Analyses, studies and field feedback on generative AI, AI agents and organizational transformation.
When employees ask AI questions instead of a colleague, the team becomes more fragile. A study by BetterUp Labs and Stanford, published in the Harvard Business Review, measures a 22,3% drop in coordination and a 21,9% drop in trust in teams where this reflex becomes the norm.
Claude Cowork or ChatGPT Work? November 2026 comparison: interface, execution in the cloud or on the workstation, connectors and plugins, scheduled tasks, deliverables, governance and price, with a method to test both in one week.
How to use Claude in a business environment in 2026: choosing between Team and Enterprise, settings to configure before granting access, setting up Claude by profession, initial use cases. Cowork, best practices and team training.
How much does Gemini cost in 2026? Prices in France for Google AI Plus, Pro and the two Ultra tiers, for Gemini in Google Workspace and for Gemini Enterprise, what each offer includes, three budget estimates and a selection grid.
How much does ChatGPT cost in 2026? Prices in euros and dollars for the free version, Go, Plus, the three Pro tiers, Business and Enterprise, what each offer includes, three team budgets with figures and a selection grid.
Ten projects to be entrusted to Copilot Cowork, profession by profession: what Cowork Go and look in Microsoft 365, what it produces, what it asks of you before acting, the delegation to copy and the trap to avoid.
How much will Claude cost in 2026? Price of the free version, Pro, Max, Team and Enterprise subscriptions, the model-by-model API, Claude Code and Coworkin dollars and euros, with usage limits and the method for choosing.
Create an AI agent in your enterprise without coding: choose the task, write the instructions, connect the sources, test and deploy. Example and test grid.
Mistral's Chat is now called Vibe: what's changing, the Work and Code modes, what the agent can do, the Free, Pro, Team and Enterprise prices, the use of your data, and how to deploy it in a company.
The AI Act already applies to any company that uses AI: prohibited practices and AI literacy since 2025, transparency since August 2026, high risk postponed to December 2027. Roles, timetable after the AI Omnibus, sanctions and seven-step action plan.
Copilot Pro no longer exists, Copilot Premium is not an offer: which is which between Copilot Pro, Microsoft 365 Premium, Microsoft 365 Copilot and Copilot Chat, with prices in France in October 2026, three budget figures and a choice grid.
Copilot Studio step by step: Agent Builder or Copilot Studio, licenses and harnesses, eight steps to create, test and publish your first agent in Teams, express version and costs. Guide from October 2026 based on official Microsoft documentation.
Free Copilot: what the consumer version of Copilot and Copilot Chat, included in Microsoft 365, offer, their limitations, six useful ways to use them without paying, and when a license becomes worthwhile. Guide from October 2026 based on official Microsoft documentation.
Copilot CoworkInstructions: prerequisites, first task in five steps, writing a good delegation, approvals, skills, planned tasks, limits, and consumption. Guide from October 2026 based on official Microsoft documentation.
It was thought that AI would free up time: the time saved is mostly returned to work. A review by the Academy of Public Policy and Economics shows that 85% of users reinvest this time in other tasks, and that the productivity gains do not automatically translate into higher wages.
With AI, we make fewer mistakes but we are less able to identify them. A study from Aalto University on 366 participants shows that AI improves performance (+7,22 points out of 40) without improving lucidity, and even widens the self-assessment gap characteristic of the Dunning-Kruger effect.
Summarize a conversation, extract decisions and deadlines into a table, then rewrite a message before sending it: Copilot in Teams tested step by step, with screenshots, 10 prompts to copy and pitfalls to know.
Summarizing an email thread, writing a response based on instructions, correcting the draft and then getting coaching: Copilot in Outlook tested step by step, with screenshots, 10 prompts to copy and the pitfalls to know.
Vibe coding: definition, tools, documented risks, what ANSSI, GDPR, and the AI Act say, and the traffic light system to allow business units to build their tools with AI without creating shadow IT. Free kit with builders' charter.
Claude Skills: what a skill is, how AI uses it, where to install it (Claude, Claude Code, Copilot) Cowork), 10 ready-to-use French skills to download, and the rules for writing and framing your own.
Claude Code in the workplace: what it does, which offer to choose, how to install it, the first day with a coding agent, what business units can do with it, and how the IT department oversees it. Guide from October 2026, compiled from official documentation.
