Skip to Content
AI in production, one analysis a day

AI in production, one analysis a day

Every day, a short analysis to understand what makes an AI agent hold up — or fail — in an enterprise: conditions, costs, controls, decisions.

One lesson per video, added automatically as soon as a new analysis is published. French and English subtitles in the player.

Responsible TAMPOLLA Serge Destin
Last Update 11/02/2026
Completion Time 2 hours 29 minutes
Members 1
  • The webinar announcement
    2Lessons · 4 min
    • The 8 December webinar: what you will learn
    • Le webinaire du 8 décembre : ce que vous allez apprendre (anglais)
  • Analyses
    12Lessons · 1 hr 23 min
    • AI agents in production: the 3 conditions
    • AI agent: 3 cost lines the quote leaves out
    • Your agent executes. Who authorised it?
    • The model is not the first question
    • Your three data questions: quality, access, rights
    • AI agent: the 4 log lines your auditor will demand
    • AI agent: 4 rights nobody wrote down
    • AI agent: 3 actions your agent will never approve itself
    • Data sovereignty: 3 questions, not a slogan
    • A proof of concept that decides, not one that demos
    • On-prem LLM or API: 3 things decide
    • AI total cost of ownership: 3 cost lines, not a price
  • Short formats
    40Lessons · 1 hr 3 min
    • What an AI agent is not allowed to touch
    • What is not logged should not run
    • Human validation is not temporary
    • Three costs an AI agent quote leaves out
    • No figure holds without a scope
    • An assistant writes an email. An agent writes in your ERP.
    • What you pay every month after go-live
    • Four stages, only one is about the model
    • The project budget does not cover year one
    • What the agent reads is not current
    • A service account reads everything
    • Four lines, and the scope exists
    • The rights stay after the project ends
    • What the log gives, what the audit asks for
    • Three mistakes that cost you the audit
    • What a signature must contain before it validates
    • What turns a human validation into a formality
    • The split that protects your data
    • What you take with you the day you leave
    • Hosting your data is not data sovereignty
    • Three confusions about data sovereignty
    • A total cost can be counted. Value cannot
    • Scoping, bounded pilot, contested acceptance, handover
    • The pilot that demos decides nothing
    • IT, business, security: three signatures
    • 7 roles, 6 refusals: where an AI agent acts
    • An assistant proposes, an agent writes
    • HR, production, reporting: the condition that holds
    • CIMA reporting: what AI prepares, what stays signed
    • What an agent never signs
    • Three signatures a tool cannot replace
    • Telecoms: the cause before the complaint
      New
    • Your alarms see the symptom, not the cause
      New
    • An agent warns; it does not correct
      New
    • Utilities: the unpaid bill starts at the meter reading
      New
    • Meter reading and dunning never talk
      New
    • Three roles, one file
      New
    • Import CEMAC: 5 documents, 3 gaps that stop the file
      New
    • The bill of lading says one thing, the invoice another
      New
    • What the agent never signs
      New