French SMEs, 50 to 500 people
Fast or right? You no longer have to choose.
Everyone already has their own AI and is faster on their part. The file, though, is not moving any faster: what the AI gets wrong, nobody counts, so everyone re-reads other people's AI. We put AI on the whole file: it proposes, a program checks the rule, your teams judge.
Twenty minutes, free. You leave with the piece of work to start with.
Or in writing, an answer within 48 hours: tell us which document comes back most often.
The quote, the tender response, the report, and every piece that leads to them: AI writes them fast, and someone re-reads them in full, because a mistake there is costly. You have gained nothing: you pay for the tool, and for the re-read on top. An AI that speeds up one person does not speed up a firm.
What gets a file out right is what your firm knows, from the one who builds it to the one who bills it; nobody has ever written it down. We write it with your teams, on your files: each person receives only what needs their eye, the rest goes out checked by a program, with no re-read, and what is written stays yours. In three years your competitors will have the same models. They will not have your know-how.
Are your teams already using AI?
On your last twenty files, we count what it gets right, what it misses, and on which questions. You know what goes out before you hand it one more file, and who in your firm may use it for what.
No AI in your firm yet?
On three of your files, with the people who know, we write down what follows a rule and what takes their judgment. You know what to build, and what this work costs you today, before paying for any software.
What we offer
Measure before building.
An assistant does the same thing, faster, for one person. Here the software knows the whole file: what follows a rule goes to it, judgment calls to your teams.
What it changes in your firm: the file is delivered sooner, not just each part of it. It crosses the house without queueing in front of the people who know. Each one re-reads only what takes their judgment; the rest goes out, every rule checked by a program on every file, the tenth as carefully as the first, with a measured error rate.
When to hand AI the keys, when to take them back, and who decides
When do you give AI the keys?
When the question follows a rule: anyone would find the same answer by looking at the file. “Has the client supplied the five documents we asked for?” A program checks it, its error rate is measured on your files, and that file goes out without a re-read.
When do you refuse AI?
When a plain calculation is enough. And when the question is a judgment call: “Is this price defensible on this job?” Someone answers for it, and no model, however good, takes it over.
When do you take the keys back?
When the AI changes version. It is measured again on your files before anything moves, and if it slips back, the question goes back to your teams.
Who decides?
You. What goes back to your teams is written in advance, not left to the software's discretion, and you validate every change.
What you are usually offered is someone re-reading behind the AI, or nobody, and no measure of its mistakes. Either way, you do not know what goes out. We write that line, what goes to the software and what stays with your teams, on your own files, and we measure it.
The same models for everyone. Your know-how, though, is yours alone.
Your files stay on your premises or on sovereign infrastructure, and nothing confidential leaves without your consent. The software is designed for the requirements of the EU AI Act.
The same line holds in every firm. Where everything is re-read, it says which files still need a judgment call, and the time of the people who know goes back to those. Where each person builds and sends their own file, it says the same thing, except that this time it points at files nobody was opening.
Nobody knows where that line runs in their own firm. Writing it, on your own files and with the people who know, is the work.
ANSWER WITHOUT US
This month, of what your teams handed you, how much was prepared with AI? And how much did you have to redo?
The first number, nobody told you: AI came in at every desk, not through yours. The second, nobody counts, in any firm. It is the first one we pull from your last twenty files, and on its own it says whether this piece of work is worth the project.
If AI is already in place at your firm
The measure — three weeks, €3,000 excl. VAT, not due if the result does not let you decide
In detail
Copilot, ChatGPT or your vendor’s module: we measure it on your last twenty files, where it is right, where it is wrong, and on which questions. You receive the table, and one page: what it can be given, what it must never be left with, and who in your firm may use it for what.
You decide whether to keep it, frame it or stop it. If the answer is to keep it as it is, that is what we will write.
We need the tool as your teams use it, and your last twenty files, the last twenty and not the best, with what the right answer was. We have it redo each one, and we count. Those twenty files stay with you: when the tool changes version, you run it on them again. The measure runs on your premises or on sovereign infrastructure.
An example, built for illustration. A technical distributor of eighty people. Requests for a quote reach Éric, the sales engineer, by email. He prepares the answer with Copilot, then re-reads all of it, because he does not know where the tool went wrong. He gives us his last twenty answers as sent, as he corrected them, and the twenty requests.