Claude Code, Codex, or Cursor are "harnesses." What does this word encompass, how does it differ from the model and the inference server, and why does it explain why the same model gives such different results from one tool to another?
Copilot Chat or Microsoft 365 Copilot? Comparison from October 2026 based on official Microsoft documentation and a licensed account: work data, Word and Excel, templates, agents, Cowork and price in France.
Claude Code or Codex? Comparison from October 2026 based on the official documentation of Anthropic and OpenAI and the two interfaces: surfaces, models, security, customization, parallel work and price.
Gemini or ChatGPT? Comparison from October 2026 based on official documentation from Google and OpenAI: models, context, Workspace integration, delegation, multimedia and price.
Claude or ChatGPT? Comparison from October 2026 based on official documentation and three tests conducted with the same prompts in both interfaces: Excel files, sourced search, writing.
The AI Act does not mandate a single AI register for all organizations, but neither is the GDPR register sufficient. This article examines what each text requires, where they overlap, and how to maintain a shared, unduplicated map of AI usage.
Create a presentation from a Word note, add a slide, write the speaker's notes: Copilot in PowerPoint step by step, with real captures, 10 prompts and pitfalls to avoid.
Summarizing a document, rewriting a section, creating and formatting a table, writing a recommendation: Copilot in Word step by step, with real captures, 10 prompts and pitfalls to avoid.
How to use Copilot in Excel, step by step: open the pane, add a calculated column, ask a question about your data, create a pivot table and a chart, highlight rows. Includes screenshots, 10 prompts to copy, and pitfalls to avoid.
AI doesn't destroy jobs; it shifts demand towards more experienced professionals. A study by Stanford and King's College of 1,25 billion job postings shows that generative AI benefits senior managers (+6,7% hiring) and restricts entry for junior candidates.
A Stanford study dissected the history of 52 volunteers: most underestimate the extent of the judgment they delegate to AI. Oracle, complacency, perceived warmth: six mechanisms of a dependency that develops imperceptibly.
Luna, Haiku, Flash, Sonnet, Sol, Opus, Astra, Fable: what are the template tiers used for, why is the most powerful one almost never the right choice for an email or a summary, and which tier to choose for each type of office task? Includes independent rankings, publisher pricing, a printable chart, and how the choice of template affects your electricity and water bills.
According to economic scenarios from The Anthropic Institute, pay for knowledge-based jobs could fall by 11,5% by 2030, while pay for manual labor would increase by 33,6%. Six major changes to anticipate in the most extreme scenario.
Claude, ChatGPT Work, Copilot Cowork Mistral Vibe Work and Mistral Vibe Work share the same promise: to describe a result rather than ask a question. Behind this lie four very different choices: execution location, access to tools, approvals, scheduled tasks, skills, actual cost, and data. This overview was updated on September 20, 2026, based on official documentation, reflecting changes that have occurred in the past two weeks.
In eight months, Anthropic, Microsoft, and OpenAI each launched their own teammate agent. Same promise, three very different architectures: a complete comparison and selection criteria for businesses, from execution location to governance and costs.
Training, storage, hosting, connectors, GDPR and AI Act after the July 2026 Omnibus: what Anthropic and regulators say offer by offer, the five questions that come up before deploying Claude in business, and the framework that allows you to move quickly.
Ten tasks that organizations actually delegate to Claude CoworkFunction by function: what you provide, what you get, the trap to avoid and the first deliverable to start with, with the test of the three criteria to know what delegates well.
