A short film.
One idea per module, shown on a real screen, not a slide deck.
▶ About the length of a coffeeAI training for employees
Teach the tools once?Teach them as they change.
82 short modules on the work people already do: a film, a short read, one prompt to try on a real task and a check scored on the server. Your team learns the version of the tool on their screen this week.
Priced per seat, per year.
The short answer
What is the best way to train employees to use AI?
Start everyone on the same foundations, then put AI to work on each person's real tasks, and check the skill rather than the attendance. Keep the training current, because the tools change every month.
Train Space does that in three programs: AI essentials, AI on real work, and Agents and automation. Each ends in a credential your people can share.
- Modules
- 82, with 33 in the scheduled core
- Each module
- Watch, read, try on real work, check
- Pass mark
- 80%, scored on the server
- Pricing
- Per seat, per year
- Tools
- Copilot live; Claude, ChatGPT and Gemini in preparation
Inside every module
Short enough to finish.
Real enough to use.
Most AI training is a recording people half watch. Every Train Space module follows the same four steps, and the rules behind them are checked by the build, so a lesson can't ship if it breaks them.
A starting-point check places each person at the right level and in their department's version of the lessons. Nothing is locked by level: placement suggests, it doesn't gate.
350 words, at most.
The build fails any lesson whose read runs longer, or any single block over 80 words.
✓ Enforced, not promisedOne prompt, on your work.
The exact prompt to paste, with its own sample input, and what a good reply looks like.
✦ Sample input includedA check that means something.
At least eight questions, scored on the server. 80% to pass, and a refresh check when it fades.
↻ Mastery that stays currentThree programs, one path
From a first prompt
to a team that runs on it.
Every program ends in a credential with a public verification link and an Add to LinkedIn button. Here is the scheduled core of each one, as module titles.
EVERYONE, FROM DAY ONEStart here
AI essentials
What these systems are, how to ask well, what may go where, and how to tell an answer is wrong before you act on it.
- What is a large language model doing when it answers?
- How do you tell an answer is wrong before acting on it?
- What may be pasted into which tool?
- How do you write a request that returns a usable result?
- What does human oversight actually require?
- When is a task actually worth handing to AI?
- Is AI actually saving you time, or does it just feel that way?
- What is a token, and why does it set the bill?
- What happens when the conversation outgrows what the model can hold?
- Why does it matter which account you paste work into?
- What is RAG, and when does it beat fine-tuning?
- Why do two examples beat a page of instructions?
- How can a document someone sends you take over your assistant?
- When and how should you tell people AI helped?
PEOPLE USING AI ON THEIR WEEKLY WORKPer seat
AI on real work
Documents, data, meetings and inboxes: briefing AI like a contractor, turning repeat tasks into reusable assistants, and reviewing what comes back.
- How do you hand AI a task the way you'd brief a contractor?
- Why should AI ask you questions before it drafts anything?
- How do you get AI to criticise your plan instead of agreeing with it?
- What makes AI work look done but not be done?
- How do you get through a 300-email Monday without letting AI reply for you?
- How do you turn meeting AI into action, not just notes?
- Why does "search everything" give worse answers than "use these three files"?
- How do you get straight answers out of a 200-page document?
- How do you get a deep research report you can actually trust?
- How do you turn a task you repeat into a reusable assistant?
- What changes when AI can read your email, drive and CRM?
- Which of your tasks does AI do well, and which does it quietly botch?
- When does connecting one more app make your assistant dangerous?
- How do you work out what AI changes in a role, task by task?
- How do you turn "watch me do it" into a written procedure?
A few from the library
- Why ask for ten versions when you only need one?
- Why shouldn't you ask AI to judge a conflict you're in?
- How do you practise a difficult conversation before it happens?
- How do you turn a messy walk-and-talk into a usable first draft?
- What changes when AI edits your document directly instead of chatting about it?
- How do you get a deck that's on-message, not just pretty?
CHAMPIONS, MANAGERS AND BUILDERSPer seat
Agents and automation
Agents, automations and pilots that prove something, and turning one person's habits into the whole team's.
- How do you check work an agent did while you were away?
- How do you run an AI pilot that can prove something?
- How do you turn one person's AI habits into the whole team's?
- How do you review AI work when it produces a hundred items, not one?
A few from the library
- How do you find the errors in a spreadsheet model you inherited?
- How do you test a scenario without rebuilding the model?
- How do you turn a thousand survey comments or tickets into themes you can defend?
- How do you turn twelve customer interviews into evidence, not a story?
- What is a browser agent good for, and where does it go wrong?
- Where does AI fit inside an automated workflow, and where does a person sign off?
Written for the work in front of them
Not “AI for everyone.”
AI for your HR team. And legal. And yours.
The same skills, in the scenarios each team actually meets.
01 / DEPARTMENT LENSES
HR, legal and GRC versions.
Lenses rewrite each module's scenario, example and callouts for the team. Placement picks the right one, so a recruiter never reads a finance example.
02 / DEPARTMENT MODULES
Their own use cases.
“How do you get a first pass on a contract without missing what isn't there?” and “How can AI help build a fair, structured interview without deciding who gets hired?”.
03 / YOUR OWN CONTENT
Modules only your company sees.
An organization can add modules of its own beside the shared curriculum. They are held to the same reading, sourcing and check rules before they ship.
The tools your team actually uses
One place for every assistant.
Dated, sourced, and honest about what's ready.
The general programs teach skills that carry across tools. The tool courses teach each product as it is this month.
Live now
Microsoft Copilot
25 modules from Chat to Word, Excel, Teams, agents and Cowork, as of 30 October 2026.
