← All insights

Is Your Team Using AI the Right Way? A Quick Check for Member-Based Organizations

Here is a question worth asking at your next staff meeting: "Who here has used an AI tool for work in the last month?"

In many organizations, more hands go up than leadership expects. Someone uses a chatbot to tidy up member emails. Someone else summarizes committee minutes. A third person drafted the annual report outline over a weekend on a personal account.

None of that is necessarily wrong. But if it is happening without guidance, you have what is often called "shadow AI": real use of AI tools that the organization can't see, hasn't approved, and can't support. For organizations that hold member trust and member data, that is worth a closer look.

Why member-based organizations should pay extra attention

Your members give you their names, contact details, employers, payment history, and sometimes far more sensitive information: grievances, health or benefits details, certification records, or political views. They trust you to look after it.

When a staff member pastes a spreadsheet of member records into a free AI tool, a few things may be true that they never considered:

• The tool's terms may allow the provider to keep or use what was entered.
• The data is now outside your systems, your access controls, and your retention rules.
• Nobody in the organization knows it happened, so nobody can answer a member who asks.

The fix is not to ban AI. Bans tend to push use further out of sight. The fix is to make the safe path the easy path.

What an acceptable-use policy should actually cover

A good AI acceptable-use policy is short enough that people read it. One or two pages is plenty to start. It should cover:

• Approved tools. Name them. Say whether staff should use organization accounts rather than personal ones.
• Off-limits information. Be specific: member records, donor or payment data, personnel matters, legal matters, anything marked confidential.
• Human review. A person checks AI-assisted work for accuracy and tone before it goes to members or the public.
• Disclosure. When, if ever, you tell members that AI helped produce something.
• Off-limits uses. For example, using AI to make decisions about an individual member's status, discipline, or eligibility.
• Who to ask. One named person or role for questions, new tool requests, and mistakes.

Where your organization has legal or contractual obligations about data, have counsel review the policy. A policy is a practical tool, not a substitute for legal advice.

A 10-point quick check

Go through these honestly. Each "no" or "not sure" is a place to start.

1. Do we know which AI tools our staff are using today?
2. Do we have a written AI acceptable-use policy that staff have actually seen?
3. Have we named the AI tools that are approved for work use?
4. Are staff using organization accounts, not personal ones, for work-related AI use?
5. Have we spelled out what member and donor information must never be entered into AI tools?
6. Have we read the data and privacy terms of the tools we approved?
7. Does a person review AI-assisted content before it reaches members or the public?
8. Is there a named person staff can go to with questions or mistakes, without fear of blame?
9. Have staff received practical training on using AI well, including how to check its output?
10. Do we revisit our AI rules on a regular schedule, such as every six months?

If you answered "yes" to most of these, you are ahead of many organizations. If you answered "no" or "not sure" to several, you are in good company, and the gaps are fixable.

Turning the check into action

Start with the first two items. Find out what is really being used, in a no-blame way, and put a short written policy in place. Those two steps reduce most of the risk on their own.

Then train people. A policy tells staff what not to do. Training shows them how to do the work well: writing good instructions, spotting confident-sounding errors, and keeping member data out of the wrong places.

Finally, set a date to look again. AI tools change their features and terms often. A rule you wrote this year may need updating next year.