Script, AI or agent?
Complete plan
To automate a task, the right tool is the simplest one that works: written rules when they are clear, AI only for fuzzy language, and an agent only with few rights, a human approval and a log.
- 3 min
- 3-question check
- English · Français
- Attestation
You will be able to
- Tell apart a rule-based script, an AI (language model) and an agent
- Choose written rules when they are clear, and AI only for free text, with a check
- Use an agent only with the fewest rights possible, a human approval before any irreversible action, and a log
The module
3 min-
Script, AI or agent? 2 min
-
Knowledge check
3 questions, pass mark 80 %. Explanations after each answer; you can try again.
-
Attestation
One per person, with the date, the score and the frameworks covered — anyone can check it online with its code.
Sources
- Anthropic, Building effective agents, 19 déc. 2024 : « finding the simplest solution possible, and only increasing complexity when needed. This might mean not building agentic systems at all »
- Moffatt c. Air Canada, 2024 BCCRT 149, Civil Resolution Tribunal (Colombie-Britannique), 14 févr. 2024 : la compagnie répond des informations données par son chatbot
- OWASP Top 10 for LLM Applications 2025 : LLM06 Excessive Agency (fonctions, droits, autonomie excessifs ; approbation humaine des actions à fort impact), LLM09 Misinformation
- ANSSI, Recommandations de sécurité pour un système d'IA générative (ANSSI-PA-102, 29 avr. 2024) : proscrire l'usage automatisé de l'IA pour des actions critiques (dont les transactions financières)
- À retenir : un agent reçoit le moins de droits possible, demande une validation humaine avant toute action irréversible (paiement, suppression, envoi externe) et journalise chaque action
Version 1 · updated October 6, 2026