Tokenacademy
Sign inBook a demoDemo

AI practice, at enterprise scale

Hands-on AI practice for every employee, coached and assessed automatically

Book a demoTry a session below
  • No setup
  • No AI licences to buy
  • Runs in the browser

Building with AI / Your First Chart

33,926 / 320,000 tokens286,074 left

Beginner 45 min

Your First Chart

Altea Group is renovating the break room at its head office. The office committee asked employees what they would like to find there, and 192 people answered. The results sit in a tiny file of 6 rows.

Your mission: ask the agent to turn this file into a chart that the committee can show at the next team meeting:

  • a chart that shows the votes for each option, read from the file;
  • a clear title and labels, so anyone understands it in a few seconds;
  • the winning option easy to spot.

Look at the preview after each request, and check that the numbers match the file.

320,000 tokens for this exercise: finish under 200,000 for +50 XP.

Instructions

  1. Open the data file data/break_room_survey.csv in the Files panel to see what it contains.
  2. Ask the agent to create a page with a chart of the votes for each option, using the data from the file.
  3. Ask for a title and readable labels, so someone who has never seen the file understands the chart.
  4. Ask the agent to make the winning option stand out (a different color, sorting, the number shown on the bar…).
  5. Check one number in the preview against the file.

Claude

Make a bar chart of the votes for each option, read from the file, with a clear title and labeled axes, and highlight the winning option in a different color.

33,926 tokens · 3 steps by the agent

Done: six bars from the survey file, a title and both axes labeled. Espresso machine (48 votes) stands out in a darker purple.

Ready to send: press Enter or the arrow

Office committee

Break room survey: what employees want

Number of votes for each option (192 employees answered). The most popular option is highlighted.

Votes for each break room option

01020304050EspressomachineComfortablesofasFreshfruitPing-pongtableBoardgamesPlantsOptionNumber of votes

Altea Group · workshop project

CCoachWatching quietly · steps in if you get stuck

Applied AI practice, not another video course.

AI is changing how work gets done, and how people learn it has to change too. A course recorded last year teaches last year's tool, and a multiple-choice quiz proves someone remembers a definition, not that they can use it.

The only way to learn is to get your hands on it: ask, watch the agent, check what it built, and do better on the next try.

A classic course

Video lessons like Udemy's, then a multiple-choice quiz.

Applied AI practice

A real agent, a real project, on Tokenacademy.

  • Learning to ask

    Classic course: Watch a video about writing prompts

    Applied practice: Write a request, and see what the agent does with it

  • AI's cost

    Classic course: A slide that defines a token

    Applied practice: See what each request costs, and manage your budget

  • AI's mistakes

    Classic course: A warning that AI can be wrong

    Applied practice: Check the result against the data, yourself

  • The proof

    Classic course: A multiple-choice quiz at the end

    Applied practice: Milestones checked on what you built

  • Keeping up

    Classic course: Recorded once, out of date at the next model

    Applied practice: Claude Code, Codex and GitHub Copilot as they are today

Every way to learn, on one platform.

Sessions to build, games to play, an assessment to prove it. Your designers can make more of each in the Studio.

Build a page

The learner describes a page, a chart or a tool; the agent builds it live. Every milestone is checked on the result.

Build an agent

An assistant with a role, rules, tools and skills, written in plain words, then tested on real customer messages.

AI Essentials

What an agent reads, counts and costs: tokens, context, the right model.

Copy the Target

Describe a page only you can see until the agent builds its twin, scored on likeness.

Crack the Vault

Get a password out of an assistant that guards it better at every level.

Lock It Down

Attackers want a secret from the learner's agent: make it hold, and keep helping real customers.

Final assessment

A free project, scored on what it does when the learner hands it in.

Rules you can add to any session

  • Token budgetA cap, and XP for finishing with fewer
  • ParEagle, birdie, par: the attempt against a good one
  • Request limitsA few requests, or a few characters each
  • Model choiceHaiku or Sonnet, and the effort, priced in credits
  • Context windowSee it fill, and clear it at the right time
  • Request structureContext, task, guidelines, constraints: scored out of 4

Learn to connect AI to your tools, without the risk.

Learners connect the agent to Outlook, SharePoint, Slack or Salesforce, and decide what it may read. Every tool runs on a made-up company: the skills are real, and no real data is ever at risk.

