Data & AI consultancy · Haute-Savoie · Geneva · remote
Your data is scattered across inboxes, your CRM, your business tools and a dozen spreadsheets. We bring it together in one place, we make it trustworthy, then we make it usable: dashboards that are always current, reports that arrive without anyone building them, and an assistant you can question the way you would question a colleague.
We start with a fixed-price audit of one to two weeks. You know what the next step costs before committing to it.
Your sources today
Single data foundation
Centralised · historised · reliable · current
The diagnosis
Most SMEs don't have too little data. They have too much of it, badly filed, and nobody making it converge.
Inboxes, CRM, ERP, SaaS business tools, shared drives. Each tool holds a fragment of the truth and none of them talk to each other.
Re-keying, consolidating, monthly reporting rebuilt by hand. Your team's time spent moving numbers around instead of acting on them.
You cannot cross sales with operations and finance. So you cannot decide on anything but instinct.
An assistant sitting on inconsistent data answers wrong, confidently. The foundation comes before the model. Always.
The service
A complete chain, from the data source to the answer on your phone. It has been our speciality for nine years, now with a layer of artificial intelligence at the end of it.
We connect to what you already use and pull your data into a single warehouse, historised and reliable.
The data comes to you, with nobody having to go and fetch it. This is the part that pays off within the first few weeks.
Once the figures are trustworthy, an assistant answers your questions in plain language, on desktop and on mobile.
The point that makes the difference
The language model computes nothing and invents nothing. Its only job is to understand your question and connect it to a validated calculation on your warehouse. The figure on screen comes out of your own database, with its source and its date.
Business calculations are defined and validated with you, once and for all. The AI picks the right one, it does not invent it.
Source, extraction date, scope. You can verify any figure on screen.
Explicit scope and guardrails. A question outside coverage gets “I don't have that data”, not an invention.
In practice
Use cases
Here are five situations every company director recognises. Each looks trivial, each costs several hours a week, and each defeats improvised automation. This is exactly the kind of problem we work on.
Day to day
A hundred and fifty messages a day. The urgent client request is wedged between two newsletters and a supplier chaser. You deal with what you see, not with what matters.
What we put in place
Every incoming message is read, attached to the right client and the right case, then sorted by nature and urgency. Anything needing a reply rises to the top, with the context of the exchange already gathered. The rest is filed away.
Why it isn't trivial - Understanding what a human actually meant in a message, without mistaking a subject line in capitals for urgency. Classic filtering rules break their teeth on this.
Day to day
Quotes, invoices, purchase orders, supplier PDFs. Someone in your company is re-keying amounts, dates and references into a spreadsheet or your management software by hand.
What we put in place
The document is read on arrival, the useful information extracted, checked, then placed into your system. Anything doubtful goes to a human for validation - never guessed.
Why it isn't trivial - Every supplier has its own layout, and it changes without warning. You need a system that recognises what a field means rather than where it sits on the page, and that knows when it isn't sure.
Day to day
“I'll send that over on Monday.” That sentence exists in an email, in a meeting note and in two people's heads. It exists in no task list at all.
What we put in place
Exchanges and meetings are analysed, commitments extracted with who, what and by when, and a task list keeps itself up to date. You see what was promised, by whom, and what has slipped past its date.
Why it isn't trivial - Telling a real commitment apart from a polite turn of phrase, and working out that “that” refers to something three messages earlier. Obvious to an attentive human - and that is where all the work is.
Day to day
“What did we agree with this client last year?” The answer is in an email belonging to someone who has left, in a file with an incomprehensible name, or nowhere at all.
What we put in place
The whole history - exchanges, documents, meeting notes, client records - becomes searchable in a single sentence. You ask the question, you get the answer and the source it came from.
Why it isn't trivial - Everything has to be gathered and attached to the right entities first. Without that groundwork, clever search over disordered data returns answers that are confident and wrong.
Day to day
The same questions come back every week: availability, the status of an order, the lead time on a job, a quote that needs chasing. They arrive by email, by form, by phone, and someone on your team spends their days answering what they already know by heart.
What we put in place
An agent handles inbound requests on your channels from your data and your rules, not from its imagination. It absorbs most of the volume end to end. As soon as a request falls outside its scope or calls for a judgement, it stops and notifies the right person, with the history of the exchange and what it has already checked.
Why it isn't trivial - Knowing when to stop. An agent that answers everything eventually gets it wrong in public, in front of a customer. The work is calibrating the boundary - what it handles alone, what it hands over, and to whom - and then keeping an eye on it over time.
All five rest on the same foundation. Once the data is gathered and attached to the right entities, they can be added one after another - which is why we always start there.
Worked scenarios
The use cases above describe irritants. These describe companies: their size, their tools, what the audit turns up, what gets built week after week, what it costs and what it changes. Enough to place your own situation before you even talk to us.
These are modelled scenarios, not client references. They draw on situations we meet and on the method we actually apply; the figures are orders of magnitude, not the results of a signed engagement. Our track record is our own software, further down this page.
45 staff · €12M revenue · 3,000 SKUs
Monthly revenue lands twelve days after close, stockouts are discovered the day the customer calls, and four tools give four different definitions of revenue.
Monthly revenue available at D+1
Read the scenario
80 staff · 400 trade customers · 120 requests a week
Four people in after-sales spend their days answering the same six questions: lead time, order status, part compatibility, warranty, documentation, scheduling a visit.
60 to 70% of the volume handled end to end
Read the scenario
3 venues · 60 staff
Occupancy, RevPAR and average spend live in three tools that don't talk to each other, and booking requests that arrive outside opening hours get lost.
Three venues on a comparable basis
Read the scenario
4 branches · 35 permanent staff · 450 temps a month
The right person has already worked for the group, but their skill sits inside a PDF CV: the consultant goes off to post an advert instead of calling them back.