Week 1, with him: the fifteen or so questions an answer has to settle. Week 2: the tool redoes the twenty answers, and we compare, column by column, with what he sent. Week 3: the table.
| Question | The tool had it right |
|---|---|
| Reference and quantity | 20 of 20 |
| Lead time | 14 of 20 |
| Discount | 9 of 20 |
| … and twelve more columns (figures invented for the example) | |
Éric knows what the tool takes care of, what it must never be left with, and what he still has to look at. Instead of re-reading everything, he re-reads the discount and the lead time.
One signature, and you choose where to stop.
- Where to start — free, an answer within 48 hoursIn writing: the piece of work to start with, what it costs you today using your own figures, and the ones that do not lend themselves to it.
In detail
The right first piece of work comes back often, depends on a few people, and part of its questions follow a rule. There are two ways to find it, your choice.
In writing, without a callYou describe in five minutes what comes back in your firm. We answer within 48 working hours.
Describe what comes back in my firm →In conversation, in twenty minutesThe same choice, and we also check who can say yes and where your past files are.
Let's talk about you → - The three reference files — six weeks, €4,500 excl. VAT, the first instalment of the project, deducted from the software if you continueThree of your last twenty files, gone through with the people who know, half a day a week. You leave with your three files annotated, three numbers, and the plan.
In detail
We start with a few interviews with everyone the file passes through, from the person who puts it together to the person who invoices it, to confirm it is the right piece of work. Then you give us the list of your last twenty jobs sent out, the last twenty and not the best, and say for each whether it was good or less good. We pick three together: an ordinary one, one you would no longer send, and an edge case. We go through them in detail with the people who know, half a day a week for six weeks at most.
What you receive, and what stays with you if you stop here
- Your three files, annotated. Question by question, what made the file good, what you would no longer send, and why. It is the first time your firm writes down what it calls a job done right.
- Three numbers you do not have today. How many days a file takes to go out, how much of that time it spends waiting on someone, and how much it changes depending on who put it together. They come from your archives, not from your teams’ time.
- The plan for the project. For each question in a file, it says who takes it, and for whom. What follows a rule goes to the software, with the cheapest method that holds up: AI when it is needed, a plain calculation when that is enough. Judgment calls stay with your teams. A question the program can check may stay in the hands of someone new to the trade, for as long as it takes them to learn it: going fast and learning are not the same use, and you set which one, for whom. The plan also writes down the time of those who answer for the files, and in what order to build.
These three files lead to the plan. The plan prepares the specification, whether you have it built by us, by another provider or in-house.
This stage does not yet say how many files the software will handle on its own. That is the first thing the next stage measures, on the other seventeen files on your list.
You can stop there, with no penalty and owing us nothing more.
- The software in place — priced at the end of the three filesOn your premises or on sovereign infrastructure. No file goes out until every rule has been checked by a program, the tenth as carefully as the first. The software says which ones still need a judgment call.
In detail
Before building, we apply the plan to the other seventeen files on your list: you see how many the software would handle on its own, and what it would miss, before you pay. It applies each person's part: what it shows the newcomer is not what it shows the one who knows. When it does not know, it does not guess: it stops and sends the file back to your teams, to whom and how is written in advance. It keeps a record of everything it decided on its own. It runs where the file already travels, with the people who look after your tools. You hold the keys: it evolves without us, and you have what it takes to extend it to a second piece of work.
Want to see what it looks like on a real file? Read AI for a tender response: technical proposal, DPGF, DC1 and DC2. You will see what it writes, where it goes wrong, and why.
- The follow-up — includedA register: what the software handled on its own, what it sent back to your teams, its error rate, and what came of each change.
In detail
The three numbers from the start are measured again. The software works; it answers for nothing. In your firm, each person answers for what comes to them, through steps written in the plan, with their time: who sets the acceptable error rate, from what a mistake costs you; who goes through the sample of files drawn at random each month, looking for what the software missed; who stops it if the rate drifts, and on which figure; who rules on what it sends back. You check one number, not every file. Every version of the rule is kept, and each file carries the version that handled it: you see what came of a change, and where the rule has to move. The register is yours and can be exported.
- Once in place — the software keeps learningEvery file your teams rule on sharpens the line, and you are the one who approves each change. When the AI changes version, it is measured again on your files before anything moves.
In detail
Cases nobody foresaw always come in. The register says where the rule has to move. A new version, of the rule or of the model, is first replayed on the files already checked: you see what it changes before you say yes. Nothing changes in the software without you, and every change is measured like the first. We come back when you call us.
