Articles·Strategy and support

“A glorified Google search.” Why AI training for employees has to start with agents.

A classroom of adults at laptops while a robot teaches at a chalkboard and one student points at her screen.

AI training for employees works when it moves your team from asking AI questions to handing AI work. The fix is short lessons on how the tools actually work, then practice on your team’s own tasks, where they tell an agent to do the job on the computer instead of pasting answers around. Most teams still treat ChatGPT, Claude, or Gemini as a glorified Google search: question in, answer out.

Why isn’t my team getting more out of AI?

Your team isn’t getting more out of AI because they’re using the chat window for questions and answers. They upload documents into the chat instead of telling their agent to go work on something on the computer. They copy the response, paste it into another system, and edit it there.

That copy, paste, edit loop is where the time goes right back out. An account manager who pastes a renewal letter into ChatGPT, copies the draft into Outlook, and rewrites half of it hasn’t saved much. That habit is the biggest learning curve.

What do most agencies try first?

Most agencies try a company account and a lunch and learn. Everybody gets a login, somebody shows a few good prompts, and the team goes back to asking questions in the chat box.

The other common try is a shared prompt library: a document full of good questions to copy. Prompts help. A prompt library still keeps AI in the question and answer box, and nobody on the team sees what an agent can do when you hand it a folder and a job.

Wanna fix it?

A team treating AI like a glorified Google search gets fixed with AI training for teams: short lessons on how the tools work, practice on your own tasks, and a clear line on what a person still checks. Here’s how I teach it at DoBetter.

  1. Learn the difference between chat and agent. A chat answers. An agent does the work: opens the files, fills in the spreadsheet, writes the draft where it lives. Chat vs Agent is lesson 4 in my AI Basics library.
  2. **Stop uploading. Point the agent at the folder.** Claude’s desktop app has a mode called Cowork that works on files on your computer. Tell it which folder and what to do with it.
  3. Break the copy, paste, edit loop. If you’re rewriting every answer, give the agent your rules up front: your voice, your format, and what it can never say.
  4. Train on your own tasks. The team workshop is a prepared 2 to 3 hour session built on the work your team actually does, with notes and one follow up.
  5. Keep lessons short. The AI Basics library at dobetter.plus/concepts has 30 lessons, about 2 minutes each, in six chapters that run from the basics to staying safe. The first 3 are free. Every client gets the full library on day one, plus new lessons as I add them.

What does it look like once your team gets it?

Once your team gets it, the question changes from “what should I ask AI?” to “what should I hand it?” A coordinator can tell an agent to pull the renewal dates out of a folder of PDFs and into a spreadsheet, then check the result. A producer can have the agent write the recap email in the right format the first time.

A person still checks the work before it reaches a client. Lesson 21 in my library is called Checking Its Work, and lesson 30 is When to Stop and Ask a Human. Training that skips those two produces confident mistakes faster.

Can you train your team on AI yourself?

Yes, you can train your team on AI yourself if one person on staff is already curious and ahead. Free tutorials from the AI companies cover the buttons. Pick one repetitive task, have that person run it with an agent for two weeks, and have them show the team what changed.

Self training stops working when the tutorials teach a general office and your team works in carrier portals, census files, and renewal season. It also stalls when the one curious person gets busy. One curious person, let them tinker. A whole team typing into a chat box, train them together on real work.

Quick answers

What should AI training for employees cover first?
The difference between chatting with AI and handing an agent a task on the computer.
What’s the biggest mistake teams make with AI?
Using it as a glorified Google search, then copying answers into other systems and editing them by hand.
What is Claude training?
Learning to use Claude, Anthropic’s AI assistant, including the desktop app’s Cowork mode, which works on files on your computer.
How long are DoBetter’s AI Basics lessons?
About 2 minutes each. There are 30, and the first 3 are free at dobetter.plus/concepts.
How long is a team AI workshop?
DoBetter’s is a prepared 2 to 3 hour session on the team’s own tasks, with notes and one follow up.
Does AI training mean nobody checks the work?
No. A person still checks what the agent produces before it reaches a client.