AI solutions for industrial SMEs
Concrete results with AI.
I help industrial SMEs deploy artificial intelligence where it truly pays off. Starting from concrete problems, and automating the tasks that cost you the most. Always measuring: you decide based on numbers, not promises.
The problem
AI is everywhere. Results, almost nowhere.
You are promised a revolution: spectacular demos, magical tools, major transformations. Yet look around you: how many businesses can show what AI has actually brought them? Disappointment is common. And the causes are specific.
The problem is not artificial intelligence. It is the way it gets deployed: everywhere at once, with no precise problem to solve, plugged into disorganized information. AI only delivers value when you give it two things: a real problem, and reliable information to solve it.
My approach
Always start from a concrete problem.
No grand transformation: we pick one specific problem, the one that costs you dearly, and we solve it.
The problem in this example: warranty claims
Before writing a single line of the reply, a service expert has to cross-check the contract, its amendments, the machine's history and the nature of the failure. Then fill in the claim file, and only then write the reply. It takes hours every time, while the customer waits. Let's follow one of these claims, step by step.
ChatGPT doesn't know your company
It helps: a letter outline, a well-turned sentence. But it has never read your contracts, your terms of sale, your product data sheets. As soon as the answer calls for specific information, it stalls, or worse, it makes things up.
Remember to check the warranty period in the contract, any extensions, and the maintenance history.
✓ The form ✕ The substance
✕ A general-purpose tool doesn't know your context ⚠ Risk of leaking confidential data
Your context: the company memory
We gather the documents that count and turn them into a knowledge base that stays under your company's control. Your AI uses it as its only source: for every question, it finds the relevant information and writes its answer from it. Your teams ask their questions, and the checks that blocked ChatGPT now take a few seconds, with the source attached.
✓ It never invents: you can trust its answers
Entire tasks handled autonomously
From a reliable knowledge base, we can move to the next step. This time, the gains are counted in hours, sometimes days. The AI no longer just answers questions: it takes on the whole task. That's what we call an agent. The file lands on your service expert's desk already prepared: he checks the work, settles what calls for judgment, adds the finishing touch.
- Handled by the AI agent
- Gathers the documentsCustomer ABC's contract, the extension amendment, the terms of sale and the conveyor's service history.
- Checks the coverageThe 24-month base warranty has expired, but the 36-month extension is still running. The gear motor is not one of the excluded wear parts.
- Fills in the claim fileConveyor references, dates, parts involved, reason for the service call.
- Drafts the documentsThe draft reply to the customer and the work order for the technician.
- Flags the points of attentionA chassis impact, noted in January in the service history: possible warranty exclusion.
- Handled by your service expert
- Checks the workHe reads through the prepared file: every statement cites its source.
- Settles the point of attentionDoes the chassis impact void the warranty? That decision is his.
- Adjusts and sends the replyCustomer ABC gets an answer. The half-day of research is gone: your expert spent his time deciding, not compiling.
The same pattern applies to any problem: reliable information first, then automation.
The ground rules
You stay in control of your documents.
Connecting an AI to your documents raises a fair question: what exactly does it do with them? Here are 3 rules, and they hold from the first day to the last.
You decide what goes in
We do not connect the AI to every one of your IT systems. We select the documents that count for the problem at hand: procedures, technical data sheets, contracts. We sort them, set aside what is outdated, and nothing goes in without your approval.
The AI never modifies your documents
It reads your documents once, builds its memory from them, and works from that memory afterward: nothing gets rewritten, nothing gets deleted. When an agent generates a document, it produces a new file and leaves your originals untouched. And when a procedure changes, you replace the old version in the knowledge base: it is never cited again.
Your tools stay as they are
Your teams keep working with the tools they know: no migration, no 18-month IT project. A copy of the selected documents goes into the knowledge base, but the original stays where it always was: in your own systems. Nothing locks you in: if you switch solutions in 2 years, you leave with everything, intact.
Concrete situations
Situations you will recognize.
Here are a few examples of situations I can solve with AI: representative scenarios, drawn from 8 years inside industry. The details will differ in your company, but I am willing to bet that at least once, you will think: "this is us."
The countdown
Your workshop manager retires in 18 months, and 30 years of knowledge leave with him. He has already been asked to write down what he knows: 6 months later, 11 pages. The countdown continues.
The defect you had already solved
A batch blocked at your biggest customer, an analysis report demanded within 10 days. A similar defect was already solved 3 years ago, but the engineer who led that analysis has since left the company, and his report is nowhere to be found. The team is about to pay again, in hours and stress, for work already done once.
The Saturday night breakdown
A cold room breaks down on a Saturday night. Your on-call technician is seeing this site for the first time: he has never set foot in it, and the expert who knows it by heart cannot be reached. The goods spoil while he searches blindly for answers he should have had from the start.
The audit that threatens your supplier status
Your biggest customer's auditor holds 2 versions of the same procedure, one in each hand. They do not say the same thing. Your quality manager did re-check everything before the audit, but years of growth have made your documentation too dense for anyone to catch everything. 3 weeks later, the purchasing department demands an action plan before renewing your supplier status.
The resignation after 5 months
Your new recruit hands in her resignation after 5 months, when it took you 8 to find her. Her reason: "I never had the information to do my job properly." Every question went to Bruno, the only one who really knew, several times a day. The next recruit will ask him the same questions.
The tender that blocks everything else
A critical tender comes in, the kind that can make your year. The customer demands a compliance matrix: 300 requirements, each one to be proven, document included. Your engineering team spends 2 weeks on it full time, and most of that time goes into finding the supporting evidence and compiling the information.
How can I help you solve them? I explain it in the executive guide: a 30-minute read, free
My method
A path with no unpleasant surprises.
At every step, you know where you stand, what it costs, and you remain free to continue or stop.
