Frequently asked questions

We commit to every client reaching the break-even point on their investment within six months of deployment.

This commitment rests on a rigorous selection of the processes to automate based on their impact-to-effort ratio, deeply customised solutions that ensure a high adoption rate, and methodical post-deployment monitoring to maximise their value over time.

The return on investment shows up concretely in the time recovered, fewer errors, better responsiveness towards your customers, and the reallocation of your employees' time to higher-value work.

Our work targets the repetitive, time-consuming, low-value tasks that weigh on your teams' day-to-day.

The time savings show up at many levels depending on your business: automatic document generation (quotes, invoices, reports), automated meeting minutes, handling and managing customer requests, consolidating data across software without manual re-keying, competitive and regulatory monitoring, among many others.

The scale of the gain depends directly on your current processes. That is exactly what our on-site operational assessment is designed to quantify: we estimate the number of recoverable hours and prioritise automations by their impact-to-effort ratio.

We invite you to browse our use cases for concrete examples that illustrate our approach.

Unlike the 20-minute video-call audits that most players on the market offer, we carry out on-site operational assessments designed to grasp your working methods and real needs in depth. Every deployment comes with rigorous post-delivery support, intended to ensure adoption by your teams and the longevity of the solutions put in place.

From the design stage of our services, we build in responsible-AI and GDPR-compliance requirements, a commitment that remains rare in a market where deployment speed often takes precedence over regulatory rigour. We commit to full transparency with our clients on these matters.

Finally, because AI should not be a blind bet, we guarantee a return on investment (ROI) achieved within 6 months of the engagement.

Your business is unique, and we are keen to understand its specifics in detail so we can tailor our services accordingly. That is precisely why general-purpose AI models, however powerful, aren't enough on their own: every company has its own expertise, its own operational constraints and its own data, which these models don't natively understand.

Our job is to specialise these tools so they embed your know-how and become genuinely operational in your context. Together with you, we build custom applications that harness the power of AI well beyond a simple chatbot, to address concrete use cases specific to your organisation.

If this isn't perfectly clear, we invite you to look at our concrete use-case examples, or to contact us directly (the first conversation is free and with no commitment).

In concrete terms, we identify the repetitive, time-consuming tasks that require neither your expertise nor your judgement, to give you back the time and attention your strategic priorities demand.

The exact scope is defined case by case. The tasks to automate first depend on your business, your processes and your tools. That is precisely the purpose of our on-site operational assessment.

We invite you to browse our use cases for concrete examples that are representative of our approach.

Our solutions integrate with all of your software, whatever its vendor or age.

The orchestration platforms we use (n8n in particular) come with native connectors for several hundred common applications. For more specific software or in-house tools, we build custom integrations via their APIs (application programming interfaces), or failing that via other interfacing methods (webhooks, database access, file reading).

At the core of our setup is n8n, an automation application server that we host and that lets us design workflows triggered on a schedule or in response to external events. Claude Code is the second pillar of this setup: we use it to develop bespoke components to integrate into our automations, or full-fledged software when that approach is more appropriate than n8n. This is complemented by PostgreSQL for data storage, administered via Adminer, and by PgVector for vectorising data that needs semantic search. The architecture also includes web servers (WordPress in particular), file servers and lightweight application servers (PHP) depending on each project's needs.

Our automations connect to your existing work environment: instant-messaging apps (Slack, Telegram), email (Microsoft, Google), cloud storage (Google Cloud, OneDrive) and online files (Google Sheets). The goal is to fit into your tools rather than force a migration on you.

When it comes to artificial intelligence, we use two types of model depending on the requirements of each use case. For vectorisation, we use Nomic AI locally or OpenAI's text-embedding models online. For language processing, we deploy local models (Alibaba Cloud's Qwen3, Mistral 7B) or some of the best-performing online models on the market (OpenAI, Anthropic with Claude Opus, Sonnet and Haiku, Google Gemini, Microsoft Azure, Mistral).

No, no technical skills are required. Our solutions are designed to fit naturally into your existing processes. We deliver tools that are immediately operational, designed to simplify your teams' day-to-day rather than weigh it down. We see no point in building complex tools if teams can't make them their own.

