+33 (0)1 87 66 00 65 · Monday to Friday, 9am–6pm Free audit (15 min)
● Business offer — DUERP & risk prevention

The single risk assessment document finally kept current, work unit by work unit

France's single occupational risk assessment document (DUERP) is required of every employer from the first employee, and it is worth what its updates are worth, not its existence. Your agent keeps the assessment current work unit by work unit, links every rated risk to the dated fact behind it, follows prevention actions through to closure and warns on deadlines. Hosted in France, locally or on an isolated resource. The employer approves.

Hosted in France Health and safety data protected GDPR & AI Act: governed deployment The employer approves

Updated on

Deployed in a few weeks
DUERP & prevention assistant · hosted in France
Our risk assessment document is two years old. Where do we start?
With the gap, and I have already measured it: your document describes 5 work units, your company now has 7.
What changed since the last version: a finishing workshop opened in March, a cutting machine replaced, and on-site fitting work that did not exist before.
What your own records say: 3 accidents with time off and 61 shop-floor reports over the period, of which 22 concern a unit the document does not describe.
My proposal: rebuild the 7 real units, rate each risk against the dated fact behind it, and hand you the whole thing ranked.
⛓ Sourced · current assessment document, accident register, shop-floor reports
And the update — is it every year?
In your case yes: the at-least-annual update applies from eleven employees; below that threshold it is still required on any change to working conditions or as soon as new information about a risk appears. You have 28 employees: both apply, and the second is the one that fires most often.
What I do continuously: I watch for the events that call for an update — new machine, new substance, new post, accident, report — and hand you the modified version, never overwriting the previous one.
What is yours: the approval. The employer approves: it is their responsibility, and it is not delegated to software.
✎ Framework · art. R4121-2 of the Labour Code — annual update from 11 employees
Local inference · no data outside the EU
Data hosted in France
Sovereign by designLocal inference or hosting in France
GDPR & AI Act: governed deploymentTraceability & human oversight
TurnkeyDesigned, installed and operated for you
The employer approvesThe agent prepares, never commits in their place
✦ In brief

A Blue Lemon Agent prevention agent keeps your single occupational risk assessment document current work unit by work unit, links every rated risk to the dated fact behind it — accident, near miss, measurement, shop-floor report —, proposes ranked prevention actions and follows them through to closure. It prepares prevention plans when outside contractors intervene, warns on deadlines and keeps every successive version of the document without ever overwriting the previous one. Hosted in France, locally or on an isolated resource, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity: your employees' health and safety data stays with you. The employer approves: risk assessment engages their responsibility, never a piece of software's. Live in a few weeks.

100 %
hosted in France in the target architecture
0
transfer outside the EU in the target architecture
8
use cases ready to deploy on this scope
0
decision taken without human approval

Benchmarks describing our offer, not results measured at a client. The scale of the gain is confirmed by a pilot on your own scope.

The context

A document required from the first employee, worth what its updates are worth

The single risk assessment document is required of every company as soon as it employs one person. What protects is not that it exists but that it matches the real work: a frozen document describes a company that no longer exists, and that is exactly what an inspection or a dispute brings out.

! The stakes

The document is almost always written once and then forgotten: nobody has the time to reopen it for every machine replaced, every post created, every near miss reported. The shop-floor reports do exist — in emails, minutes, workshop logbooks — but they never reach the document. Handing that material to a consumer tool would also expose occupational health and safety data to a third party often hosted outside Europe and subject to the Cloud Act.

Our answer

The agent does the work nobody has time for: it rereads your records, ties every rated risk to the dated fact behind it, brings the work units back into line with the real organisation, proposes ranked actions and follows them through to closure. It keeps every successive version without overwriting the previous one, which the law requires over a long period. Local inference or an isolated resource hosted in France: your employees' data does not leave the company. The employer approves — risk assessment engages their responsibility.

The decisive point

Occupational health and safety data: sovereignty & compliance

Accidents, fitness restrictions, exposures, named reports: prevention handles sensitive information about identified people. Here is how our architecture protects it.

Local inference

The agent can run on a machine inside the company: no minutes, no named report leaves the network.

Hosting in France

Otherwise, a dedicated and isolated resource hosted in France under French law — your data: processing and access within the European Union targeted by the architecture.

Reduced extraterritorial exposure

Occupational health and safety information: architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity, even with a US provider hosting in Europe.

Isolated resource per company

No pooling: an environment strictly dedicated to your company and its work units.

Minimisation and role-based access

The document describes posts, not people: named material stays walled off, access is role-based (RBAC), with encryption and logging.

AI Act: governed deployment

Strictly a support agent; no assessment approved automatically; traceability and human oversight end to end.

What depends on the architecture chosen These points are not general guarantees: they are settled deployment by deployment, in the quotation.

