From “Everyone Has Tried ChatGPT” to a Trained Team

“Is your team using AI?” I ask every IT director this, and the most common answer is some version of “Oh, yeah, everyone’s tried ChatGPT.” It’s said with a little pride, and I understand why. Two years ago, nobody had.

But tried isn’t trained. In practice, “everyone’s tried it” usually means each person used it for a handful of emails, got a mediocre answer to a vague question, and either decided it was overhyped or kept using it in the same shallow way. What’s missing is an AI training plan for the IT team: a structured path from casual use to people who can use AI well, support others using it, and spot when it’s wrong.

Why IT has to go first

Your IT team isn’t just another group of users. When the business adopts AI, IT will be asked to support it, secure it, and increasingly build with it. Users will bring their AI questions to the service desk. Leadership will ask IT whether a tool is safe. Someone will need to review the script a colleague generated with AI before it runs in production.

Self-taught habits make all of that harder. The most common risk I see isn’t a lack of skill. It’s confident skill with bad habits: pasting a configuration file with credentials into a public chatbot to troubleshoot it, or running AI-generated code without reading it. Structured training fixes habits as well as building skills.

Here’s how the assessment asks the question, and the 0 to 4 ladder I score it against:

K1. How hands-on is your IT team with AI tools?

0. Little or no hands-on use
1. A few enthusiasts, self-taught
2. Most have tried AI tools; no structured training
3. Most of the IT team has completed structured AI training
4. Ongoing, role-based AI training with hands-on labs

Level 2 is where “everyone’s tried ChatGPT” lands, and it’s where most mid-size IT teams are today. The jump to level 3 doesn’t require a big budget. It requires a plan, protected time, and training on the tools you’ve actually approved.

What IT people actually need to learn

A generic “intro to AI” course wastes most of your team’s time, because different roles need different skills. Start with a shared foundation, then split by role.

Everyone in IT: how to write a useful prompt, what never to paste into any AI tool, how to verify an answer before relying on it, which tools are approved, and what your AI acceptable use policy says.

Service desk: summarizing tickets, drafting knowledge-base articles, turning technical answers into plain language, and supporting users who are learning the approved assistant.

Infrastructure and cloud: using AI to draft and explain scripts, reviewing generated code before it runs, troubleshooting with logs that have been scrubbed of sensitive data, and the basics of cloud AI services.

Security: AI-specific threats such as prompt injection and data leakage, DLP for AI tools, reviewing AI vendors, and reading AI audit logs.

Builders and developers: calling models through APIs, building simple agents and automations, testing AI output systematically, and keeping prompts and configurations under version control. Growing at least one person into this role is its own project, covered in you need one AI builder on staff.

The asset: a six-week AI training plan for an IT team

This plan assumes about two hours a week of protected time per person. It works for a team of five or fifty.

Week 1: baseline and ground rules

• Each person completes three realistic tasks from their own job, without AI, and records how long they took. Keep the results; you’ll repeat them in week 6.
• Walk through the approved tools, the acceptable use policy, and a one-page “never paste” list: credentials, customer data, employee data, anything labeled confidential.

Week 2: prompting fundamentals

• A 90-minute hands-on workshop: giving context, stating the goal, providing examples, setting constraints, and iterating on the answer instead of accepting the first one.
• Homework: use the approved tool for five real tasks and save the best prompt from each.

Week 3: role tracks

• Split into the role groups above. Each group works through three exercises drawn from its own daily work.
• Collect the best prompts into a shared team library in your wiki. This becomes one of the most-used pages you own.

Week 4: hands-on lab

• Each person picks one recurring task of their own and uses AI to speed it up or automate part of it.
• End the week with a 30-minute show-and-tell. Peers learn more from each other’s real examples than from any course.

Week 5: verification and risk

• Deliberately test the tools’ failure modes: ask questions with wrong premises, request scripts for your environment and review them line by line, and compare answers against documentation.
• Security leads a short session on prompt injection and data leakage, using examples relevant to your environment.

Week 6: capstone and re-measure

• Repeat the week 1 tasks, this time with AI. Record time and have a peer rate the quality.
• Each person writes down the two ways they’ll use AI in their job going forward.

At the end, most of the team has completed structured training, which is level 3, and you have before-and-after numbers to show leadership.

Measuring whether it worked

Use four measures, all of which you’ll already have from the plan: the change in time on the week 1 tasks, the peer quality ratings from week 6, active use of the approved tool at 30, 60, and 90 days from its usage reports, and how many prompts people have added to the shared library. The first two show skill. The last two show whether the skill turned into a habit. Report all four to leadership in one short page; it’s the evidence that justifies the next round of training.

Making it stick

Training fades without reinforcement. Three habits keep it alive: a monthly 30-minute session where someone shares a new technique, the shared prompt library that people are expected to add to, and role-based refreshers when your approved tools gain significant new features. That ongoing, role-based rhythm with hands-on practice is what level 4 looks like. It needs a budget line and protected time, which is the argument in the case for a standing AI training budget.

For people who want a formal credential on top of the internal training, certifications add structure and an outside benchmark. Which AI certification your IT team should get first covers how to match them to roles, and which AI certification to get compares the options in detail.

Mistakes I see at this stage

The one-off lunch-and-learn. A single session creates enthusiasm for a week. Without follow-up, habits don’t change.

Training on tools you haven’t approved. If the training uses a consumer tool and the company approves a different one, you’ve trained people on the wrong thing and encouraged shadow AI in the process.

No protected time. “Do the training when you have a spare moment” means nobody does it. Put it on calendars.

Skipping verification. Teams that learn to prompt but not to check produce confident mistakes faster. Week 5 is the most important week of the plan.

Ignoring the skeptics. The person who thinks AI is overhyped is often the best reviewer of AI output. Give them that role instead of trying to convert them.

Where does your team actually stand?

Hands-on AI skill is one of 24 questions in the AI Readiness assessment, which covers six dimensions: data, security, infrastructure, skills, use cases, and governance. The free version is 10 questions and gives you a score in a few minutes.

Get your free AI Readiness Score →

Want to see what the full assessment covers first? Flip through a complete 38-page sample report.

Related guides

• Publish Your AI Acceptable-Use Policy This Month (3-Tier Template)
• You Need One AI Builder on Staff. Here’s How to Grow One
• The Case for a Standing AI Training Budget
• Which AI Certification Should Your IT Team Get First?
• The 6-Dimension AI Readiness Framework, Explained