hutek · Lyon, France · AI strategy & engineering
Most AI projects fail before the first prompt: the problem was badly posed. We pose it, then we build the agents that solve it.
Pose the right problem. The agents do the rest.
Lyon · Paris · Las Vegas. On site, wherever you are.
Three objectives. Start with the one that serves yours.
Agents take over multi-step work, from repetitive tasks to high-value ones: analysis, synthesis, decision prep.
Bring back in-house what you outsource today: content, analysis, first-level support, dev tasks. Agents make it affordable.
Create services and products that were not possible before.
A speaker who builds agents, not one who recites slides: live demos on your own use cases.
Keynote “AI: a human revolution”. Places AI within three million years of human evolution: not a rupture to endure, but the next step of a long human adventure. Delivered to a partner’s clients.
We start by stating what you actually want to obtain, then pick the first project worth funding.
We build the whole system around the model: tools, tests, limited permissions, a stop switch.
We measure what ChatGPT and Google actually say about you, and which sources they trust.
hutek is an AI strategy and engineering firm founded by Sébastien Hubert in Lyon, France. It helps leadership teams of SMEs and mid-sized companies state the real objective and decide what to delegate, then designs and builds the AI agent systems that deliver it, with guardrails. hutek gives keynotes and masterclasses for executive teams, leads a guided trip to CES Las Vegas in January 2027, and is a Bpifrance-approved expert for the Diag Data IA programme (“Osez l’IA” plan). Its agent work serves three objectives: automate work, from repetitive to high-value tasks; re-internalize outsourced work; and invent new services.
Agents now read, decide and act. They change your business model and how customers find you, in AI answers as much as in search. Before any tool, we state the real objective: what you want to obtain, not the means you assumed. You leave with that objective written down, a prioritised roadmap, a first costed project and an honest go/no-go.
Objective · Roadmap · Go/no-go
A model alone is not a system. What works is the model plus its harness: tools, memory, tests and permissions, and several agents that check each other’s work. We design and build that system like aviation does: sandboxed, limited rights, observable, reversible, with a stop switch.
Harness · Guardrails · Stop switch
When agents execute, what stays human is setting objectives, arbitrating between them and owning the result. We redesign roles around that: a founding moment, then rituals and decision rules your teams own.
Roles · Rituals · Decision rules
Push a first door, with measured stakes and a clear scope. Each one lights up the next.
01 · Consulting
About three weeks in your real work: tools, teams, and the AI already used “under the table”. The objective written down, a roadmap, quick wins, a first costed project, a clear go/no-go.
→ Automate · Re-internalize · Invent: we find yours
Talk to us about “Operational AI diagnostic”02 · Consulting
How you appear in generative AI answers. Two weeks of analysis with our own tools, a one-day workshop, and a dashboard you keep: your visibility and the authority sources cited.
Talk to us about “Strategic GEO diagnostic”03 · AI agents
Automations and agent systems that run on their own, with harness and guardrails, plus independent advice on LLM architecture: models, orchestration, observability, reversibility.
→ Automate · Re-internalize · Invent
Talk to us about “Agent engineering”04 · Keynote
1h30 for boards and executive committees. A closed-door briefing: what AI agents change in your model, disintermediation (GEO), risks, and the decisions to make now.
Talk to us about “The executive wake-up call”05 · Masterclass
3h for seminars and conventions. High-energy: live AI agent demos on your use cases, de-dramatisation and immediate action.
Talk to us about “Team masterclass”06 · Recurring
Every month, an issue that decodes the weak signals of AI and digital and turns them into a few clear decisions. Currently distributed to hundreds of employees across 30 countries at a major pharmaceutical group.
Talk to us about “Monthly AI & digital watch”Selected projects.
01 · StrategyWhere to start
In a sector on the front line of AI, we helped the leadership team rethink its value chain and brand narrative: positioned as an innovator, not a disrupted victim.
02 · Agent pipelineAgents
From a single company ID, a pipeline produces a fully personalised audio message in minutes, with no human intervention: research, LLM portrait, script, synthetic voice, music, self-evaluation.
03 · Technical advisorAgents
For a startup facing a major change of scale, a clear technical diagnosis and roadmap to turn a promising POC into a product that handles very high volume without blowing up cost or reliability.
04 · AI visibilityGEO
For a laboratory present in 30+ countries, we mapped how the brand appears in generative AI answers, identified the sources models cite, and built a roadmap to weigh in.
05 · Automated analysisAgents
For a certification body, an application that checks hundreds of semi-structured documents a week against regulatory rules that only natural-language processing can verify.
06 · ProcessAgents
Expert reports that took a full day now take one hour: the system drafts a complete first version that the expert finalises.
07 · Strategic watchWhere to start
Since January 2026, a monthly English-language AI & digital watch for a major pharmaceutical group, distributed to hundreds of employees across 30 countries, turning weak signals into a few clear decisions.
Applied AI and transformation leader · Founder of hutek · Lyon
Yesterday, machines executed our instructions. Today they generate, handle and understand meaning, and agents test and refine their own reasoning.
The companies that succeed will not have the best algorithm; they will have stated the right objective and rethought their foundations. That is why hutek combines strategy with hands-on engineering.
Based in Lyon, working in France and abroad.
Sébastien on LinkedIn ↗hutek is an AI strategy and engineering firm in Lyon, France. We help leadership teams state the real objective and decide what to delegate to AI, then design and build the AI agent systems that deliver it, with guardrails.
Executive committees, boards and leadership teams of SMEs and mid-sized companies, in any sector.
Live AI agent demos prepared on your own use cases, then the decisions that follow: what to delegate to agents, what to keep, where to start. 1h30 for executive committees, 3h for teams and seminars.
We build the system around the model: tools, memory, tests, limited permissions, agents that check each other, observability, reversibility and a stop switch.
Yes, through Bpifrance’s Diag Data IA, part of the national plan “Osez l’IA” (France 2030), for which hutek is an approved expert. Public terms: 8 days with an approved expert over up to 3 months, quote capped at €10,000 excl. VAT, 40% subsidised (remaining cost €6,000 excl. VAT), for SMEs and mid-caps of 10 to 2,000 FTE with at least €1M revenue, over one year old, based in France. Bpifrance assigns the expert; you can name hutek as your preferred expert, but it is not guaranteed.
Our office is at 132 rue Bossuet, 69006 Lyon, France. We work on site wherever you are: Lyon, Paris, elsewhere in Europe, and Las Vegas for CES.
Lyon · Paris · Las Vegas. On site, wherever you are.
hutek · 132 rue Bossuet, 69006 Lyon, France