A practical way to decide what to tackle first.

You have a clear picture of where staff are already using AI, including public tools, built-in app features, meeting notes, browser assistants, or workarounds.

Staff know which AI tools are approved, what data is off-limits, and who to ask when a use case is unclear.

You are confident AI will not expose shared files, email, CRM records, finance data, donor/member/client data, or staff data to the wrong people.

Your first AI pilot is narrow enough to test safely: a clear owner, a small user group, allowed data, a success measure, and a stop/go decision.

AI learning is more than one training session: staff practice on real work with champions, cohorts, office hours, or role-based support.

There is a way to collect, compare, and approve AI ideas based on value, risk, data readiness, effort, and who will own the change.

The work you want AI to improve is understood well enough to redesign and measure, not just described as "make us more efficient."

For agentic or custom AI, you have a specific workflow candidate — such as client service, document intake, or internal operations — with known data sources, handoffs, human review, and an owner.