Internal learning
Agentic Arena
Practice AI skills and climb the leaderboard.
Sign in with CGI Use your @cgi.com account.Preparing your learning arena…
Choose your challenge
Open-ended rated challenge
Streak Answer across all three roles. Each correct answer grows your Streak; the first miss ends the run.Use AI responsibly
AI User
- Tools & modelsChoose approved tools and know model capabilities.
- Prompting & guardrailsWrite clear prompts and use safe guardrails.
- CGI guidanceFollow CGI policy; ask AI Hub or AI SPOC when needed.
- Skills & agentsBuild simple Markdown skills and find agent use cases.
Build and run AI
AI Engineer
- LLM essentialsKnow model behavior, context limits, and inference.
- RAG & MCPConnect governed knowledge and tools to AI apps.
- Agents & deliveryRun agents locally or remotely and deploy safely.
- Model fit & telemetryUse evidence; track quality, cost, delay, and faults.
Design enterprise AI
AI Architect · Technical
- Guardrails & orchestrationDesign controls for suitable multi-agent systems.
- MCP platformsCreate reusable MCP servers with trusted contracts.
- Evaluation & hostingPlan release gates and cross-platform AI hosting.
- Capacity & observabilityPlan for hardware limits and end-to-end monitoring.
Elapsed
00:00
Current Streak
0 correct in a row
Leaderboard
Current rank: Not ranked yet
Set the pace for your colleagues
0 answered
0%
You did not make the Top 20 leaderboard.
Best leaderboard rank—
Completion time—
Questions answered—
Review and reinforce
Answer explanations
Use the decision rule in each explanation, not only the answer text.
Leaderboard
Ranking rules
How the leaderboard works
- Seven separate boards. Streak is a cross-role, open-ended challenge. Each role also has a Sprint (10 questions) and Marathon (20 questions).
- Streak rewards consistency. Every correct answer extends the run and updates its score. The first wrong answer ends it. Higher Streaks rank first; equal Streaks are ordered by when they were achieved.
- Only your best try is kept. Each person appears once per board. A stronger retry replaces the displayed result; a weaker retry does not lower it.
- Sprint and Marathon use score, then speed. Higher percentage ranks first. Equal scores are ordered by shorter completion time, then by the earlier completion date.
- Each board is a Top 20. Five rows are visible at once. Scroll inside a board to see ranks 6–20. Results below rank 20 are not shown.
- Podiums score for teams. Across all seven boards, gold earns 5 points, silver 3, and bronze 1. Teams are ordered by points, then gold, silver, and bronze counts. A complete tie uses the configured team order.
Your current thumbnail, cluster, and team appear with your result. You can change them in your private profile.
Loading scores…
Across every Arena role
Streak
Who can keep answering correctly for the longest?
No Streak scores yet.
| Rank | Colleague | Streak | Achieved |
|---|
Across all seven challenges
Team medal table
🥇 5
🥈 3
🥉 1
| Team | 🥇 | 🥈 | 🥉 | Points |
|---|
AI User
Sprint
No scores yet.
| Rank | Colleague | Score | Time |
|---|
Marathon
No scores yet.
| Rank | Colleague | Score | Time |
|---|
AI Engineer
Sprint
No scores yet.
| Rank | Colleague | Score | Time |
|---|
Marathon
No scores yet.
| Rank | Colleague | Score | Time |
|---|
AI Architect
Sprint
No scores yet.
| Rank | Colleague | Score | Time |
|---|
Marathon
No scores yet.
| Rank | Colleague | Score | Time |
|---|
Your learning progress
Rank
Build your AI knowledge, one level at a time.
Loading your rank…
The 0–10 scale
Rank titles describe your progress. Learning paths choose your question topics; leaderboard positions compare individual runs.
| Level & title | What this level covers |
|---|
How you move up
Everyone starts at rank 0. Your first review follows 20 qualifying answers, then another review follows every 10. Streak answers and completed Sprints and Marathons count from this release onward.
Only your first answer to each question per UTC day earns credit, whether right or wrong. Repeats still help you practise. Higher ranks need evidence from questions at that level or above.
Each promotion needs at least 20 relevant answers across 5 different questions. We assess up to 60 recent relevant answers, with at most 4 per question, and use a conservative accuracy check.
You can gain one level per review and at most one every 24 hours. Each level also needs 20 more qualifying answers in total. Your earned rank never goes down.
The accuracy check uses a one-sided 95% Wilson lower bound of at least 60%. For example, 16 correct out of 20 passes; 15 does not. This is a learning milestone, not a professional certification or a measure of practical job performance.
Arena administration
Arena administration
Loading the Arena overview…
Access
Arena administrators
Loading the question library…
Import JSON
Questions
0 totalNo questions match these filters.