Déclic
Your teams move from "AI is a vague idea" to "I use it every day".
- Data access
- None
- Lead time
- 2 to 3 weeks
None of your data is involved at this stage.
your teams already use AI daily. Go straight to the diagnostic.
From
€1,500/ day
AI audit and deployment · European SMEs
Usage audit, private AI hosted in the EU, team training. Thirty seconds to find what your teams currently spend thirty minutes looking for.
Reply within 24 working hours · First call with no commitment
100%
of the infrastructure hosted in the EU: OVH and Scaleway in France, or your own servers
Where we have shipped
70+
French SMEs supported
Since the team started
5 years
of experience on AI projects
Étienne Guimbard, founder
2
SaaS products built and operated by the team
DocAgora (healthcare), CoProFlex (property management)
0
client records sent to a US API
Written into the contract
01
The problem
§ 01
Your company
AI is already inside your walls. You did not decide it.
Your staff use AI assistants on their personal accounts, with your documents and sometimes your clients'. No one has set them a limit. In the event of an incident, the employer answers.
Your tools do it without telling you.
Your CRM, your video tool, your email have enabled AI features that send data to subprocessors you never approved. Verifiable within a day.
The subject belongs to no one. So it belongs to you.
Your IT provider keeps the machines running, not the usage, and no one has taken stock of it. What an employer is held to account for is rarely having done things badly, it is having framed nothing.
Three data flows already active. None of them was decided.
02
The offers
The more of your data the AI can reach, the more value it creates. Always within your walls.
From training to a copilot connected to your ERP. Each offer includes the previous one, your data stays with you.
Fig. 01 · Three levels of data access. Nothing leaves the enclosure.
Your teams move from "AI is a vague idea" to "I use it every day".
None of your data is involved at this stage.
your teams already use AI daily. Go straight to the diagnostic.
From
€1,500/ day
Your own AI, installed on your servers, with command of your entire company knowledge.
100 % local. Where most solutions route through a non-EU cloud, nothing leaves your walls.
your documentation fits in a few dozen files. A well configured search engine will cost less and do the job.
From
€8,000
Setup. Maintenance contract required, see § 08.
An AI that does not simply answer: it executes.
Always under your control, never fully autonomous.
your ERP exposes no API and cannot be changed. With no access to the data, the copilot has nothing to act on.
On quotation, from
€20,000
Scope defined after the diagnostic.
From diagnostic to managed services, in the order an engagement unfolds. Each milestone stands on its own, or plugs into any offer.
Your management software stays the system of record. We do not replace it: we connect to it through the access it exposes, the AI reads and writes there, and the document comes out in the screen your teams already work in.
That is the entry point, not the ceiling. Some companies stop there. Others go further: a private AI on their documents, then a copilot connected to their ERP. Those are the three levels above, and the step is decided at the diagnostic.
One condition, checked at the diagnostic: your software must expose an access. If it does not, we tell you — we then work from your exports, or not at all.
Five things that do not change when we connect to a tool you already run.
No migration
No data to move across, no switchover month, no risk of losing everything on the way: your management software keeps its data and stays the system of record.
The screen does not change
For this particular move, your teams have no new tool to learn: the document is waiting for them in the software they already open. Raising their everyday use of AI is a subject in its own right, and that is the Déclic offer.
Your official documents
Quotes, invoices and certificates keep coming out of your own tool, in your own format and under your own responsibility.
Unplug it, nothing is missing
Your quotes, your invoices and your history never left your management software.
Nothing to move before starting
We connect to what is already running: there is no switchover to prepare, and rollout times remain the ones shown above, offer by offer.
The move is the same for a quote, a contract, a certificate or a report: you describe the situation in one sentence, the document comes out written to your own formats, and it waits for you in your own software. The exact scope — one document first, the others after — is set at the diagnostic.
What you dictate
“Enquiry from the Keller group for 12 November: a twenty-person offsite, meeting room for the day, lunch and one break, two rooms the night before.”
Dictated or typed, it makes no difference.
The proposal then waits in your software. Nothing has gone out: you review it, you adjust it, and you are the one who sends it.
Fig. 02 · One sentence in, one document waiting for approval in your own software.
Time saved can be measured, and it counts. What cannot be measured, and costs more, are the errors: they appear nowhere in a set of accounts, because they are absorbed one file at a time.
And each one costs three times over: the money lost, the time spent putting it right, and what the client remembers.
A document produced from your own data, and read before it goes out, removes what these errors have in common.
03
Operations
Models evolve, your data changes, an unmaintained infrastructure degrades. Our operational support keeps it updated and supervised.
The foundation: your AI stays stable and up to date.
From
€390/ month
Continuous supervision and improvement over time.
From
€990/ month
For critical and regulated environments.
From
€2,500/ month
Hosting included on a sovereign European cloud, France by default. Or deployed on your own server.
04
Sovereignty
When your data goes to an American provider, what protects it is a signature. Here, what protects it is first of all the place where it is processed: the contract records that, it does not have to make up for it.
Everything we build runs on European Union infrastructure, or on your own servers. You choose which one during scoping: OVH or Scaleway in France, another EU provider, or your own machines. The models used by default are European.
These three points appear in the contract, together with the exit clause: models used, location, conditions for leaving. If you leave, nothing locks shut: you take the installation over, or hand it to whoever you choose.
Default architecture
Your data
quotes, case files, clients
European infrastructure
France, the EU, or your own servers
European models
by default, written into the contract
No client data to an American API, including during demonstrations.
Fig. 03 · Data processing chain
The building blocks, and where your data lives
Open, standard building blocks, each one replaceable, and it is written into the contract.
Every deliverable maps back to a text. Your inventory, your risk classification, your policy and your impact assessment are built to be verifiable by a lawyer or an auditor.
Affiliations
IAPP member · iapp.org
05
What clients say
Public reviews left by directors we supported on their websites and internal tools. Each one states the work it refers to.
From the very first contact, the team was attentive, responsive and extremely professional. The final result exceeded our expectations. They do not just "build a website", they think strategy, visibility and performance.
Their professionalism shows at every stage of the project: rigorous follow-up, attentive listening, and a real understanding of what we needed. We felt in safe hands throughout.
The support was genuinely tailored, and the result is a professional, high-quality website. What I value most is their responsiveness: adjustments are made as soon as they are needed.
These reviews cover web and internal tooling work, not AI deployments: they say the team keeps its commitments, and nothing more. Our first AI engagements are too recent to publish measured results. They will appear here as figures, not as rewritten prose.
06
The method
A consulting engagement, not a large programme. Each stage produces proof before the next one is committed. You decide at each step, on figures measured at your premises.
Diagnostic.
4 weeks
Interviews and inventory. Your policy drafted, your search costs measured. The deliverables are yours to keep.
Scoping & compliance.
2 to 3 weeks
We close the grey zone: company accounts, signed policy, teams trained, register maintained.
Pilot.
3 to 4 weeks
A restricted, reversible scope, demonstrated on public data first: measured results before any extension.
Deployment.
4 to 6 weeks
The engine in production, on your side: thirty seconds for what your teams spend thirty minutes on.
Managed operations.
Ongoing
The foundation maintained over time, and the automations that pay off, connected one by one. Priced per unit.
Fig. 04 · Engagement protocol: five stages, four decision points
07
The team
Four people, not a bench. They are the ones who design your infrastructure, deploy it, host it and come and keep it running on site, from diagnosis through to operations.