A first conversation, 45 minutes, free
You describe your problem. Together we assess whether an artificial intelligence solution can address it and what it would involve, I give you a concrete preview built on a typical case, and you leave with a price range, from this very first conversation. No string of 3 meetings before we talk numbers, no endless PowerPoint deck. A conversation about one concrete problem.
The diagnostic, free of charge
If you wish to continue, we carry out a diagnostic, with your teams: we define the problem precisely, measure what it costs you today, set the scope, design the solution, confirm technical feasibility and estimate the gains. We also set measurable success criteria before starting, so that in the end we judge on facts, not on an impression. You receive two documents: the diagnostic report, and the proposal. That is where you decide, with the numbers in hand.
Development and training
We develop your solution, tailor-made, building on proven technology components adapted to your need. We structure the information relevant to your problem, bring the solution online, validate the tests defined in the proposal, then train your teams.
The life of the solution
The solution runs, your teams use it, and I watch over it: hosting management, updates and backups, monitoring the quality of the answers, and adjustments over time, because your company evolves, and the technology itself shifts several times a year.
Transparency
How does my pricing work?
Have you noticed how digital service providers love the "custom quote" that only gets revealed at the third meeting? I make the opposite bet: I explain the structure before we even talk.
And the amounts? I give you a price range from our very first conversation, adapted to your situation. Spoken plainly, not buried in fine print.
- A price range from the first conversation. 45 free minutes, and you know where you stand before committing to the diagnostic.
- A free diagnostic. It leads to a clear offer, in 2 blocks: the development cost of the solution, and the monthly service cost, known in advance.
- Payment on go-live. Development is only paid once the solution is live at your company and the tests have passed.
- No per-seat license. Whether 20 or 200 people ask questions, the price does not scale with headcount.
Your questions
Engineer's answers, not a sales brochure.
What if the AI gets it wrong?
Yes, it can happen, and claiming otherwise would be dishonest. But everything is built so that errors stay visible and recoverable: every answer cites its source, verifiable in one click. When the information does not exist in your documents, the AI says so instead of making it up. For automated tasks, your experts remain responsible for the review and add the finishing touch. Reliability is tested with your teams, on real cases, before going live. You do not have to take my word for it: you can verify everything yourself.
Will my teams actually use it?
The tool is custom-built for your teams, with one constant requirement: it must stay simple to use. You ask a question the way you would ask a colleague; there is almost nothing to learn. And above all, the usefulness must be immediate: we always start on a process that hurts, where the benefit is felt from the first week. We do not impose a tool: we remove a pain.
What types of documents can it read?
Text documents, very well: procedures, technical data sheets, minutes, reports, whether in PDF, Word or your emails. More complex files, such as technical drawings or large spreadsheets, are handled case by case: some integrate easily, others require extracting the useful information into another format first. This is exactly what we check during the diagnostic.
Where is my data hosted, and which AI runs the whole thing?
It depends on your project: hosting is chosen with you, during the diagnostic, according to your constraints and preferences. What never changes: your documents remain your property, the hosting is dedicated to your company alone, never shared with other clients. We also choose the AI model together, to strike the right balance between quality and cost.
How long does the project take, and what does it require from my teams?
The first conversation lasts 45 minutes. The diagnostic then involves a few of your people, for a handful of interviews, so that I understand the process in depth. Then comes development: this is when your teams are most involved, to gather the documents that will feed the knowledge base. Depending on the chosen solution, your IT department may also step in, for a limited time, for example to set up access rights. Your teams come back at the time of testing and training. The total duration of the project is estimated during the diagnostic phase. But rest assured, we are talking weeks, not months.
Why don't you display any prices?
A single figure would make no sense for a tailor-made solution: your case will not cost the same as the company next door. It depends on how complex the problem is, and how complex the documents are. What I commit to instead: you never have to wait. From the first conversation, you leave with a price range adapted to your situation.
Is the development price fixed, or can it drift?
Fixed. That is precisely the role of the diagnostic: pinning down the scope and the solution before giving you a number. If a new need appears along the way, outside this scope, we discuss it before adding it, never after the fact on the invoice. No unpleasant surprises at the end.
Couldn't we build the same thing ourselves with ChatGPT?
Technically, yes: ChatGPT can connect to your documents. But 2 differences matter in practice. First, your documents: to be usable, they have to be copied and indexed on the vendor's own infrastructure, and for a European company that means storage in the United States today. Then the cost: this kind of subscription is billed per user, one license per person, which climbs fast with your headcount. A tailor-made solution solves both: your documents stay under your control, and the price does not depend directly on the number of people asking questions.
Is this suited to a company our size?
What matters is not the size of your company: it is having a precise, costly process to tackle. Time lost searching for information, knowledge that depends on a few people, answers that come too slowly. The solution starts with a limited scope and grows from there: it fits a single-site SME as well as a multi-site group. Either way, write to me: we will find out together in the first conversation.
Will this solution reduce our headcount?
That is not the goal, and it is not what I recommend either. The time freed up gets reinvested: fewer hours lost searching for information, less time spent on repetitive tasks once automation is in place, and more time on what really moves your business forward.
Who I am
Expert in AI deployment for industrial SMEs.
I design and deploy artificial intelligence solutions for industrial SMEs: the company memory that answers from your documents, then the agents that take on your most costly tasks. I come from the industrial world, not the software world: 8 years in aerospace and defense, improving processes on the ground, with data to back it up. Your workshops, your audits, your tenders: I know them from the inside. My method comes from there: measure before promising, start small, prove it, then extend. My background in detail →
Which problem would you start with?
Describe your problem in 2 lines, and let's book 45 minutes. You will leave with a frank opinion on your case and a price range. Free, no commitment.
Or directly by email: contact@remy-giraud.com