Who you will be dealing with
Julien Pierre Martin
The trade, before the tool
Solutions in production since 2000
AI for 20 years, not 20 months
In detail
You work directly with Julien Pierre Martin. A PhD in AI, he has been building tools since 2000 alongside the people who use them, in logistics, animated film, investigative journalism and law.
The trade, before the tool. A file passes through several hands, and each one knows something the others do not. Before building, we understand who does what, who depends on whom, and what each person needs to know for the whole to hold. We take the problem as you live it, and we turn it into a solution that holds, with software exactly where it pays.
Solutions in production since 2000. He designed film production pipelines. He is behind the search application and the secure network of the ICIJ, the consortium behind the Panama Papers. Each time, the tool had to work that same day, for people who had no time to fix it. Laureate of the Brown Institute (Stanford / Columbia). Distinguished by the jury of the CNIL-Inria prize.
AI for 20 years, not 20 months. A PhD in AI, he has led AI engineers and spent ten years in constrained optimisation. He knows how to choose an AI tool and judge it on your files, not on its demo. We take the cheapest method that holds: AI when it is needed, a plain calculation when that is enough.
Which piece of work comes back most often in your firm, and costs the most when it goes wrong?
Twenty minutes, free. You will know whether it can be done in your firm.
Frequently asked questions
Who is this for — and who is it not for?
For the person in charge of a piece of work that comes back in a French SME, and who can decide to change it: the owner or managing director, the operations or plant director, a branch director, a technical director. You don't need to be the AI specialist — that's the starting assumption.
The right piece of work to take on has four traits: it is still partly done by hand; someone often has to decide in it, and that is not written down anywhere; a mistake costs something you can put a number on; and you can decide to change it without a tender.
It isn't for you if ten files out of ten are handled by applying the procedure to the letter — a plain automation costs less than we do, and we'll tell you so. Nor if the method belongs to a third party, a software vendor or an accreditation body, whose sign-off any change requires. Nor if the goal is to cut jobs: those projects turn on you at the first awkward file.
What happens during the twenty minutes?
You describe a piece of work that comes back, or your business if you do not know yet which one to pick. We ask five questions: how often it comes back and how long it takes, who makes the judgment calls in it, what a mistake costs when one gets through, where your past jobs live — bids, costings, lessons learned — and in what shape, and who can say yes.
The call ends with a straight answer: the piece of work we would start with, what it costs you today per year, estimated from your own figures, and what we would do in your place. No deck, no diagnosis to buy, no surprise quote: a quote only comes afterwards, and only if both sides want to carry on.
The person who would do the work is the one who takes the call.
What does it cost?
L'Ordre du jour is free. Counsel is €99/month, fourteen days' trial with no card. The twenty minutes cost nothing and commit you to nothing.
The project takes a single signature. The three reference files cost €4,500 excl. VAT: that is the first instalment of the project. If you go on, it is deducted from the price of the software. The software in place is quoted at the end of those three files. We cannot price it earlier, because the price depends on the work and we do not know it before looking at it with you. The follow-up is included: it is a register, and it belongs to you.
The measure of the AI already in place costs €3,000 excl. VAT, not due if the result does not let you decide.
You can stop at the end of the three files, with no penalty. You keep your annotated files, your three numbers and the plan.
Before we start, we fix with you what this work costs you today (real lead times, rework, files turned down) and what will count as success. Those are the numbers you will follow the software against afterwards.
Where do our data and our client files go?
Two answers, because these are two different things.
The solution we build for you runs on your premises or on sovereign infrastructure, it can evolve without us, and there is no subscription to our presence. Your files never have to leave your walls.
We choose each AI model according to what your files contain. Whatever is confidential goes through models installed on your premises or on sovereign infrastructure: you keep control of them, and their cost does not move at a vendor's whim. A commercial model can serve on a step where it does better, provided it sees nothing confidential. It then works on documents that are not, or on files from which we have first removed the names and whatever you want protected. You decide what leaves, and the record shows which model served at each step.
The EU AI Act requires, depending on the use, that a record be kept of what a piece of software decides and that a person answers for it. The software keeps that record, and the judgment calls stay with your teams. What the regulation requires depends on your use: we look at it with you from the three reference files onward.
Counsel, our online product, is more exposed and we would rather say so: the text of the decisions you submit to it is analysed in the United States by Google LLC (Gemini API) and Tavily Inc., processors within the meaning of Article 28 of the GDPR, who do not use it to train their models. Everything else — account, reading, billing — is hosted in France with OVH and does not leave the European Economic Area. The full list is in our privacy policy.