Can we explain it?
Can the person responsible explain the result and why it fits our work?
Something to take home · HyperChimp
Let AI help.
Keep your team capable.
Capability Debt & the nonprofit
AI can draft the report, sort the spreadsheet, or get an idea moving. Someone on your team still has to understand the result well enough to check it and decide whether to use it. This is about keeping that someone.
A made-up example
The report is ready.
A question from the funder isn’t.
Your nonprofit starts using AI to help draft grant reports. It saves time, especially when deadlines pile up. Over several months, a newer coordinator learns how to produce the reports but gets little practice checking the numbers behind them.
One report says attendance increased by 40 percent. The team reviews the wording and sends it.
Then the funder asks
“That’s a big jump. What do you think drove it?”
The coordinator opens the report but can’t find the answer. They know the program well. They just haven’t spent time with the attendance records, so they can’t say which months were compared or whether the draft counted repeat visits as different people.
Now someone has to go back through the records, work out where the increase came from, and find out whether the claim holds up.
What would help
Before the next report, an experienced colleague walks through the attendance calculation with the coordinator. They check which months are being compared and whether they’re counting visits or individual people.
They keep that calculation with the report, and they add a line to the AI policy: AI can help draft the report, but someone confirms the attendance numbers against the original records before it goes out.
Next time, the coordinator works it out and asks a colleague to review it. AI can still help write the draft, and the coordinator can explain where the numbers came from.
What just happened
Some work can go to the tool.
Some needs to stay with you.
In our Keen framework, capability debt is the gap that grows when AI takes on work while the people responsible lose, or never develop, the skills to understand and check it.
Nobody on that team decided the attendance number wasn’t worth understanding. It never got decided at all. It’s a decision worth making on purpose. Go through the work AI is helping with and ask, for each part, whether someone here still needs to be able to do this without the tool.
The numbers you report to a funder belong on that list. So does anything you’d have to defend to your board, and any judgment call about the people you serve. A first draft, the formatting, a summary of a long email thread: those can go to the tool, and your team is better off for it. The gap opens when the two kinds of work get mixed together and nobody has looked.
If you’d like help sorting your work, that’s a conversation we’re glad to have.
Start a conversation ↗Bring these to your next team conversation
Pick one task where you already use AI.
Can we check it?
Can someone trace an important claim back to a record we trust, and spot it if it’s wrong?
Can we keep going?
If the tool is down or wrong, do we still know how to do the essential part ourselves?
These are conversation questions, not a test. If someone can’t answer one, ask what would help them learn it.
One useful next step
Review one piece of work together.
Pick one thing AI helped write. Ask whoever sent it to explain one important claim, show where it came from, and say what they changed. Give them room to ask questions. Keep a short note of the correction for the next person who does that job.
Give people time to check
Checking takes time, records you trust, and permission to question an answer that sounds right. A deadline with no room for that is where the gap starts.
If this raised a question for your team, come find us at our table or start a conversation ↗. There’s a longer piece on capability debt on our site: read the full article ↗.