  • The applicable location is that of the architecture set out in the quotation and verified before commissioning.
  • Local execution is announced only for the configuration explicitly described and accepted in the quotation.
  • The applicable isolation depends on the deployment mode set out in the quotation; no dedicated isolation is presumed.
  • Roles and permissions are configured and accepted for the identities and systems actually connected.
  • The events logged, their content, their retention period and who may access them are defined for the deployment chosen.
For the most sensitive data, SecNumCloud and HDS options are available depending on your requirements. A single architecture is designed to answer both the GDPR and extraterritorial exposure. Designed for deployment in line with the GDPR and the AI Act, after the processing, roles and context-specific risks have been assessed.
Demonstration

See the agent at work

4 real situations, taken from those that come up most often. Pick one: the exchange unfolds as it would in your organisation.

A scripted demonstration. These exchanges show how the agent behaves — its sources, its refusals, what it leaves to your teams. Nothing is sent from this page, no model is queried here, and the matters named are fictional. That is precisely what we promise your data.
The behaviours shown here — monitoring, automation rules, routing and reminders — are configured with you during deployment, from your tools, your rules and your thresholds.
The architecture points named in these exchanges — location, local execution, isolation, encryption, role-based access, logging — are not a guarantee attached to the demonstration: they are those of the architecture set out in your quotation, and verified before commissioning.

The company in this demonstration

Fictional company

Ateliers Marceau — industrial joinery and shopfitting

Sector
Industrial joinery and shopfitting, workshop manufacture and on-site fitting
Headcount
28 employees, of whom 19 in the workshop, 5 fitters, 1 workshop manager in charge of safety
Clients served
Retail chains and interior architects, sites across the region
Pace of the work
7 real work units, 3 accidents with time off and 61 shop-floor reports over twelve months
Tools in place
Shared folders, accident register, workshop logbook, safety data sheets for products
Who decides
The owner approves the assessment document and decides between actions; the workshop manager reports from the floor
Room for improvement
Document unchanged for 2 years, 5 units described against 7 real ones, 22 reports attached to nothing

The company did its risk assessment properly… two years ago. Since then a finishing workshop has opened, a machine has been replaced and on-site fitting has become an activity in its own right. Shop-floor reports do exist — workshop logbook, emails, meeting minutes — but nobody has the time to attach them to the document. The agent is connected to the shared folders, the accident register and the safety data sheets. It works continuously; the employer approves.

This company, its figures and the exchanges that follow were invented for the demonstration. They illustrate a common situation; they describe no real client.

DUERP & prevention assistant · hosted in France
I have reread your assessment document and your last twelve months of records. Your document describes 5 work units; your company now has 7. And the gap is not administrative: it sits exactly where it hurts.
The two units absent from the document:
· Finishing and varnishing workshop, opened in March. 14 of the 61 shop-floor reports concern it, nine of them about solvent vapours and extraction.
· On-site fitting, now an activity in its own right with 5 fitters. 8 reports, and one of your three accidents with time off happened there — a same-level fall on a cluttered site.
What changed inside the units already described: the panel saw was replaced in January — the document still describes the old one, with a guard the new one does not have — and two adhesives were reformulated, safety data sheets to prove it.
What that adds up to: 22 of the 61 reports attach to no line of the document. In other words, a third of what your employees report goes nowhere.
My proposal: I rebuild from the 7 real units, rate each risk against the dated fact behind it, and hand you the whole thing ranked. You approve unit by unit, not in one block — a document approved in one block is never read. gap_assessment-document-and-shop-floor.pdf2 units missing, 22 orphan reports
⛓ Sourced · 2-year-old assessment document, accident register, workshop logbook, 61 reports, safety data sheets
We have 28 employees. Do we redo it every year, or only when something changes?
Both, and the second will concern you far more often.
What the Labour Code says: the update is at least annual from eleven employees; below that threshold it is no longer required every year. But in every case it is required on any significant change affecting health, safety or working conditions, and as soon as further information relevant to assessing a risk comes to light. At 28 employees, both bind you.
What that means here, concretely: opening the finishing workshop in March was a significant change — the update was due then, not at the annual date. The same goes for the new saw and for the two amended safety data sheets.
What I do so it does not happen again: I watch continuously for the six events that trigger an update in your company — new machine, new substance, new post or activity, accident, repeated reports on one subject, amended safety data sheet. On each trigger I open an update proposal and put it to you. It is one of only two things I do without asking, and it only opens: nothing enters the document without your approval. The employer approves.
✎ Framework · Labour Code, art. R4121-2 — annual update from 11 employees, and on every trigger
Go ahead, rebuild the 7 units.
7 work units brought back into line with the real organisation, and I will tell you how I drew the boundaries — because the boundaries govern everything else.
The principle I applied: a unit groups employees exposed to the same risks, not those carrying the same job title. Your two forklift drivers therefore moved to « handling and storage » rather than staying in « workshop », because their exposures are those of storage: vehicle movement, loads at height, ambient noise.
The 7 units retained: cutting and machining · assembly · finishing and varnishing · handling and storage · on-site fitting · design office · administration.
What the boundaries change, measured on your own reports: under the old split, 22 reports found no line. Under this one, all 61 attach, and three attach to two units — precisely the interference risks between workshop and handling.
The point I put to you rather than settle: should « on-site fitting » be split in two depending on whether the site is occupied or empty? The exposures differ markedly — co-activity, public movement — and your 8 reports split 6 against 2. Two units describe better; one unit is easier to keep current. That is an organisational call: it is yours. work-units_seven-units.pdf61 reports out of 61 attached
⛓ Sourced · organisation chart, site schedules, 61 reports attached one by one
Local inference · no data outside the EU

