What we do
We refuse AI for AI's sake. The first step is to identify a specific, time-consuming, repetitive task, then assess whether a language model handles it better than conventional automation. When it does, we build the solution with explicit safeguards and human approval wherever mistakes are costly.
-
1
Identifying use cases
We examine your processes to find where AI brings a real gain, and where simple automation would be enough.
-
2
Preparing your data
Documentation, history, catalogues: we structure the sources that will feed the model for reliable answers.
-
3
Prototype and measurement
A prototype is tested on real cases and compared with human handling, with a measured error rate.
-
4
Integration into your tools
The approved solution is connected to your existing tools: website, messaging, CRM or business software.
-
5
Monitoring and improvement
Conversations are analysed to correct imperfect answers and gradually refine the system's behaviour.
Tech stack & tools
The technologies and platforms we use for this type of assignment.
- OpenAI API
- Claude API
- Python
- LangChain
- Hugging Face
- N8N
Methodology
Prototype measured before deployment, human approval for sensitive decisions and logging of conversations for continuous improvement.
What's included
A detailed list of what you receive at the end of the assignment.
- Use case identification workshop
- Preparation and structuring of your source data
- Prototype tested on real cases
- Measurement of the correct answer rate
- Integration into your existing tools
- Safeguards and human approval on sensitive points
- Documentation and user training
- Monitoring and fine-tuning during the first month
Indicative timeline: 4 to 10 weeks depending on the use case and data preparation
Why choose us
Use cases, not promises
We only deploy an AI solution if it addresses an identified problem with a clear cost.
Your data stays under control
Together we define what can be sent to an external model and what must stay in-house.
Human oversight preserved
For anything that involves your liability, the AI suggests and a human approves.
Transparent running costs
The real cost per request is estimated from the start, to avoid nasty surprises once in use.
Our packages
Indicative prices, adjusted after we discuss your project.
Essential
from €990
First use case
- 1 targeted use case
- Assistant based on your documents
- Simple integration
- Documentation provided
- Training included
Pro
from €2,900
Business assistant
- Assistant connected to your tools
- Structured knowledge base
- Safeguards and human approval
- Conversation dashboard
- Fine-tuning for 1 month
- Team training
Custom
from €6,000
Complete solution
- Several use cases
- High-volume document processing
- In-house hosted model possible
- Advanced integrations
- Long-term support
All our prices are indicative. Contact us for a free personalised quote.
Relevant industries
The industries for which this service is most often used.
Service FAQ
It depends on the architecture chosen. Together we define what can go through an external API and what must stay on your servers. Fully in-house hosting is possible for sensitive data.
Yes, which is why we always measure the error rate on real cases before deployment, and require human approval wherever a mistake would have consequences.
Models are billed by usage, usually a few cents per exchange. For an assistant handling 500 questions a month, allow often less than €30 a month.
No, and that is not the goal. It handles simple, recurring questions, which frees up human time for the requests that deserve it.
AI processing is still data processing: it must appear in your register, be mentioned in your privacy policy and respect data minimisation. We build these points in from the design stage.