30 to 50 % of AI initiatives fail to return what they should: abandoned projects, under-used licences, agents stuffed with data that burn tokens and hallucinate. A consultancy specialised in AI, Idun Group structures your AI support engagement around value: we analyse the real usage of your AI platforms with our log-analysis tooling, then we deliver strategic advice with a roadmap that says what to industrialise, what to consolidate and what to cut. Our references include LVMH, Airbus, L'Oréal and Randstad.
AI diagnostics run in a year
users measured
qualified use cases
of AI initiatives fail to return what they should (Gartner: at least 30 % of proofs of concept abandoned).
AI is already everywhere in your company: Claude, ChatGPT, Gemini, Copilot, in-house agents. What is missing is visibility on real usage: nobody knows which use cases create value, which cost without serving, nor what the whole represents in consolidated cost. Without that measurement, industrialisation decisions rest on intuition and stall at the first hard trade-off.
Subscriptions taken out then dropped, redundant tools on the same use case, inherited zombie projects: waste is structural as soon as several AI tools coexist. This is a field finding: across the estates we analyse, 20 to 40 % of the AI licences paid for go unused. Analysing real usage makes it visible, quantified and open to arbitration at renewal time.
A structured AI support engagement rests on 3 non-negotiable requirements. Without these 3 conditions, the technology changes nothing about how profitable your AI is.
Decisions taken on measured real usage, never on intuition
A roadmap prioritised by return on investment before any development
Results aggregated by team and by tool: we analyse usage, never people
30 minutes
Your AI context, the method, the likely scope: no access to your data before signature
Mandatory interviews
IT department and business units: the tooling does not replace them, it makes them factual
Logs and exports
From what you already produce, on the platforms you administer, installing nothing
Roadmap
Recommendations to industrialise, consolidate or cut, ready for the steering committee
That is what the journey looks like at Idun Group. Everything starts with a 30-minute discovery call: here are the questions it covers.
No score, no automated verdict: these questions are handled in conversation, with your context. The call opens no access to your data, pre-sales is handled by people.
The next section sets out the three blocks of the engagement that answer those questions.
Each block answers a distinct objective: measure what your teams really do with AI, decide on figures what to industrialise, consolidate or cut, then follow the effect of those decisions over time. Consulting built on tooling rather than generic training, with three principles: an isolated network is possible; results are aggregated by team and by tool, never by person; and the logs you already have are read without installing anything on workstations.
We start from the logs and exports you already produce, on the platforms you administer (Azure OpenAI and in-house models, Dust, Copilot, custom agents), installing nothing and re-instrumenting nothing. Handing over exports or a read-only connection: the exact arrangement is settled in the scoping meeting. Interviews with the IT department and the business units are a mandatory step: our usage-analysis tooling does not replace them, it makes them factual. You come out with:
All of it is delivered in a report ready for your steering committee, with explicit recommendations: industrialise, consolidate or cut. We sell neither the agents, nor the agent platform, nor the training we recommend: the analysis serves your decision, not a catalogue.
Building on the analysis, we prioritise your use cases by return on investment, effort and adoption, we arbitrate licences at renewal time and we flag the same use case found across several tools, which deserves one single implementation. The engagement runs through to the decision in the steering committee, with a roadmap. To keep the deliverables current over time, the continuous insights option (recurring re-analysis, the same deliverables kept up to date) extends the engagement.
Downstream of the advice, we tell you which use case to industrialise, which training to run and for whom; we do not do it for you. Execution stays with your teams, your integrators or a partner, and we can help you choose one. Idun Cockpit, the tool used for the analysis, can stay with you to measure the effect. The full approach is set out on our AI agency page.
How do these three blocks fit together in practice? The table below answers at a glance.
From the discovery call to the follow-up over time, each block has its format and its deliverable, with no grey area about what you receive. Nothing is committed until the analysis has proven the value.
Book a discovery callThe first conversation offered by an institutional scheme helps narrow a still-vague need. It stops there, however: no measurement of your real AI usage, no recommendation grounded in your data. Funding-support schemes complement that first conversation, without the analysis tooling that lets you decide on figures.
An institutional scheme delivers general guidance, valid for any company. At Idun Group we analyse the logs and exports of your AI platforms instead, and we deliver quantified recommendations (what to industrialise, consolidate or cut) with a roadmap. And if the analysis does not prove a return on investment, we tell you so in writing. Handing over exports or a read-only connection, results aggregated by team and by tool, never by person. And we sell neither the agents, nor the agent platform, nor the training we recommend.
The regulatory question weighs more and more in the choice of an AI support engagement, including for companies that thought they were out of scope.
The deadline for the obligations covering high-risk AI systems (annex III of EU Regulation 2024/1689: recruitment, credit, education) has been deferred from 2 August 2026 to 2 December 2027, following a European Commission proposal of 19 November 2025. The penalties remain unchanged: up to 35 million euros or 7 % of worldwide turnover. We anticipate that classification from the analysis of your AI usage onwards, as an extension of the engagement, never as its starting point.
perfume and cosmetics
An Idun engagement on the generative AI platform of all the perfume and cosmetics brands.
real activation
At this consulting and training firm, the analysis of real Claude usage established 33 users for 39 seats.
critical use cases uncovered
AI Flash Audit won through a public tender, run in its social-housing subsidiary. Engagement in progress.
The first contact with Idun Group is a 30-minute discovery call: your AI context, the method and the likely scope of the engagement. We ask for no access to your data before signature: pre-sales happens in conversation, with people.
We run scoping interviews with the IT department and the business units: a mandatory step, our tooling does not replace them, it makes them factual. The interviews give the intent and the context, the logs give the facts: it is the combination that makes the audit. The scoping meeting also settles how data is collected: handing over exports or a read-only connection.
Before the discovery call, list the AI platforms you administer (Azure OpenAI and in-house models, Dust, Copilot, custom agents). The more complete the inventory, the faster the scoping and the more the debrief reflects your real usage.
We start from the logs and exports you already produce, on the platforms you administer (Azure OpenAI and in-house models, Dust, Copilot, custom agents), installing nothing and re-instrumenting nothing. We analyse usage: results are aggregated by team and by tool, never by person, and the exact status of the processing is validated with your data protection officer.
We present the analysis on your real data: a map of usage, use cases quantified, the waste identified and explicit recommendations (industrialise, consolidate or cut), with a roadmap. You decide, on figures.
A 30-minute discovery call, scoping with your teams, analysis of the logs and exports you already produce, then a quantified debrief with a roadmap. You industrialise only what pays.