If you've been told to "implement AI" at your association and you're not sure where to start, you are in good company. Most association leaders I talk to are somewhere on a spectrum between "we tried ChatGPT and it didn't stick" and "we have a policy but no one reads it."

So let's skip the panic and the hype. Here is the shortest defensible path from "we should try AI" to "we have evidence and momentum," in six moves.

Before you pick a tool, recruit a pilot, or write a policy, answer one question with your ED and a couple of department heads:

Document the answer in one paragraph. That paragraph belongs in your pilot kickoff message, your AI policy preamble, and your pre-pilot board briefing.

You need two documents and one decision before any staff member uses AI for real work.

The two documents:

• A formal AI policy. Governance, data privacy, ethical use, vendor requirements, incident reporting. IT and legal review it. (See our guidelines).
• A practical staff guide. The version your team will actually consult before pasting member data into a tool. (See our template).

What must never go into a public AI tool:

• Member data of any kind
• Financial data, budgets, forecasts
• HR information and personnel files
• Confidential documents and board materials
• Legal matters and contracts
• Passwords and credentials
• Proprietary strategy
• Personally identifiable information

If a staff member is unsure whether something belongs on this list, the answer is yes until proven otherwise.

The decision: Free accounts are fine for low-risk experimentation. Anything touching the prohibited list above belongs on a Team or Enterprise tier where your inputs are not used for training (be sure to adjust the appropriate settings in your tools - many times they default to "on") and your organization controls the data. This is non-negotiable for your base AI assistant - you want this on a paid tier.

Three months. Three phases. No mystery.

**Month 1: **Normalize and prepare. ED message to staff, policy drafted and legal-reviewed, one-pager out, 2 to 10 pilot members recruited

Month 2: Pilot, learn, and train. Show-and-tells, microlearning, Hallucination Hunts, personal logs, mid-pilot retrospective

Month 3: Scale, sustain, and measure. Champions named, first custom assistant built, broader staff enrolled, confidence survey repeated

Things to measure.

Two kinds of metrics, both required:

• Adoption metrics (immediate): staff trained, use cases shared, time saved, confidence scores before and after.
• Strategic outcome metrics (slow, 12 to 24 months): Did AI work move a member-facing number? A mission-level goal? A strategic plan KPI? Allow staff to innovate?

Adoption tells you the program is running. Outcomes tell you it's working.

Most associations end up with two tools in regular use: one embedded in their productivity suite, one standalone chatbot.

• Already in Microsoft 365 or Google Workspace? Start with the embedded option.
• Need a stronger writing or custom-assistant partner? Add ChatGPT or Claude.
• Have a specific high-friction workflow? Layer in one to three specialty tools (three we used at my last association: Gamma for slides, Descript for video creation and editing, and Venngage for infographics).

The tool(s) you standardize on this year may not be your best option two years from now. So train staff on prompting and output evaluation, not just on one platform's interface.

Most failed AI rollouts trace back to one of these.

• Skipping the strategy conversation. AI without strategy produces tool sprawl, not transformation.
• Leadership says "use AI" but doesn't model it. Visible curiosity from the top is worth more than any training program.
• Long workshops instead of microlearning. Five to ten-minute lessons embedded in real meetings build durable skill. Two-hour trainings get forgotten in a month. At one association, I created periodic Scribes (SOPs or how-to's) showing the team how to create custom prompts, for example; these took about 5 minutes to read and 20-30 minutes to follow. Here are a few staff activities you can run for about 30-45 minutes each to teach basic prompting skills.
• Trusting AI output without verification. AI output is a draft, not a decision. Pair AI use with stronger editorial standards, not weaker ones. Make sure AI output is reviewed by a human, preferably a subject matter expert on the content. It's also a good idea to disclose significant AI use on AI-generated output (typically with the exception of marketing).
• Losing momentum after the initial rollout. Build sustainment in from the start: monthly show-and-tells, quarterly champion check-ins, annual policy refreshes. During the roll-out phase, I emailed staff biweekly with AI tips to keep momentum and interest going.

Boards care about fiduciary duty, reputation, and member trust. So frame AI work as risk management, not a tech project.

I recommend four touchpoints across the year:

• Pre-pilot briefing (10 minutes): why now, what staff can and can't do, what we're protecting, when the policy comes back.
• AI policy review: real board input on risk tolerance, disclosure, and reserved decisions.
• Post-pilot update: adoption metrics, risks surfaced and addressed, what's next.
• Ongoing: AI becomes a quarterly paragraph in the CEO's report, until something material changes.

The board weighs in on policy and risk. The board does not weigh in on which day-to-day tools staff use.

You don't need a transformation. You need a defensible 90-day plan, a one-page staff guide your team will actually read, and one strategic priority AI is going to serve this year. The rest builds from there.

I built this resource so you don't have to start from scratch:

Free Quick-Start Pilot Guide

Questions about adapting any of this for your staff and policies? Book a free 30-minute consultation. I can help your association move from AI curious to AI capable in just 90 days.