- AI practice
- Technical vs non-technical AI training
Technical vs non-technical AI training, for every team
Technical vs non-technical AI training on one platform: business teams chat with Claude or ChatGPT, developers work in Claude Code or Codex. Everything is customisable, so each team gets sessions made for it.
- Business teams: a chat, no code
- Developers: the real terminal
- Sessions made for each team

Technical vs non-technical AI training: the same session, two ways
One request, one chart built from the same survey file. Drag the seam: on the left, the agent in a chat; on the right, the same agent's work in its terminal. Pick another agent for either side.
✻ Welcome to Claude Code
cwd: ~/project
> Using data/break_room_survey.csv, make a bar chart of the votes for each option. It's for a team meeting.
⏺ Read(data/break_room_survey.csv)
⎿ Read 7 lines
⏺ Write(index.html)
⎿ Wrote 48 lines to index.html
⏺ Done: index.html shows a bar chart of the six options, sorted by votes. Espresso machine leads with 48 of 192.
>
? for shortcuts
Your First Chart
Using data/break_room_survey.csv, make a bar chart of the votes for each option. It's for a team meeting.
Read break_room_survey.csv
Created index.html
Done: index.html shows a bar chart of the six options, sorted by votes. Espresso machine leads with 48 of 192.
Reply to Claude…↑
For non-technical teams: no code, ever
Sales, HR, finance and operations work in a conversation, in the look of the AI app they already use. The agent writes code to build their page or chart, but they never see it: only what it did, in plain words, and the result.
- Claude in the look of the Claude app.
- ChatGPT in the look of the ChatGPT app.
- GitHub Copilot in the look of its chat, as in VS Code.
What the agent wrote
Never on their screen
On their screen
- Read the survey file
- Created the page
A bar chart of the votes: the espresso machine wins.
For technical teams: Claude Code and Codex in their real terminal
Developers, data and IT teams see everything, code included, and train where they will work: the vendor's own command-line agent, run in a sandbox of their own, with its commands, its shortcuts and its way of showing each step. A session can even ask for a command, such as /model or /review, and validate it when the learner types it.
- Claude Code Anthropic's agent, as its terminal shows it.
- Codex OpenAI's agent, as its terminal shows it.
- Commands a milestone met when the learner types the one the session asks for.
- Driving a coding agentrequests, commands, reviewing its changes
- Choosing the agent's modelwith the commands the agent has
- Working to a token budgetevery step counted
Which look for which team
Each session is set to one of three looks. The milestones, the coach and the checking stay the same.
| The platform's screen | A chat app's look | The terminal | |
|---|---|---|---|
| Suits non-technical teams | YesPlain words, no code | YesThe app they already use | NoCommands and a prompt |
| Suits technical teams | PartlyFine for a first session | PartlyFine for a first session | YesWhere they will work |
| Agents | YesAll five | YesAll five, each in its app's look | PartlyClaude Code, Codex |
| A milestone on a command | NoNo commands | NoNo commands | YesValidated when typed |
| Checked on what the learner built | YesMilestones | YesMilestones | YesMilestones |

Set per session, in one sentence
Learning designers pick the agent and its look in the Studio, by asking its agent or in Settings. The next learner who starts the session gets it.
For the developers, run it with Codex in its terminal.
The agent and its lookClaude, platform's screenCodex, terminalFor the sales team, make it look like ChatGPT.
The agent and its lookCodex, terminalChatGPT, ChatGPT appAdd a milestone: use the command /review.
Milestones3 milestones4 milestones
The rest of the session stays as it is:Mission and stepsFiles and dataMilestones and their questionsToken budgetCoachBadge
One platform for technical and non-technical learning paths
Each team follows its own path, on the same programs and the same reporting.
Business teams
AI Essentials and Prompting with AI in a chat, then Building with AI: a page or a chart, made by talking.
Technical teams
The same programs in Claude Code's or Codex's terminal, with sessions that ask for the agent's own commands.
Both, compared
Milestones and tokens per learner, whatever the look: one report for the whole company.
Questions about technical and non-technical AI training
What is the difference between technical and non-technical AI training?
What people see and drive. Non-technical teams talk to an agent in a chat and judge the result; technical teams drive a command-line agent, read each step and use its commands. What both must learn is the same: ask precisely, check the result, spend fewer tokens.
Which agents can learners use?
Claude, ChatGPT, Claude Code, Codex and GitHub Copilot, run by the platform. Each session names its agent; your organization chooses.
Is the terminal the real Claude Code or Codex?
Yes: the vendor's own command-line agent runs in the learner's sandbox, and the learner types in its real terminal. The platform runs it and counts its tokens; nothing is installed on the learner's computer.
Does a non-technical learner ever see code?
Not on the platform's screen or in an app's look (Claude, ChatGPT, GitHub Copilot): the agent's actions show as short rows (a file read, a page changed) and the result in the preview. The terminal shows what the command-line agent shows, which is for technical teams.
Can a milestone check that a learner used a command?
Yes, in a terminal session: the designer names the command (for example /model) and the milestone is met when the learner types it.
Can one session serve both audiences?
Make it twice in the Studio, set once for each: a chat for one team, the terminal for the other. Each team's learning path names its own.
Who sets the agent and its look?
Your learning designers, in the Studio: by asking its agent ("run it with Codex in its terminal") or in Settings. Both do the same.

See both looks with your own teams.
A 30-minute call: we get to know your project, then show you the platform.
Book a demo