Fig. 05 · É. Guimbard, founder
Étienne Guimbard
Founder — architecture and product
Websites, ERP systems and platforms built for European SMEs: skilled trades, healthcare, real estate, hospitality. PropulAI is the logical next step, with the same sovereignty requirements.

Fig. 06 · S. Estache, implementation
Samuel Estache
AI infrastructure implementation
He deploys the infrastructure designed during scoping: installing the models, integrating them with the systems already in place, moving them into production. On your premises as on your servers.

Fig. 07 · Lyes, hosting
Lyes
Infrastructure and hosting
He builds and maintains the platform your models run on: servers, sovereign hosting, security and operations. He is the one who makes sure the infrastructure stays under your control.

Fig. 08 · Lucas, on site
Lucas
On-site presence
One day a week on your premises, included in the operational support: adjustments, working through the cases that stall, bringing your teams up to speed as usage grows.
5 years
of experience in AI
70+
European SMEs supported
4
core sectors: skilled trades, healthcare, real estate, hospitality
08
Self-assessment
Immediate answer on screen, without leaving your email address. It is only asked for if you want the written summary.
Question 1 of 8
09
Questions
The transparency obligation only concerns companies that expose a conversational agent or generated content to the public. If that is your case, becoming compliant is still quick.
And for everyone, AI literacy measures have been required since February 2025, whatever the size of the company.
10
The guide
What is already running in your company, what the law says, and the method to put things back in order in 90 days. Seven chapters, zero jargon. Get it by email.
Contents
Written for executives of SMEs from 10 to 250 employees, with or without an in-house IT department.
11
Deciding
A straight answer: what applies to you, and whether a project makes sense. If the answer is no, you leave with the answer.
Regulation (EU) 2024/1689 · End of file