See the Copilot courseIn preparation
Claude
Claude on the web, the desktop app, Claude in Chrome and Claude Code, with every new feature as it ships.
The planned outline
- Where Claude lives now. The plans, the web, desktop and phone apps, and the single Claude experience that is replacing separate Chat and Cowork.
- Everyday chat. Uploads, web search, choosing a model and effort, incognito chats and visuals in replies.
- Context that sticks. Memory, chat search and Projects, so Claude knows your work without retyping it.
- Making things. Files, artifacts, Claude Docs, Slides and Design, and dashboards from your own data.
- Connected work. Connectors to Google Workspace and Microsoft 365, Research, and Claude Tag in Slack.
- Inside your apps. Claude for Excel, PowerPoint and Google Workspace, used on a real file.
- Making Claude yours. Skills and plugins: package a method once and reuse it.
- Hand it off. Cowork tasks, scheduled tasks and Dispatch from your phone.
- Claude acts for you. Claude in Chrome and computer use, with the safety habits both need.
- Claude Code for non-developers. The desktop Code tab, then the CLI: project memory, skills, MCP and cloud sessions.
In preparation
ChatGPT and Codex
ChatGPT on the web, the desktop and phone apps, and Codex for code, with every new feature as it ships.
The planned outline
- What changed. Chat, Work and Codex, the plans, and the names that are gone: Atlas, custom GPTs, agent mode.
- ChatGPT on the web and phone. Chat, web search, images, voice, memory and projects for one ongoing piece of work.
- Finished work from ChatGPT Work. Documents, decks, spreadsheets, deep research and visualizations, then reviewing what comes back.
- Connect your tools. Plugins, skills, scheduled and event-triggered tasks, and ChatGPT in Slack and Teams.
- The desktop app. The built-in browser, Appshots, Computer Use, Record and Replay, Space pages and Sites.
- Delegation and agents. Dots, long-running work and notifications, and staying in control with permissions and approvals.
- Codex. Codex in the desktop app, the cloud, the CLI and your editor: reviews, AGENTS.md and safe changes.
- Admin and governance. For workspace admins: plugin controls, model availability, analytics and compliance.
In preparation
Google Gemini
The Gemini app, Gemini in Gmail, Docs, Sheets and Meet, NotebookLM and Gemini in Chrome, with every new feature as it ships.
The planned outline
- Where Gemini lives now. The web, phone and new desktop apps, Workspace, Chrome and Gemini Notebook, and which plan or account gets what.
- Everyday Gemini. Chat and models, Gemini Live, files in chat, Canvas, and images and video.
- Gemini in Workspace. Gmail, Docs, Sheets, Slides, Meet notes and the Drive side panel on your own work.
- Research. Deep Research over Gmail and Drive, and Gemini Notebook (formerly NotebookLM) with overviews and quizzes.
- Make it yours. Skills, which replace Gems from November 2026, and Connected Apps.
- Automate. Workspace Studio agents, Tasks, and Gemini in Chrome with auto browse.
- Builders' corner. Antigravity for non-developers, and what happened to Jules and Gemini CLI.
Why this matters
Buying the licence was the easy part.
Using it well is the work.
Three findings from the research, each linked to its source.
78%
of people who use AI at work bring their own tools. Only 39% have had any AI training from their company.
36%
of employees say the AI training they got was enough.
2.1x
as likely to use AI often when their manager backs it. Only 28% strongly agree their manager does.
A few good questions
Training a team on AI?
Start here.
Still wondering about something? Ask us and we’ll answer it here.
Ask your questionWhat should AI training for employees cover?
Four things, in this order: what the tools are and where they go wrong, what company data may go into which tool, how to ask for and check a usable result, and how to use it on the person's own weekly tasks. Train Space's AI essentials program covers the first three in 14 short modules; AI on real work and Agents and automation cover the rest.
Which AI tools does Train Space train people on?
The Microsoft Copilot course is live, dated as of 30 October 2026. Courses for Claude, ChatGPT and Codex, and Google Gemini are in preparation, with their outlines published. The three general AI programs teach skills that carry across tools.
How long does it take?
Each module is a short film, a read of at most 350 words, one prompt to try on real work and a short check. The scheduled core across the three AI programs is 33 modules; the other modules are a library people open when the task comes up. There is no fixed number of hours: people move at their own pace and managers see progress.
How do we know people actually learned it?
Each lesson ends with a check of at least eight questions, drawn and scored on the server so the answer key never reaches the browser. A module counts as mastered at 80%. Mastery fades: a few weeks later it needs a short refresh check. Use-case modules are completed with a record of the work the person did.
How is it priced?
Train Space is a subscription per seat per year, bought per program: AI essentials, AI on real work, Agents and automation, the tool courses and the security program. Ask for a quote for your team.
Can it be tailored to our company or department?
Yes, in two ways. Department lenses rewrite the scenarios and examples for HR, legal and GRC teams, and the starting-point check picks the right one. An organization can also have modules of its own, added alongside the shared curriculum and held to the same rules.
How do people sign in?
With Google, Microsoft or a six-digit email code. Access is invite-only, or an organization can let anyone with a verified address on its email domains join automatically.
Is AI training required by law?
Article 4 of the EU AI Act asks providers and deployers of AI systems to take measures to ensure a sufficient level of AI literacy among their staff. Whether it applies to you depends on where you operate; ask your counsel. Train Space gives you a dated record of who completed which training.
Put the learning to work
Give everyone a starting point.
Then keep them current.
Start your team on AI essentials, and take them all the way to agents.
Priced per seat, per year.