Most professionals use generative AI reflexively: a question, an answer in ten seconds, then they move on. This mode suits the vast majority of uses. But it overlooks a completely different kind of capability: Deep Research. For five to thirty minutes, the AI consults dozens of sources, cross-references them, revisits its hypotheses, and delivers a…
In any organization, a crucial part of the value lies in know-how that isn't written down anywhere: the way your best lawyer reviews a contract, the process your sales director follows to qualify an opportunity. These methods reside in a few minds. When those minds leave, the methods leave with them. "Skills," a recent feature of AI assistants, offer a way…
For two years, the key skill in generative AI was prompt engineering: the art of formulating the right query, with the right role, the right format, and the right examples. This skill remains useful. But it is no longer the primary driver. The decisive question has shifted: it's no longer "how do I formulate my request?" but "what do I give the model to read?" It's about context…
The Model Context Protocol (MCP) is the standard that connects AI to your business tools: CRMs, calendars, document databases, and websites. Launched in late 2024 by Anthropic and subsequently adopted by the entire ecosystem, this standard plays the same role for AI as the USB-C port does for our devices: a universal connector that replaces a multitude of proprietary cables. The problem with MCP…
Generative AI promised to free us from tedious work. Two years after its widespread deployment in businesses, studies converge on a disturbing finding: when poorly integrated, it intensifies work instead of simplifying it. An analysis of a phenomenon still…
Article published in June 2026, updated on September 27, 2026: Cowork now runs in the cloud and has merged with the chat. Memory usage, scheduled tasks, Dispatch, and permission modes have been updated.
Difference between Copilot Cowork and Claude Cowork Article published in June 2026, updated on September 27, 2026: Cowork It now runs in the cloud and has merged with the chat. Claude's presentation Cowork and Copilot […]
Claude Cowork as a deliverables factory. Article published in June 2026, updated on September 27, 2026: Cowork now runs in the cloud and has merged with the chat. Memory usage, scheduled tasks, and […]
Cowork now runs in the cloud, launches from the web, desktop and mobile, and has merged with chat since September 16, 2026. This rewritten guide details the prerequisites, /setup-claude, folders and projects, the new memory, the three permission modes, the integrated browser, scheduled tasks and the NextStart method for building a workspace rather than a prompt.
In the race for AI, most companies focus on the tools but forget the human element. Yet, an organization's psychological infrastructure is more fragile than its technological architecture. The result: demotivation negates the expected gains. This is the cost […]
Should we ban AI? Wrong question. What we need to change is how we use it. AI makes us faster, but if we're not careful, we risk transferring to the machine the mental work that makes us competent. This phenomenon […]
We often hear that AI frees us from tedious tasks, allowing us to engage in more strategic thinking. However, by systematically delegating problem-solving to machines, we reduce the cognitive processes essential for learning. ⚡️The […]
Which jobs will become automated, reorganized, grow, or evolve less than expected? We talk about jobs "exposed" to AI. This is a starting point, but it's insufficient. The real question is: Which jobs will become automated, reorganized, grow, […]
Generative AI is silently locking the door to the job market. By analyzing the pay stubs of millions of US workers, Stanford researchers have identified the infamous "canaries in the coal mine." And […]
Until now, it was thought that the danger of AI stemmed from its "hallucinations." But an MIT study highlights a different peril: AI is too polite, and this is causing us to lose touch with reality:1️⃣ The […]
LLMs have monopolized our attention in recent years. But according to Yann LeCun, the AI revolution is only just beginning… The goal? To move from a “generative” AI that recites to a “planning” AI that understands and acts in the real world: 1️⃣ The […]
Will AI destroy or transform our jobs? To move beyond mere speculation, Anthropic has just published a study that measures the actual adoption of AI by profession. Instead of limiting themselves to what AI could theoretically accomplish, they cross-referenced this with real-world usage […]
Arthur Mensch, CEO of Mistral AI, offered a clear-eyed assessment of AI and its implications during his presentation at École Polytechnique: 1️⃣ The commoditization of technical expertise. Software development as we learned it 10 […]
The promise of generative AI is enticing. Yet, some organizations are observing a surprising phenomenon: the "Productivity Paradox." Instead of reducing mental workload, AI tends to intensify it! 1️⃣ The myth of workload reduction: We think we're saving time, […]
Mistral AI is financing a 44 MW datacenter in Bruyères-le-Châtel, equipped with 13,800 Nvidia GB300 chips, through debt financing (€722 million, or $830 million). Ten reasons why this project matters for European digital sovereignty.
RAG (Retrieval-Augmented Generation) makes AI respond based on your documents: definition, operation in five steps, comparison with fine-tuning and long context, use cases, limitations of RAG and method to make it reliable in business.
Agentic AI refers to systems capable of achieving a goal with limited supervision: they plan, use tools, and correct their work. Definition, differences from generative AI and automation, levels of autonomy, industry-specific examples, tools for 2026, and governance.