Every deployment comes with a structured knowledge transfer to ensure a smooth handover from day one. And if you also subscribe to post-deployment support, we remain available over time to help your staff, answer their questions and resolve any usage difficulties. The aim is to reduce friction as much as possible so that adoption happens naturally, without disrupting existing work habits.

Post-deployment support covers debugging, ongoing training for your teams, the costs of AI providers and regular updates to your automations with the best-performing models on the market.

Without this service, you would have to cover the costs related to n8n (self-hosting or subscription) and the AI models used (subscriptions or API access), along with the associated technical management.

What's more, even if your technical team has the skills to handle debugging, it won't necessarily have the availability required to support users day to day or to keep up a continuous technology watch on how AI models evolve.

As specialists in these areas, we offer to bring our expertise to bear to free you from this burden and save you significant operational time.

With post-deployment support, your usage is unlimited. The monthly subscription covers all the costs related to the automation orchestrator (n8n in most cases) as well as the software used, in particular the AI models. In the event of exceptionally high usage, the applicable terms are set out in the contract signed between the parties.

Without post-deployment support, your usage remains unlimited, but the infrastructure costs (n8n, AI APIs) are at your expense.

Yes. Any other approach would be illegal and contrary to our vision.

When it comes to the protection of personal data, we act as a processor within the meaning of the GDPR: you remain the data controller and the owner of your data (T&Cs art. 13). Every engagement comes with a legally binding DPA, appended to the service contract, which sets out the nature of the processing, the categories of data, the retention periods, the security measures and the sub-processors used. You receive the full list of these on delivery and have a right to object within 15 days (T&Cs art. 13.3). You also have a right of audit, which gives effect to the accountability principle set out by the GDPR. The security of processing is governed contractually, supplemented by a five-year confidentiality obligation (T&Cs art. 12.5). In the event of a breach, we notify you within 48 hours and assist you with your obligations to notify the CNIL and the data subjects. We also assist you with requests to exercise data subjects' rights. European hosting is preferred; outside the EEA, transfers are framed by the European Commission's standard contractual clauses.

On the AI Act (Regulation (EU) 2024/1689), each automation is classified by risk level, which is communicated in the contract (T&Cs art. 14.2). We refrain from any practice prohibited by the AI Act. Use in a high-risk domain is strictly forbidden unless expressly stipulated (T&Cs art. 9.3), in which case it triggers effective human oversight, log retention for a minimum of six months and European registration where required. Transparency obligations towards end users are built in from the design stage. The transfer of skills to your teams covers the AI-literacy requirement set out by the AI Act (T&Cs art. 14.4). Complete delivery documentation, which forms part of the Deliverables, formalises this compliance (T&Cs art. 14.7).

Our privacy policy can be viewed from the bar at the bottom of the page. For any question about the protection of your data, write to us at rgpd@codesia.fr.

Under article 4.2 of our DPA, every member of Codesia who accesses your data or your clients' data is bound by a contractual confidentiality obligation. This obligation applies from the very first exchange between the parties, including the pre-contractual phase, survives the end of the working relationship and continues for five years after the contract ends (article 12.5 of the T&Cs).

No. Article 12.3 of our T&Cs commits us to favouring AI providers whose policies explicitly exclude using data sent via their API to train their models. We answer your questions about the data-handling policies of the providers we use with full transparency. And if one of them changes its policy in a way that could affect you, we let you know without waiting for you to ask.

Article 13.5 of our T&Cs commits us to favouring solutions hosted within the European Union. When a provider outside the EEA proves functionally unavoidable, we put the appropriate safeguards in place: the European Commission's standard contractual clauses, and the Data Privacy Framework where applicable. You are always informed of this in the delivery documentation.

Under article 13.3 of our T&Cs, the delivery documentation gives you the full list of sub-processors used: orchestrators (n8n, Make), AI providers (Anthropic, Mistral) and third-party services. You have fifteen days to object to any of them. Any later addition or replacement is notified to you and opens the same right to object.