LinkedIn

AI consulting firm · SMEs & mid-caps

AI that frees up your team's time. The gain in hard euros, before you automate anything.

We map your processes, then I build what needs building: AI agents, business applications, automations, wired into the Microsoft tools you already use every day (Teams, Outlook, your business databases). I don't just advise, I write the code and ship it. No demo that ends up in a drawer.

Let's talk about your processes

Contact on LinkedIn

See the method ↓

01 · Method

The same path for every client

Start from business processes, stop what no longer justifies itself, prioritise and put numbers on it, then ship to production. AI is the means, never the starting point.

  1. Map

    A workshop with leadership and operational teams: where the hours go, on which documents, in which tools (M365, SQL Server, line-of-business apps).

  2. Prioritise and quantify

    First the steps nobody can justify any more: they come out without you buying anything. Then an impact × effort matrix and a costed 90-day roadmap. We only automate what has a demonstrable return, and bad ideas are ruled out, in writing.

  3. Build and industrialise

    AI agents, business applications and workflows built then shipped to production inside your IT system, with guardrails, traceability and human validation. And supervised over time.

Non-negotiable principles

  • We automate processes that are understood, prioritised and measured, never hunches.
  • Wired into your existing Microsoft estate, no parallel platform to maintain.
  • What gets built is yours: your code, your repo, your infrastructure.
  • Human validation on sensitive decisions (human-in-the-loop).
  • GDPR-compliant hosting and data processing, in Europe.

02 · Services

30-day AI diagnostic.

Before you spend a euro on AI, you know what it will return. In 30 days: your processes mapped, how much faster each one can run, what you can stop doing right now, and for every priority case a return in hard euros with a payback measured in months. You leave with an investment decision, actionable with or without me. Then I build whatever you decide to launch.

The full diagnostic process →

Starting investment

Fixed fee, all-in

Scope and price agreed in writing before work starts, set against the tens of thousands a poorly scoped AI project burns.

Book a conversation

Contact on LinkedIn

Guarantee: if the diagnostic surfaces no opportunity with a payback under 12 months, I refund half the fee.

AI agents & business applications · fixed fee

The agents and applications on the roadmap, built from the first line of code through to production in your IT system. Scope, price and acceptance criteria agreed before work starts.

See the service →

.NET AI integration · fixed fee

For mid-caps and software vendors with a .NET estate: AI developed inside your application and your SQL Server, not next to it.

See the service →

Maintenance & supervision · monthly

Supervision of agents in production, fixes and improvements as your needs and the models evolve. No long-term commitment.

AI visibility · free check

A separate service: when your customers ask ChatGPT, Perplexity or Copilot who to hire, your name should come up. Free check, then a fixed-fee upgrade.

See the service →

03 · Case studies

Real systems, built, shipped and in daily use

Two examples of what 'AI in production' means in practice: one integration into an existing IT system, and one web platform built from scratch. Two projects I designed and shipped to production in 2025 and 2026, before founding GiLabs. In both cases, code I wrote myself, running every day.

Diagram of the document AI pipeline: ingestion, AI extraction, human review, transactional import

industry · anonymised · 2026

Document AI pipeline

Two people were tied up full-time re-keying hundreds of paper timesheets by hand every month. Today, AI reads the documents for them: it extracts the data automatically, a human validates in a few minutes per batch, and everything flows straight into the business database, with no duplicates and no half-finished entries.

  • ≈ 2 full-time roles of manual re-keying redeployed to higher-value work, data-entry errors all but eliminated
  • Built in .NET inside the existing estate: validation interface, business rules, transactional import
  • AI proposes, humans validate: never a 100% automatic import
  • The system learns from corrections: quality improves with use
  • Wired into the existing estate: network folders, directory, business database
Read the full case →
Diagram of the Convention Online AI chain: collection, AI enrichment, generated visuals, publishing

web platform · built from scratch · 2025

Convention Online · professional events

A worldwide directory of trade shows and congresses that enriches itself every night: the AI collects the events, writes their descriptions in 4 languages and even generates the visuals, with no human input at all. Public site, back office and AI chain: all of it built from scratch.

  • A complete application built end to end: public site, back office, enrichment chain
  • Fully automatic: content and images produced every night, with nobody at the controls
  • Clean, structured data you can use directly: not rough free text
  • Runs unattended: recovers on its own after an incident
Read the full case → Visit the live site ↗

04 · The Copilot objection

"We already have Copilot. Why you?"

Fair question. Copilot and I don't operate at the same level: it speeds up your teams inside Office, while I redesign the whole process and build what's missing, all the way into your line-of-business apps, where Copilot stops.

05 · Questions

Frequently asked questions

How do you choose an AI consulting firm as an SME or a mid-cap?

Three questions to ask before signing. Who will write the code: many firms stop at the recommendation and subcontract the build. What you own at the end: the code, the documentation and the workflows have to stay with you and be picked up by another developer. And what the commitment is about: a gain measured in hard euros, process by process, beats a productivity promise. At GiLabs, mapping the processes comes before the tool, and the person who does the costing is the person who writes the code.

What does an AI engagement cost, and how is it billed?

A fixed fee, agreed in writing before work starts, never time and materials. The amount depends on the scope and is discussed in the first conversation, once your processes have been looked at. What is fixed in advance is the logic: the diagnostic puts a euro figure on the expected gain, and you only commit to a build if that figure justifies it. For eligible French SMEs and mid-caps, Bpifrance's Diag Data IA scheme can separately co-fund a diagnostic carried out by a Bpifrance-approved expert: it is distinct from this offer, and GiLabs will tell you whether it applies to you.

What sets an AI consulting firm apart from an IT services company or an agency?

An IT services company sells development days and expects you to already know what to build. An agency often starts from the tool, then looks for somewhere to put it. An AI consulting firm starts from your processes: what they actually cost in elapsed time, what can simply be stopped, what is worth automating. GiLabs does both halves, the framing and then the build, which removes the handover where half the context is lost.

Do your data have to be clean before you start?

No, but they have to be usable, which is not the same thing. An AI project does not need a spotless data warehouse, it needs the information to exist somewhere in a readable form, however imperfect: documents, a business database, a history inside your Microsoft tools. The diagnostic looks at that real state before promising anything, and if the foundation is missing it says so then, rather than halfway through a build.

06 · About

Yann Gilliot, founder of GiLabs

Yann Gilliot

Founder of GiLabs, AI consulting and engineering

15 years designing and building enterprise systems in the Microsoft ecosystem (.NET, Azure, SQL Server). I'm an engineer first: I don't stop at recommendations, I write the code and put it into production. My differentiator: AI already running in industrial production, a document extraction pipeline (scans → Azure Document Intelligence + LLM → business rules → database) used daily, not a demo.

Based on the French Riviera, working across France: workshops on site or remote, implementation fully remote.

  • 15 years of .NET / Azure / SQL Server development and architecture
  • Complete applications delivered from the first line of code to production
  • Document AI pipeline in industrial production
  • Microsoft Applied Skills: generative AI apps with Azure OpenAI and Semantic Kernel (AZ-2005)
  • At home in demanding contexts: industry, banking, regulated sectors

Why GiLabs

No demo left in a drawer. Agents in production.

The gain in hard euros, before you automate anything.

Let's talk about your processes

Contact on LinkedIn

30-day AI diagnostic · 30-minute video call

  • The one who quantifies it builds it.

    No context lost in a handover.

  • Wired into what you already run.

    Microsoft 365, SQL Server, business apps.

  • You own the code.

    Your repo, your infrastructure. Fixed price.