Articles, guides and field reports on AI agents

EN
Mistral or ChatGPT for business: where your data goes, plan by plan, and who uses the tool you pick
Comparisoncomparatifrgpd

Mistral vs ChatGPT: which one should your company choose?

ChatGPT keeps the lead on models; Mistral (Le Chat, now Vibe) wins the public sector and custom agents. Performance, data, GDPR, usage.

Aurélien Sabaly
20 min
EN
Cover of the Idun Group article: Claude (Anthropic) and ChatGPT (OpenAI) compared for business, as a work assistant and as a coding agent
Comparisoncomparatifadoption-ia

Claude vs ChatGPT: which one should your company choose?

Models are close, plans and data differ: what 151 companies told Idun Group about Claude and ChatGPT, and how to decide on real usage.

Geoffrey Harrazi
20 min
EN
Claude enterprise pricing: an Enterprise seat includes no usage, every token is paid at API rates
Guidecouts-ialicences-ia

Claude pricing for business (2026): subscription or tokens?

Claude Team: $20 to $125 per seat per month. Enterprise: $20 plus every token. Measured by Idun Group, the same usage would cost nearly 30 times more per token.

Geoffrey Harrazi
23 min
EN
Copilot vs ChatGPT: with an eligible Microsoft 365 subscription, Copilot Chat is already included; the choice is made on measured usage
Comparisoncomparatiflicences-ia

Copilot vs ChatGPT: which one should your company keep?

Copilot vs ChatGPT at work: what Copilot Chat includes, what the licence adds, and what we see across 151 companies.

Geoffrey Harrazi
18 min
EN
AI adoption: 85% real activation across a Claude fleet, measured in usage logs
Servicesadoption-iaclaude

AI adoption: a Claude fleet measured seat by seat

85% real activation at a consulting firm, read in its Claude usage logs. Which seats produce, which sit idle, and what the measure cannot say.

Geoffrey Harrazi
7 min
EN
AI training in companies: from one company to another, weekly Copilot use ranges from 6.3% to 70%, according to Microsoft Research
Servicesformation-iaadoption-ia

AI training in companies: who to train, and on what?

AI training in companies should be decided on actual usage: who to train, on what, what the AI Act requires, and how to measure the effect.

Geoffrey Harrazi
10 min
EN
AI ROI: 39 of 93 equipped companies do not measure how their AI is used, 14 rely on the vendor's statistics
Servicesroi-iamesure-usage-ia

AI ROI: how to calculate it from actual usage data

39 of 93 equipped companies do not measure how their AI is used. How to calculate AI ROI from actual usage, and the mistakes to avoid.

Geoffrey Harrazi
10 min
EN
Company AI policy: many policies date back to early 2023 and never caught up with the AI tool rolled out since
Guidecharte-iaconformite

Company AI policy: write it, then check real usage

An unchecked AI policy stays an intention. What it changes in how your teams use AI, and how to check each rule against what really happens.

Geoffrey Harrazi
10 min
EN
GitHub Copilot pricing: a bill multiplied by ten with usage-based billing, every request paid at the chosen model's rate
Guidecouts-ialicences-ia

GitHub Copilot pricing: what usage changes on the bill

GitHub Copilot pricing for companies: what the Business or Enterprise seat includes, what usage beyond it costs, and two settings that let the bill run.

Geoffrey Harrazi
10 min

Get our guides on the value of AI every month

Field notes on real AI usage in companies: adoption, costs, what it returns and what to do about it. No spam, ever.

1 email per month. No more.

Unsubscribe at any time from the link in every email. How we handle your data: privacy policy