Why an AI workshop did not change daily work, and what makes it stick

The workshop went well. People asked good questions, a few tried the tools that same afternoon, and the feedback forms were warm. A month later, the inbox, the reports and the quotes are done exactly the way they were before. Nobody objected to AI. The old routine simply won, and it is hard to say which part of the day failed.

The short answer

A one-off workshop rarely changes daily work, because it teaches a skill and then leaves people alone with a full week. Training sticks when it runs on the team's own recurring tasks, when the accounts and settings are ready before the session, when there is a follow-up after real use, and when one person owns the result and checks it. Plan the 30 days after the workshop as carefully as the workshop itself, and measure use, not satisfaction.

Why one-off workshops fade

• Generic examples. A demo that writes a poem or plans a trip proves the tool works. It does not show anyone how to answer their own Tuesday emails with it.
• No follow-up. The first real obstacle shows up days later, at a desk, with no one to ask. Without a second session, that obstacle ends the experiment.
• Settings and permissions not in place. If people leave the room without a work account, the right connections and a clear rule on which data may go where, the careful ones stop and the rest improvise with their own tools. In Microsoft's 2024 Work Trend Index, 78 percent of AI users said they bring their own AI tools to work.
• No time. Practice competes with deadlines. If nobody sets aside time in the first weeks, deadlines win every time.
• No owner. When nobody is responsible for adoption, nobody notices it stalling.

The research points the same way. In BCG's AI at Work 2025, a survey of more than 10,600 workers in 11 countries, only 36 percent of employees felt adequately trained, and those who received five or more hours of training, especially in person and with coaching, were significantly more likely to become regular users. Microsoft found that only 39 percent of AI users had received AI training from their company.

A worked example

Example, for illustration. A 12-person accounting office holds a 3-hour workshop. That is 36 hours of the team's own time. At an assumed internal cost of 150 NIS an hour, the session costs 5,400 NIS in staff time alone, before any tool seats.

Now assume two outcomes after 30 days. In the first, 3 of the 12 still use the tools every week, mostly for email. In the second, the office adds what makes it stick: two follow-up sessions of 1 hour each (24 more hours), an owner who spends 1 hour a week for 4 weeks (4 hours), and 20 minutes of practice a week per person for 4 weeks (16 hours). That adds 44 hours, so 80 hours in all, or 12,000 NIS of internal time at the same assumed rate.

The second version costs more than twice as much in time. It is worth it only if it moves real work. If 9 of the 12 end up drafting client replies and monthly summaries with AI, and each saves an assumed 1 hour a week, that is 9 hours a week, which covers the extra 44 hours in about five weeks. These are assumptions to test, not results: the point of the follow-up is to find out which numbers hold in your office.

How to make the next workshop stick, step by step

1. Choose two recurring tasks per role before the session, with the people who do them. Bring real, anonymized examples.
2. Set up the accounts first. Work accounts, the connections the tasks need and a one-page rule on which data may go into which tool, all ready before people arrive.
3. Build in the room. Each person leaves with a saved prompt, project or assistant that already handles one of their tasks.
4. Write it down the same week. A short playbook per task: inputs, steps, what to check before anything leaves the office.
5. Block practice time: 20 minutes a week per person on the calendar for the first month.
6. Hold a follow-up at day 14 and at day 30. Fix what got stuck, share what worked, retire what did not.
7. Measure use. On Claude's Team plan, owners can open usage analytics from the account menu and see weekly active members against assigned seats. Other tools have similar admin views; a short weekly check-in works too.

Mistakes that cost the most

• Measuring the workshop by its feedback form. People enjoy a good session. The number that matters is who uses the tools a month later.
• Training before the tools are ready, so the first week goes to logins and permissions instead of work.
• One session for everyone. Sales, finance and operations repeat different tasks. Mixed examples help nobody enough.
• No owner with time. Adoption needs one person who checks in weekly, not a manager who asks once.
• Skipping the data rules, which leaves the careful people out and the risky habits in.

When outside help is worth it, and when it is not

If one person on the team already uses AI well, has an hour a week and can prepare examples from your real tasks, run the 30-day plan in-house. Outside help is worth it when nobody has that time, when several teams need different tasks, when the data rules are unclear, or when an earlier workshop already faded and the team needs a different approach.

The AI training page describes that approach: hands-on workshops on the team's own tasks, written playbooks, tool setup and follow-up sessions until the tools become habit.