The talent pool becomes queryable
Read the scenario
40 staff · billed per job · no finance function
The owner steers by the bank balance, the P&L lands in the spring, and the margin per job stays unknown. The management accountant role was costed, then dropped.
A monthly position, without the headcount
Read the scenario
The firm
ET Technology Consulting is a technical consultancy focused on a single subject: making a company's data converge and turning it into something usable, all the way to the AI layer. We don't do brochure websites, ERP replacements or managed IT. We do this, and we do it end to end.
Our systems run for clients who cannot afford an outage: 24/7 continuous processing services, consumer applications with more than 80,000 users, data chains handling millions of rows. What we install in your company is built to the same standard.
Data collection, warehouse, data model, dashboards, mobile app, AI layer: all of it is covered by the firm. You have no three suppliers to keep talking to each other, and no integrator quietly subcontracting half the work.
Most offers on the market sell an assistant sitting on unprepared data. We take the problem the other way round: a sound foundation first, intelligence on top of it only afterwards. That is the difference between a tool your teams use daily and a demo abandoned after a month.
Scope, timelines, deliverables and monthly infrastructure cost are quoted before the work starts. You know what you are buying, and what the system will cost to run next year.
ET Technology Consulting - a French SASU. Project team sized to the engagement, with a single technical point of contact from scoping through to day-to-day operation.
How we start
You don't sign off a platform on the strength of a slide deck. We start small, with a deliverable that serves you even if we stop there.
A map of your sources, the real state of your data, the three questions costing you the most today, and a precise quote for what follows. The deliverable is yours and stays usable, even if you stop there.
Connectors, automatic collection, warehouse, data model, and the first dashboards in production. Delivered in usable stages, not one big reveal six months out.
Automated reports, alerts, then the conversational assistant on web and mobile, plugged into the foundation. The AI layer arrives once the figures are trustworthy, never before.
Supervision, evolution and infrastructure cost monitoring. A badly tuned setup costs three times what it should: we size it and we keep it under control.
Our products
The firm builds and runs its own platforms in production, with paying customers and an on-call rota. These are not client references: they are our own software, put to the test every day at our own expense.
Built by the firm, shipped with our engagements
The firm's own mobile app: a chat wired into your systems through MCP servers, the dashboards that go with it, and alerts that surface on their own. Under every answer, the query that actually ran and how long it took: the figure comes from the warehouse, never from the model. This is the interface shipped with the data foundation and the AI assistant.
Built and operated by the firm
A conversational agent platform deployed across some twenty agencies in France and French-speaking Switzerland. Automatic handling of inbound WhatsApp, email and SMS enquiries, handing over to a human as soon as a request falls outside the intended scope.
Built and operated by the firm
A booking agent and mobile foundation for venues: reconciling information scattered across a dozen tools, letting the end customer book on their own, and a back end feeding the mobile app.
Shareholder and technical publisher since 2022
A full-stack business application designed and built entirely by the firm, which holds a stake in it: back-end architecture, API layers, database optimisation, deployment and operation. Embedded AI layer with document search and an assistant for scoping the need.
The relationship is stated for each product. Architectures and detailed figures are presented in a meeting.
The experience of our people
ET Technology Consulting was founded in 2022, but it formalises an older practice: its people have always run their own projects alongside startup roles. What follows is the track record of the people who will do the work, not commercial references of the firm.
Founder's track record - Tech Lead, 2024-2026
Technical foundation on Google Cloud for an application with more than 80,000 active users. Data and analytics layer, AI-automated moderation, traffic peaks absorbed without the bill exploding.
Founder's track record - Tech Lead, 2019-2024
Processing chains handling millions of rows from music catalogues and broadcast feeds, continuous TV and radio content recognition, automated reporting for demanding clients.
The full track record, role by role, is on the founder's page. Profile
For those who dig
If you have an IT director, a technical supplier, or simply want to know what is under the bonnet, here is enough to judge.
Sources
Idempotent ETL
Scheduled, replayable, logged
Historised PostgreSQL
Data model, materialised views, versioned migrations
Two answer paths
Deterministic path
Validated business calculation, semantic cache. Handles most requests.
Model layer
RAG on pgvector, structured outputs, guardrails. Called when it earns its place.
Hybrid architecture: deterministic code does what it does well, and the model only steps in where it genuinely adds something. That is what makes the system reliable and the cost predictable.
Questions
It depends on the number of sources, how accessible they are and the state of the data - which is precisely what the audit establishes. The principle does not change: each step is sold at a fixed price, quoted before it starts, and you get the exact quote for the next stage before committing to it. We would rather discuss it with you than publish a range that matches neither your situation nor your budget.
First dashboards in production within 4 to 8 weeks after the audit. We deliver in usable stages: you see value during the project, not only at the end.
Hosting in Europe, strict isolation, encryption and a GDPR-compliant processing agreement. Depending on your requirements, the AI layer can be designed so that no sensitive data ever passes through an external model.
No. We connect to what you already use. Changing a tool is a business decision, never a technical prerequisite we would impose on you.
That is the most common case, and precisely why this service exists. We design, we deliver, we document and we stay available to operate it. You have nobody to recruit.
Everything we build belongs to you: code, data schemas, documentation and environment access. No closed proprietary components, no dependency on us to read your own data. If you hire a technical profile, they take the system over.
The mechanics are the same everywhere: heterogeneous sources, a warehouse, a delivery layer. Our references are in real estate, hospitality, media and consumer applications - what changes from one sector to the next is the business questions, and that is exactly what the audit is for.
Contact
Thirty minutes is enough to know whether your problem is a data problem, and what it would cost to fix. No commitment and no sales deck.