AI training for employees: what it should cover
AI training for employees should start from the tasks in their own jobs, teach them to check what AI produces and cover the company's rules for using it.
Employees should practise on their own tasks in the tools your company has approved. They should be able to show how they checked an AI draft and applied your rules. Put those requirements in one brief and ask each training supplier to show how their exercises meet it.
What employers' examples can add to your brief
Zapier's rubric gives you wording for quality standards; GitHub's office hours offer a format for help with employees' own work. A part a document leaves out may still be part of that employer's training.
| Employer and document | Their own tasks | Checking the output | Employer's own rules |
|---|---|---|---|
| Zapier, AI fluency rubric | Their own tasks "Integrate AI into real workflows, not just isolated or one-off tasks." | Checking the output "Compare AI-assisted outputs against your own standards for quality." | Employer's own rules Not in its published description |
| Salesforce, AI Fluency Playbook | Their own tasks "What’s one repetitive task that slows you down?" | Checking the output A learning path called "AI Output Validation & Quality Assurance" | Employer's own rules Not in its published description |
| GitHub, Fostering AI learning opportunities | Their own tasks Office hours where employees bring "a specific, real-world problem" | Checking the output Not in its published description | Employer's own rules Recommends an onboarding module pointing new employees "to the approved tools" |
| State of Georgia, training announcement | Their own tasks Not in its published description | Checking the output Content "double-checked for accuracy, hallucinations, and bias" | Employer's own rules "guardrails aligned to Georgia's policy framework" |
| State of Delaware, training announcement | Their own tasks Not in its published description | Checking the output Not in its published description | Employer's own rules The training covers the state's generative AI policy |
| State of New Jersey, Innovation Authority page | Their own tasks Not in its published description | Checking the output Not in its published description | Employer's own rules Staff are advised to begin training before using the state's internal AI assistant |
| State of California, state staff training page | Their own tasks "Identify opportunities for using GenAI to help perform job duties better" | Checking the output Not in its published description | Employer's own rules Not in its published description |
Start from the tasks in their own job
Have each participant bring a few recurring tasks from their own week, with material they may use in the approved tool. For a Sales Engineering team, AI training for presales builds one module per task, from preparing a meeting to RFP answers.
Salesforce's AI Fluency Playbook suggests a question to start with: "What’s one repetitive task that slows you down?"
GitHub's post on learning opportunities describes drop-in office hours with no set agenda. An employee brings "a specific, real-world problem" and gets help from the programme's leads and from colleagues.
If a task involves customer or personal data, decide first whether that data may go into the tool.
Teach people to check what AI produces
A participant should leave knowing what to check in an AI draft of their own work before anyone else sees it.
The Georgia Technology Authority's announcement of its staff training says that "all AI-generated content must be appropriately labeled and double-checked for accuracy, hallucinations, and bias". According to the announcement, the training will help employees put this into practice. To move up a level, Zapier's AI fluency rubric advises: "Compare AI-assisted outputs against your own standards for quality."
Zapier's hiring update includes this example of the minimum bar in sales: "Can demo their personal AI workflow live and explain where they trust it and where they don't."
Put your own rules into the course
Include your AI policy and approved tools in the course, so participants practise within the rules they will work under.
The UK government launched generative AI courses for civil servants in January 2024. Its blog post told learners: "You are encouraged to check your department’s policy on the use of AI before using any of the tools described in the courses."
Delaware's announcement describes a training for state employees in three self-paced parts, about one hour and 45 minutes in all. The training covers the state's generative AI policy alongside responsible use and the risks. New Jersey's Innovation Authority asks state employees to begin the AI training before they use the NJ AI Assistant, the state's internal chatbot. The training also shows them how to reach it.
If the rules are not written yet, decide them before the training: which data may go into which tools, how a tool gets approved and who answers for the result.
AI training for managers: what they need beyond the staff course
Managers need to decide which team workflows should change, alongside learning to use AI themselves.
GitHub's workforce playbook says that for managers "the focus is not just about their personal AI usage, but about equipping them to lead their teams through this change." It suggests that executives "challenge managers to rethink their teams' workflows, to identify tasks ripe for automation, and to redefine what high-value work looks like."
California's state staff training page suggests that departments consider a phased approach. In the suggested order, executive leadership, generative AI leads and legal, labour and privacy specialists go first. The business owners and technical experts who oversee programmes likely to consider the tools follow, then the general workforce. The UK government's AI Playbook describes five groups of learners, from beginners to leaders. It asks that learning for leaders focus on AI's "potential impact on organisational culture, governance, ethics and strategy."
Questions to ask a training supplier
Ask the supplier one question for each part, and ask to see the material that answers it.
Which of our tasks will participants work on, and what material may they use in our approved tool?
Where in the course do participants practise checking AI drafts of their own work?
Where do our AI policy and approved tools come into the course, and who on our side confirms that the course matches them?
How will you help managers decide which team workflows to change?
In AI training, participants practise in the tools the company has approved and learn where a person must decide. I tailor the content to your company and to the department taking part.
Sources
- How to measure your AI fluency, and move to the next level. Zapier, 2 February 2026.
- Raising the AI fluency bar for every Zapier hire. Zapier, 31 March 2026.
- AI Fluency Playbook. Salesforce, undated.
- GitHub’s internal playbook for building an AI-powered workforce. GitHub, undated.
- Playbook series: Fostering AI learning opportunities. GitHub, undated.
- New AI training helps employees use technology with confidence and care. Georgia Technology Authority, 19 November 2025.
- Delaware Launches New Statewide AI Training for State Employees. State of Delaware, 19 March 2026.
- Generative AI in New Jersey State Government. New Jersey Innovation Authority, undated.
- State staff training. State of California, undated.
- Artificial Intelligence: introducing our series of online courses on generative AI. Government Digital and Data, UK Government, 19 January 2024.
- Artificial Intelligence Playbook for the UK Government. UK Government, 10 February 2025.
Facts last checked on 27 October 2026.
Updated 27 October 2026.