ChatGPT Training for Employees: Agenda and Exercises
A practical guide to employee ChatGPT training, with department-specific exercises, data-handling rules, a brand style guide and team policies.
👁 114 reads
I start ChatGPT training for employees with a list of information that must never be pasted into a chat, and only then move on to prompts. Teams often have had ChatGPT or Copilot accounts for a while, but everyone uses them differently - sometimes on personal accounts and with customer data. I run this kind of training for businesses, so here’s how I structure a workshop, the exercises each department can do, and what the company takes away afterward.
At a glance
- Format: a 4-hour online workshop for up to 15 people, or a 6-hour in-office day for up to 20, with exercises based on company materials.
- Order: restricted data and business accounts first, then writing, research, spreadsheets and department tasks, followed by team guidelines.
- After the training: prompts refined using company materials, checklists, a style guide and numbered certificates of completion.
Hour one: what you can share with ChatGPT
I start by explaining the difference between a personal account and a business account with model training on your data turned off. I show you where to check the settings that determine whether conversations can be used to improve the model, and who at the company decides which plan to use.
Next, we go through information that should never be entered in a chat. In HR, that includes candidate data and employee evaluations; in a law firm, confidential client information; in a medical practice, patient data; and in an accounting firm, client data. We use examples that have been anonymized before the session.
The third topic is hallucination. Using a task from participants’ work, I show how a model can add facts that aren’t in the source material, how to ask for sources, and how to check them. We compare ChatGPT, Claude, Gemini and Copilot on the same task.
Department exercises based on real tasks
A single training session for the whole company can leave accounting bored by graphics and marketing bored by spreadsheets. That’s why I split groups by department or prepare separate exercises for each one.
| Department | What they practise with ChatGPT |
|---|---|
| Marketing | Social posts, newsletters and product descriptions using a brand style guide; checking whether ChatGPT recommends the company |
| Sales | Researching a prospective customer’s company with sources, drafting an offer and writing a follow-up email from personal meeting notes |
| HR | Job ads, onboarding materials and FAQs based on company documents; analyzing survey responses after anonymization |
| Accounting and finance | Excel and Google Sheets formulas with explanations; extracting data from documents into a table and checking a sample |
| Customer service | Drafting replies from a knowledge base, summarizing long threads and recognizing when a customer needs to speak to a person |
Finance follows one rule: the model can organize, describe and search, but formulas in a spreadsheet or a person must always check the numbers. Models can get addition wrong.
Use a style guide instead of “write professionally”
In marketing, I see the same problem most often: everyone writes prompts their own way, yet the copy still sounds as if it came from the same machine. During training, we create a brand style guide covering words the company uses and avoids, sentence length, and examples of good and bad copy. You can give the guide to new team members and include it with every prompt.
We also make a list of patterns that can make readers recognize AI-generated text: empty summaries at the end of paragraphs, strings of three adjectives, and constructions such as “not only … but also.” Participants use the list to review their copy before sending it.
The final 30 minutes: where your team keeps its prompts
A common mistake after training is leaving prompts in one person’s chat history. So we use the end of the workshop to agree on guidelines: which tools the company allows, which data is off limits, who approves content sent outside the company, where the shared prompt file is stored and who updates it.
After the session, the company receives prompts from the exercises refined using its own materials, checklists for reviewing AI-generated copy, reports and graphics, and numbered certificates for participants. A tailored programme also includes a draft AI policy for review with a lawyer.
How to arrange ChatGPT training for employees
The online workshop lasts 4 hours and the in-office day lasts 6 hours. For the online workshop, you choose 2-3 of the seven modules; for an in-office day, you choose 4-5. To discuss a date and the details, use the form on the AI training for businesses page. I reply within 24 business hours.
If your team would rather learn independently, the guide to AI training compares three options: training, courses for teams and running training sessions in-house. For an introduction to ChatGPT itself, see the ChatGPT guide.
Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →Frequently asked questions
How long does ChatGPT training for employees take?
The online workshop lasts 4 hours; the in-office day lasts 6 hours. For companies that need more, I offer a tailored programme with several sessions for different departments, giving the team time to apply what they’ve learned at work between sessions.
Can employees use personal ChatGPT accounts for work?
It’s better not to if they’re entering company or customer data. During training, we agree on which tools and plans the company allows and what information must never be entered in a chat.
Does ChatGPT training meet the requirements of Article 4 of the AI Act?
Article 4 requires measures to support AI literacy; it doesn’t prescribe a particular training course. A workshop with certificates for participants is one such measure and gives the company a record of its activity.
