How to Plan AI Training for Your Employees

Many organizations now have some form of AI access in place: Microsoft 365 Copilot licenses, a ChatGPT or Claude workspace, or Gemini inside Google Workspace. In our experience, far fewer have a plan for teaching people to use it well. When we scope private training with an organization, the same handful of decisions comes up every time, and getting them right matters more than the length of the session.

This guide walks through those decisions in the order we take them.

Start From The Tasks Your Team Already Does

The most useful training plans begin with a list, written by the team, of the work that fills their week: the reports they assemble, the documents they read and summarize, the emails and meeting follow-ups, the spreadsheets they inherit. Ask each person for three tasks they repeat and one they dread.

That list does two jobs. It tells you which exercises will carry over to Monday morning, and it shows where AI is a poor fit. A task that depends on information nobody has written down, or on a decision that has to stay with a specific person, is better left out of the first session.

Find Out What Access People Actually Have

Access is the detail that most often derails a session. Two people in the same organization can open what looks like the same tool and see different features, because what is available depends on the plan or license, the account they sign in with, and the settings their administrator has chosen.

• Microsoft Copilot: Copilot Chat and licensed Microsoft 365 Copilot reach different information, and the account a person signs in with decides which one they get
• Gemini: some people have Gemini inside Gmail, Docs, Sheets, and Drive, and others only the standalone Gemini app, depending on the Google Workspace edition and admin settings
• ChatGPT and Claude: Projects, file uploads, research features, connectors, and sharing all vary by plan and workspace configuration

Confirm this before anyone builds an agenda. An exercise built around a feature half the room cannot see is a lost half hour.

Train On One Platform At A Time

If your organization runs more than one AI tool, it is tempting to cover them all in a single session. Even across a full day, that leaves very little time for practice. We usually recommend one agreed platform per session, taught with a method that carries over to the others: state the goal, supply the context and sources, describe the output, and check the result.

Comparing tools is a legitimate goal of its own. When a team is still choosing, a facilitated comparison on the team's own tasks is a better use of the time than a tour of features.

Make Checking Part Of Every Exercise

The habit that matters most happens after the output appears. Every exercise should end with the participant reading the result against its source: the claim against the document it came from, the total against the spreadsheet, the summary against the meeting it summarizes. AI tools produce confident drafts whether or not the draft is right, and the dependable control is a person who knows what to look for.

It also helps to agree, as a team, on what does not go into a prompt at all. Most organizations already have a data policy; training is the moment to translate it into specific examples people will recognize.

Space The Sessions Out

A first round of training can establish a shared method and take people through one real workflow. The skills that compound take longer: reusable instructions, shared Projects or prompt libraries, agents and custom GPTs, and team conventions for review. They also take practice between sessions.

If your schedule allows, split the training instead of running it in one sitting. Live online, the same six hours can run as one day, two three-hour sessions, or three two-hour sessions. Spacing them a week or so apart gives people time to try the method on their own work, and the questions they bring back are where the training gets specific.

Questions To Ask Any Training Provider

• Will the exercises use our own tasks and documents, or approved examples close to them?
• How will you confirm which features our participants can access before the session?
• What will participants leave with: worked examples, reusable instructions, a review checklist?
• Is the training independent of the companies whose tools it teaches?
• How are sessions priced, and what happens after the first one?
• Are there discounts for nonprofit or public-sector organizations?
• Does the training count toward PDH or CEU requirements for licensed staff?

How We Can Help

AIPTD is a 501(c)(3) nonprofit, and we run private, live AI team training on ChatGPT, Claude, Gemini, and Microsoft Copilot, online or on-site. Our programs are built around what a team needs to work with AI together: a common language, agreed rules of play, and shared skills. If you are planning training for your organization, see AI Training for Teams, where you can also request a free consultation. Nonprofit, city, state, and federal organizations receive a 50% discount, and our courses provide PDH and CEU credits.