Your case is not here? That is exactly what a 15-minute conversation is for. Book the free audit

Use cases

What the agent takes on in prevention

Each use case is an agent we deploy. All of them work in support: the employer approves.

Included in your agent The 6 capabilities essential to this promise are included, at no extra cost.
From 668 € excl. VAT / month

Work units kept current

Bringing the described work units back into line with the real organisation: posts created, new activities, machines replaced.

Ratings backed by facts

Every rated risk links to the dated fact behind it: accident, near miss, measurement, shop-floor report, safety data sheet.

A tracked action plan

Ranked actions, an owner and a date for each, a reminder as the deadline nears, and closure on evidence.

Prevention plans

Preparing the prevention plan when outside contractors intervene, including the cases where a written plan is mandatory.

Deadlines and versions

Alerts on update deadlines and retention of every successive version, the previous one staying readable.

Document memory

Finding a safety data sheet, a machine manual or an old set of minutes, through a knowledge agent.

Controls and safeguards These 6 controls are built into the agent: they frame what it does, whatever plan you pick. They are not chosen and are not added to your order.
Human validation, exceptions and escalation Status, safe closure and audit trail Work from a versioned corpus with citations and the law as it stood on a given date Preserve confidentiality, compartmentalisation and access logging Manage deadlines, versions, evidence and human validation Flag uncertainties and reserve advice, decision and signature for the lawyer
The gain

How much time does prevention take today?

By taking on fact gathering, drafting and action tracking, the agent moves the work towards decisions and the shop floor. The scale of the gain depends on your headcount and number of work units, and is confirmed by a pilot.

Full update of the assessment document
Today · done by hand
Prepared by the agent, to approve
Gathering and attaching shop-floor facts
Today · done by hand
Continuous, sourced
Following actions through to closure
Today · done by hand
Reminders and closure evidence
Qualitative, non-contractual comparison: the proportions shown illustrate work moving towards decision-making, they represent no measurement. Every output of the agent is read and approved by a competent person.
How it works

The stages of your AI agent project

1

Audit & scoping

15 minutes to target the use case with the best return.

2

Quote or direct sign-up

A catalogue offer is bought online; a specific need gets a costed quote.

3

Design

We design the agent and its guardrails.

4

Integration & testing

We connect your tools to the agent, which is itself hosted in France.

5

Rollout

Going live and training your team.

6

Operation

Continuous supervision and improvement.

Pricing

A prevention agent, installed and operated for you

An agent that holds the assessment document, the action plan and prevention plans, connected to your files and your action tracker. Prices excluding VAT — annual subscription, the time it takes for the gains to settle.

Agility

Setup + controlled subscription

7,320 € excl. VAT setup
then 668 € excl. VAT/month — you invest at installation and pay a reduced subscription. Ideal for keeping the cost under control over time.
  • Installation, configuration and training for your teams
  • Operation, human oversight, updates and support
  • Sovereign hosting in France, a dedicated and isolated resource
Order →
The simplest Serenity

All inclusive, no setup fee

1,078 € excl. VAT /month
all inclusive, immediate start. No upfront investment: a single subscription. Ideal for starting quickly and simply.
  • Setup included (installation, configuration, training)
  • Operation, human oversight, updates and support
  • Sovereign hosting in France, managed end to end
Order →
100% Sovereign