OutlookMicrosoft 365

Simulated, read only

What really needs me this week?

Looks through your mail for what needs you

  • Lyon hub incident – claim window closes 28 OctoberKarim Belkacem, Parcelia
  • Board pack – your CX numbers by Wednesday 9:00Élodie Marchand
  • URGENT!!! Your ad credits expire TONIGHTAdBoost MediaOnly looks urgent

Two things need you: the Parcelia claim, by 28 October, and your board numbers, by Wednesday 9:00. The ad credits only look urgent.

Model
Effort

One step of the agent: 21,000 credits

What AI costs, under control.

Every request a learner sends is counted in tokens, step by step, against the session's budget. They see what each request cost, and learn to get the same result with fewer.

  • A budget for each session. Past it, the session is over.
  • No request wasted. One that could not leave the agent room to work is refused, for free.
  • A report at the end. Request by request, with the tip that saves the most.
  • The cost in money. At your model's real prices.

Where your tokens wentExample

211,100 tokens of 320,000

  • Needed
  • Yours to save
  • The agent's own checks

Next time: say what "better" means in the first request. One message instead of two: about 94,000 tokens less.

Ready in the browser, nothing to install.

A learner clicks, and the session opens. No AI account to create, no key to manage: the agents are included.

  • Opens in the browser

    Every learner gets a sandbox of their own, ready with the session's files and data.

  • ClaudeOpenAIGitHub Copilot

    The real agents, included

    The Claude suite, the OpenAI suite or GitHub Copilot, run by the platform. No licence to buy for each learner.

  • No code on screen

    Learners describe what they want. Data reads as tables, documents as documents.

From a first request to an agent of their own.

Four programs, in the order your teams take them. About 45 minutes a session.

  1. Start here

    AI Essentials

    What an AI agent reads, counts and costs.

    • What's a Token?
    • Ask Once, Ask Clearly
    • Where Your Tokens Go
    • Same Result, Fewer Tokens

  2. Prompting with AI

    A request in four parts: context, task, guidelines, constraints.

    • The Four Parts
    • One Shot
    • Prompt Golf
    • Change Only That

  3. Building with AI

    A web page, a chart, a tool, built by talking.

    • Your First Chart
    • Sales Dashboard
    • Event Registration Form
    • Fix the Bugs in an App

  4. Build an Agent

    An assistant with a role, rules, tools and skills.

    • Meet an AI Agent
    • Give Your Agent Its First Tool
    • Don't Let Your Agent Get Tricked
    • Final Assessment: Your Agent, Live

Your own sessions, made by talking.

Your learning designers describe the session they want, with your files and your business cases. The Studio's agent drafts it while they talk: the mission, the steps, the milestones and the questions that check them.

  • Your cases, your data. Upload a file or paste an image, and the session is built on it.
  • Tried before it is published. Designers play it as a learner first; publishing stays their decision.
  • Every change is a version, and any version can be brought back.
  • Start from ours. Make your organization's version of any platform session.

New sessionDraft

PreviewHistorySettingsPublish

What should learners be able to do at the end? Tell me who it is for, and give me the files they will work with.

CSVsales_q3.csv

Ready to send: press Enter or the arrow

The session takes shape here as you talk.

A learning designer at her desk, working on her laptop, sticky notes on the wall behind her.
A team of colleagues gathered around a laptop in an office, looking at the screen together.

One-to-one, for your whole company.

Your teams, your brand, your agent, your sign-in. The platform follows every learner and checks every milestone, with no admin action to move on.

Book a demo

Altea Group

Search learners and sessions

  • Learning paths
  • Leaderboard
  • Learners
  • Studio

Your company's setup

  • Sign-inSSO with Entra ID
  • AgentClaude CodeCodexGitHub Copilot

Leaderboard

Building with AI, Altea Group only

This monthAll time

  1. 1CMCamille MartinSales3 of 3 milestones151,200 tokens
  2. 2HLHugo LefèvreFinance3 of 3 milestones163,480 tokens
  3. 3IDInès DuboisHR3 of 3 milestones171,900 tokens
  4. 4TRThomas RouxOperations3 of 3 milestones188,350 tokens
  5. 5SBSarah BenaliMarketing3 of 3 milestones196,020 tokens

Fewest tokens first, among those who reached every milestone. Demo data.

See it with your own teams.

A 30-minute call: we get to know your project, then show you the platform.

Book a demo