← Services

AI automation and n8n

A quick first use case on n8n, then a custom automation you control

Gensai designs AI automations in two stages: a first use case (POC) set up quickly with n8n to validate the idea on your real data, then, once this first use case is validated, a custom-built version for production. In both cases, a workflow reads an e-mail or a document, extracts the information, queries your tools, has an AI model draft or classify, then passes the result to the right person, with data hosted in France or on your servers.

Classic automation connects applications through fixed rules: a form comes in, a row is added to the spreadsheet. AI adds what rules cannot do: understand a freely written e-mail, read an attachment, summarise an exchange, sort a request by urgency. The workflow stays in charge of the sequence of steps, and the model only steps in where it adds something.

Gensai mainly uses n8n to set up these first use cases. The tool connects to hundreds of applications (CRM, messaging, calendar, storage, databases), accepts code when a step calls for it and offers AI agent nodes that call Mistral, Claude or OpenAI models. A workflow thus runs quickly on your data, and it can be hosted on your premises: the n8n community edition is free for internal use, under a so-called fair-code licence.

Once the first use case is validated, Gensai prefers custom development for industrialisation: code designed for your process alone reduces the dependence on external tools, their licence, their pricing and their updates, and limits potential security flaws, with fewer exposed components and connectors. Every automation ships with its guardrails: access limited to what it has to do, human validation before any sensitive action and a log of every run. And when the tools you already have are enough, such as Power BI for reporting, Gensai relies on them.

Discuss an automation

Examples of AI automations

Sorting incoming requests

Each e-mail or form is read, sorted by topic and urgency, then passed to the right person with a summary.

Customer reminders

Unanswered quotes, pending invoices, incomplete files: reminders prepared from the CRM and validated before sending.

Data entry from documents

Invoices, purchase orders, documents in a file: the useful information is extracted and entered into your tools.

Meeting minutes

A meeting transcribed, summarised and shared, with the decisions and the follow-up actions.

Assistant on WhatsApp or Teams

A chatbot answers common questions from your information, and a workflow frames each answer with control steps.

Monitoring and alerts

Sources followed every day, sorted by a model, and a summary sent to Slack, Teams or by e-mail.

HR and administrative reporting

Dashboards fed automatically, without consolidating figures by hand every month.

Content enrichment

Product sheets, customer records or descriptions completed by a model, then proofread before publication.

Our references

Frequently asked questions

What is n8n?

n8n is a workflow automation tool: it connects applications and chains steps (trigger, processing, condition, action) in a visual interface. It includes AI nodes that call a language model or drive an agent, and accepts code when a step calls for it. Its main asset for a business: it can be hosted on the company's own servers.

Is n8n open source?

Not strictly speaking. Its code is public, but its licence, the Sustainable Use License, is described as fair-code: it allows internal use, including free self-hosting of the community edition, and restricts reselling or hosting n8n for third parties. To set up a first use case on its own processes, a business usually finds the community edition enough.

What is the difference between n8n, Make and Zapier?

All three connect applications without coding everything. Make and Zapier are online services, easy to pick up and billed according to the volume processed. n8n is better suited to complex workflows, accepts code and can be hosted on your premises, which matters as soon as the data is sensitive. For first use cases, Gensai favours n8n, not least because it can be hosted in France.

Automation or AI agent: which to choose?

A workflow follows a path defined in advance, which makes it predictable and easy to control. An agent chooses its own steps, which makes it more flexible but harder to frame. For a repetitive, well-defined task, a workflow with an AI step is often enough. An agent is justified when requests vary too much to be anticipated.

How much does an AI automation cost?

As a guide, at Gensai, a first use case on real data costs between €3,000 and €5,000 excluding VAT and takes 2 to 3 weeks. A first version teams can use costs between €5,000 and €10,000 excluding VAT. Hosting and model usage come on top and depend on the volumes processed.

Does our data stay in France?

It can. Both the first use case on n8n and the custom-built version are installed on your servers or with a French hosting provider, and the AI steps can rely on models hosted in Europe or installed locally. The choice is made with you at the architecture stage, step by step.

Why not keep n8n in production?

n8n makes it possible to validate an idea quickly. For production, Gensai prefers custom development: it reduces the dependence on an external tool, its licence, its pricing and its updates, and it limits potential security flaws, since only what is needed is exposed. Every run is still logged, and sensitive actions remain subject to human validation.

Where to start?

With a frequent, time-consuming and well-defined task: sorting requests, data entry from documents, reminders, reporting. A 30-minute call is enough to check whether it lends itself to automation, then a first use case on n8n, in 2 to 3 weeks, measures the gain on your real data.

A repetitive task to hand over to AI?

Describe it in a few lines, with the tools it involves: Gensai tells you whether it can be automated, with which tool and for what budget.

Subject: AI automation project

Information used only to answer the request.