On site, you own it

11,405 € excl. VAT setup
then 888 € excl. VAT/month · + hardware from 2,491 € (one-off purchase, in addition) — a sovereign computer installed on your premises, maintained remotely. Models run locally, your data returned at the end of the contract. 36-month commitment.
  • Hardware installed on your premises (you own it)
  • French / European AI models run locally
  • Secure remote maintenance (Pro support included)
Order →
Not included in the packages: AI consumption (model tokens), re-invoiced at real cost with no margin, and tracked in real time in your client area. Maintenance and supervision subscription for an initial term of 12 months for the Agility package, 24 months for the Serenity package and 36 months for the 100% Sovereign package, renewable; support levels (SLA 72 h / 24 h / 4 h) optional. Bespoke development, additional integrations or exceptional volumes are quoted separately. Support Monday to Friday, 9am to 6pm. Prices exclude VAT.
AI model: none of the AI models offered currently carries a fixed surcharge. When the selected model carries a cost, that cost is shown when you choose it, before you order, and re-invoiced at the cost incurred, with no mark-up; usage is billed at the publisher's price. Publishers' prices are published in US dollars: the amount re-invoiced is the amount in euros actually borne by Blue Lemon Agent on the publisher's invoice, at that invoice's exchange rate, with no commission or mark-up.
Included components and additional components Components included in the base offer: the Blue Lemon Agent software foundation, the AI models listed in the order journey, the standard channels (Microsoft Teams, Slack, WhatsApp Business, email, website chat, calendars, Microsoft 365 / Google Workspace, file storage, market VoIP telephony, professional social-media pages and accounts, Google Business Profile), hosting in France for the package chosen, backups, supervision, updates and support. If adapting the AI agent to your constraints, your needs or your requests requires other paid components — a third-party publisher's software licence, paid API access to one of your applications, hosting of health data, for which French law requires an HDS-certified host (art. L. 1111-8 of the French Public Health Code), SecNumCloud-qualified hosting, a speech synthesis service, particular hardware —, they are offered to you as an option or on quotation and re-invoiced at the cost incurred; nothing is committed without your written agreement. Where the artificial intelligence model you choose entails an additional cost, that cost is shown to you before you order and re-invoiced to you at the cost incurred, with no margin.
What to expect
Go-live 2 to 3 weeks
Agent designed, channels connected, team trained.
Steady state 4 to 7 weeks
After a few weeks of real use, once the agent's behaviour matches what you expect. Indicative estimate, adjusted to the options you keep. It is not a delivery commitment.
Our commitment

Four guarantees that matter in prevention

Your employees' data stays with youLocal inference or an isolated resource hosted in France; no minutes entrusted to a foreign third party, no content used to train a model.
Data in France, under French lawOccupational health and safety: minimisation and location in France, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
The employer approvesThe agent assesses, proposes and tracks; approving the document and deciding between actions belongs to the employer, who answers for them.
No version overwrittenEvery successive version is kept and stays readable exactly as it was approved, with its date and its author.
Frequently asked questions

Your questions, our answers

From what headcount is the document mandatory?
From the first employee. The duty to assess risks and record the results in a single document applies to every employer whatever the company's size; the assessment is recorded by work unit. The document is kept available to workers, former workers, members of the social and economic committee, the occupational health service, the labour inspectorate and the prevention officers of the social security bodies.
How often must it be updated?
At least once a year in companies with eleven employees or more. Below that threshold the annual update is not required, but an update remains mandatory on any significant change affecting health, safety or working conditions, and as soon as further information relevant to assessing a risk comes to light. That second trigger fires most often, and it is the one the agent watches continuously.
How long must the document be kept?
The document and its successive versions are kept by the employer for at least forty years from the date each was drawn up. That is why the agent never overwrites a version: each stays readable exactly as it was approved, with its date, its author and the facts it rested on.
Does the agent decide on prevention actions?
It proposes them, ranks them and quantifies them in time and resources, then follows them through to closure on evidence. Deciding and approving belongs to the employer: the employer approves. That is also what the AI Act expects of this kind of use — support, human oversight, traceability.
What about the prevention plan with outside contractors?
The agent prepares it from the joint prior inspection and the interference risks identified. It flags the cases where a written plan is mandatory: where the operation represents at least four hundred foreseeable working hours over a period of twelve months or less, or where the work appears on the list of hazardous work set by ministerial order, whatever its duration.
Does the agent measure individual employees?
The document describes posts and work units, not people, and that is what makes it useful: an individual indicator would soon become a target, shop-floor reports would dry up, and the company would lose the very instrument that lets it see accidents coming. If you still want an individual indicator, the agent produces it and brings you the conditions to meet — proportionality, prior information of employees, and consultation of the social and economic committee from fifty employees. The decision is yours.
Is our employees' health data protected?
Yes. The agent is hosted in France, locally or on an isolated resource, with the deployment objective of processing and access operated within the European Union and an architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity. Named material is walled off and access is role-based; the document itself stays written by post, with no individual health data.
How long does deployment take?
A few weeks as a rule, after a free 15-minute audit that scopes your work units and your sources of facts, then a design, integration and testing phase before go-live and team training.
Let's talk

Let us size the potential in your prevention work

15 minutes to scope your work units and your sources of facts — hosted in France, supervised, no commitment.