From experimentation to mass industrialization of artificial intelligence: For more than a decade, artificial intelligence was seen as a fascinating but still largely experimental technological promise. Spectacular demonstrations, impressive prototypes, occasional productivity gains… but few truly structuring deployments […]
In recent years, artificial intelligence has emerged as a major driver of transformation for organizations, public services, and society as a whole. Yet, behind the technological enthusiasm and promises of efficiency, a central question too often remains relegated […]
2026: From Promise to Real Architecture. After several years of intense experimentation, 2026 marks a decisive turning point for artificial intelligence in business. The era of spectacular demonstrations and marketing announcements is gradually giving way to a more […]
Does AI threaten our critical thinking? As artificial intelligence tools invade our daily lives, from virtual assistants to search engines, a crucial question arises concerning the future of our cognitive abilities. Recent research conducted […]
AI in Healthcare: Conversational AI has become an essential ally for millions of Americans facing healthcare challenges. More than 5% of global conversations are about medical issues, with approximately 40 million users […]
AI, an opportunity for equal opportunities. For the first time in the history of work, a technology is not necessarily widening the gap between those who master it and those who do not. Generative AI stands out for its ease of use. […]
Artificial intelligence promises to profoundly transform the healthcare sector. From assisted diagnostics and personalized virtual assistants to the optimization of hospital operations, the potential applications seem endless. The AI market in healthcare is expected to reach 491 […]
Artificial intelligence has spread rapidly within local authorities, creating a striking contrast between the widespread interest and the cautious approach to its deployment. Many employees are already using AI tools, often outside of established frameworks, while organizations are seeking to structure these practices. […]
Conversational artificial intelligence is no longer just a tool. It has become a mental space we visit throughout the day, and the analysis of millions of interactions reveals a surprisingly human map of our habits. Two screens, two […]
The true cost of automation: monitoring and verification. AI agents promise to revolutionize office work, but the reality is more nuanced than a simple human-machine replacement. Understanding how agents and humans work differently is essential for […]
AI is no longer just a chat assistant. It's becoming a layer for executing work. More than half of organizations are now deploying agents capable of managing multi-step workflows, and 16% are even orchestrating cross-team processes. The […]
Artificial intelligence is no longer just a matter of algorithms and performance. It has become an infrastructure issue, with considerable physical impacts that are reshaping economic competition. Explosive electricity consumption: Data centers dedicated to AI are radically changing […]
AI agents are no longer just chatbots. They are becoming digital colleagues capable of performing complex tasks: analyzing data, producing reports, creating presentations, or generating code. But for businesses, the real question […]
France faces the challenge of generative AI. Artificial intelligence is no longer a laboratory subject. It has established itself as a general-purpose technology affecting all sectors, and is therefore becoming a matter of public policy as well as […]
Prompt engineering explained simply: the structure of a good prompt in five components, seven techniques with a before/after example, what changes with reasoning models, frequent errors and a method to test your prompts in business.
Cognitive Dependence and the Ecological Cost of AI: Reducing generative AI to a mere tool obscures two essential questions: its material composition and its impact on society. This simplistic view prevents us from seeing the systemic effects that […]
The French Generation Z has just reached a historic milestone: barely into their working lives, they have already embraced generative artificial intelligence tools. Nearly nine out of ten young people say they have tested them at least once […]
AI is finally gaining ground in French SMEs. The wave of artificial intelligence currently sweeping through the French business landscape is no longer an abstraction: a field survey conducted in early 2025 among 1,209 small and medium-sized enterprise managers reveals […]
Questioning the necessity, sizing the model as needed, measuring and publishing the impacts, limiting retraining, managing data, evaluating social effects: this is what the State's responsible purchasing guide for AI solutions (December 2025) requires, verified figures, and what a private company can learn from it.
In lecture halls as well as virtual classrooms, the arrival of generative AI has revealed a striking contrast: on the one hand, students spontaneously using chatbots to write, summarize, or explain; on the other, teachers who are experts in their field.
By format
By profession
By AI tool
Aeronautics and space
Insurance and social protection
Health and life sciences
Public sector and research
Assess yourself
Job-specific kits
Claude's Kits
Copilot Kits
Get the teams on board
Compliance and security
