Team practice and accountability
Safe use by one careful person is fragile. It breaks the moment work is handed off, reviewed, or published by someone else. This module is about the shared habits that make good practice stick across a team: being honest about AI use, reviewing each other's work, protecting trust and rights, and raising problems without fear.
By the end, you will be able to
- Disclose and record AI use in a way that fits the audience and keeps trust.
- Review another person's AI-supported work against a shared standard.
- Protect authenticity, rights, and donor trust in anything you publish.
- Raise incidents, near misses, and quiet workarounds without blame.
Lesson 1
Disclose and document appropriately
Being honest about how AI was used protects trust. What you disclose depends on who is reading.
Honesty about AI use is not one rule for every situation. A quick internal draft may need nothing said. A public message, a fundraising appeal, or anything a person will rely on may need it made clear. The test is what your audience would reasonably expect to know.
- Audience: what would this reader want to know about how it was made?
- Provenance: keep a simple note of what AI helped with and what a person did.
- Approvals: record who reviewed and approved before it went out.
- Records: keep enough of a trail that you could explain the work later.
Note
You are not confessing to a crime. You are keeping the kind of honest record that lets your organization stand behind its own work if anyone ever asks.
Try it
Think of one thing you might publish with AI help. Would your reader expect to be told? What short note would you keep about how it was made?
RememberMatch disclosure to what the audience expects, and keep a simple record of AI use and approval.
Lesson 2
Review another person's AI-supported work
Reviewing a colleague's AI-assisted work is a real skill. You are checking more than whether it reads well.
When you review someone else's AI-supported work, a quick read is not enough. Fluent text hides its errors, so you have to check the things a polished draft can quietly get wrong.
- Evidence: is every factual claim backed by something real?
- Permissions: was any private or restricted information used that should not have been?
- Voice: does it sound like your organization, or like a generic tool?
- Accessibility: can the people it is for actually use it?
- Claims: does it promise or state anything your organization cannot stand behind?
A colleague's thank-you letter reads beautifully and includes a warm line about a program result. You check it. The result was never verified, and the number came from the tool. You catch it before it reaches a donor. The letter still reads well after the fix. That is a good review.
Try it
Next time you review a teammate's AI-assisted draft, do not just read it. Run these five checks: evidence, permissions, voice, accessibility, claims.
RememberReviewing AI-supported work means checking evidence, permissions, voice, accessibility, and claims, not just the writing.
Lesson 3
Communications and creative rights
Published work carries extra duties: it has to be authentic, respect other people's rights, and protect the trust you depend on.
The moment work goes public, more is at stake than accuracy. Your supporters trust that what you show them is real. Other people and creators have rights you have to respect. And a shortcut that breaks either one can cost trust you cannot easily rebuild.
- Authenticity: do not present AI-made images or stories as real events that did not happen.
- Synthetic media: be careful with generated faces, voices, or scenes, and be honest when you use them. A generated face or voice can carry separate legal risk, so get sign-off, not just a stock license.
- Rights and likeness: do not use someone's image, voice, or work without the right to do so.
- Donor trust: never let a polished shortcut misrepresent your work to the people who fund it.
Important
A generated image of a beneficiary who does not exist, shown as a real person you serve, is a trust breach even if the intent was kind. When in doubt, be honest about what is real and what was made.
Try it
Think of one piece of public content your team makes. Is there any point where a generated image, voice, or claim could be mistaken for something real? What would keep it honest?
RememberPublic work must stay authentic and respect rights, because donor and community trust is hard to win back.
Lesson 4
Escalate without blame
Problems only get fixed if people feel safe raising them. Blame drives them underground.
Every team hits AI problems: a real incident, a near miss that almost went wrong, a gap in the policy, or someone quietly using a tool no one approved. What decides whether these get fixed is whether people feel safe naming them.
If raising a problem gets someone in trouble, they stop raising problems. Then the small issues you could have fixed grow into the ones you cannot. A team that treats a near miss as useful information, not a failure, gets safer over time.
Try it
Think of one AI worry you have noticed but not raised. What would make it feel safe to name? That answer is what a learning team is built on.
RememberMake it safe to raise incidents, near misses, and workarounds, or they stay hidden until they cost you.
Lesson 5
Build a learning team
Good practice spreads when a team shares what works and feeds it back into how they operate.
The strongest teams do not rely on everyone being careful alone. They build shared habits, so good practice is the easy default and not a personal effort each time.
- Approved patterns: agree on a few AI uses that are known to be safe and useful.
- Peer review: make it normal to have a second person check AI-supported work.
- Shared examples: keep good and bad examples where the team can learn from both.
- Policy feedback: send what you learn back up so the rules improve over time.
Note
None of this needs to be heavy. A short shared list, a habit of a second look, and a way to pass lessons upward will carry a small team a long way.
Try it
Pick one of these four your team could start with: an approved pattern, a peer-review habit, a shared example, or a way to send feedback up. What is the smallest first step?
RememberA learning team shares safe patterns, reviews each other's work, and feeds lessons back into policy.
What you leave with
Team AI working agreement and content approval record
A short working agreement your team can adopt. It sets what to disclose, how to review each other's AI-supported work, and how to protect trust and rights. It adds how to raise problems safely and a simple record for approving content. It turns careful individual habits into shared team behavior.
This is just for you. It saves on this device only, and nothing is scored.