Our experts in
Artificial Intelligence
bring your projects to life

Monthly access to a team that already exists, not a number of days. You know who is in your squad, how fast it delivers, and you follow what it spends as it happens. The month ends with an invoice you knew in advance and software in production.

If you want to think in day rates to compare, the dashboard shows delivered days too, but that is not the unit we sell.

One month’s notice, no penalty. There is nothing to bring back: from day one we work in your repository and on your infrastructure, and the rights transfer to you as we go. The Squad Scale plan adds an annual reversibility rehearsal, handover documentation included.

Yours. Our developers are contributors within your organisation, under the same review rules as your internal teams. If you do not have infrastructure yet, we set it up in your name, with the host you choose.

In France or the European Union by default. For sensitive processing, we can deploy open-weight models on your own infrastructure: the data then never leaves your perimeter, and no third-party supplier reaches it.

We never put production personal data into a training or test set. That rule follows the recommendations published by the French data protection authority (CNIL) on 19 June 2025 on the development of AI systems.

No, and nobody honest would. The question behind it is usually: “how do I know I am not buying hot air?” We commit to the make-up of the squad, the delivery rhythm, availability and time to restore service. The velocity actually observed appears sprint after sprint in your dashboard: after three sprints it becomes predictable, and that is what we plan on together.

Included in all three plans, at different intensities: one scoped use case per quarter on Squad Start, a continuous line on Squad Product, a dedicated workshop on Squad Scale. Inference costs for third-party models are passed on to the euro: they appear in the AI view of the dashboard, with the cost per request.

We always start with a costed assessment. Sometimes we conclude that plain automation does the job better for ten times less, and we say so.

It depends on the use, and the timetable matters. Since 2 February 2025 the prohibitions and the AI literacy obligation have applied. Since 2 August 2026 the regulation applies generally, and article 50 requires transparency: a conversational agent must say that it is an AI, and synthetic content must be marked along the lines of “generated by artificial intelligence”.

The obligations for the high-risk systems in annex III, on the other hand, will not apply until 2 December 2027, a deadline pushed back by regulation (EU) 2026/1744, in force since 27 July 2026. Be wary of anything published before the summer of 2026: it still gives the old date.

The requirements from the European Accessibility Act have applied since 28 June 2025, but to a defined scope: e-commerce, banking services, transport, audiovisual media, telecommunications, e-books. Companies with fewer than ten employees and turnover below €2m are exempt, and a transition period runs to 28 June 2030 for certain services already on the market.

In other words: it does not apply to everyone, and we will tell you if it does not apply to you. That said, we ship our interfaces at level AA of the standard whatever happens; the accessibility check is part of what runs on every release.