that." Then look at each item and pick the one that's repeated, rules-based and safe to get slightly wrong while the AI learns. That's your first AI employee.

Use a simple log like this. The example rows come from the real work described above.

Then follow these steps:

1. Pick one row. Choose the task you'd most love to stop doing that is also low risk.
2. Write the job description. What it reads, what it produces, what good looks like, who reviews it.
3. Work beside it for a week. Correct every miss in writing.
4. Step back. Check daily, then weekly.
5. Leave the room. Keep reviewing results, and move to the next row.

Treat this like hiring, because it is. Treat AI like a new hire covers the onboarding mindset, and the donor thank-you system is a good example of a first AI employee aimed at fundraising. Find more in our implementation articles.

For one practical idea each week, join the If Possible newsletter. If you'd like to build your first AI employees with us and other nonprofit teams, get on the cohort waitlist.