Value first
Opportunities mapped with management and scored on value, feasibility, risk and data. Hypotheses and success criteria before any build.

Alexandre Durand-Chabert · AI-native builder · Paris
I work out what to build with AI, then get something credible in front of the business, fast.
Sixteen years in innovation and consulting taught me where the value hides. Building with AI every day taught me what it really takes to ship. Today I sit with leadership, frame the problem, and build the prototype myself.
How I create value
Opportunities mapped with management and scored on value, feasibility, risk and data. Hypotheses and success criteria before any build.
Four-day cycles: two days on problem and solution, two days on a prototype running on real models, tested with end users.
Architecture options compared in a business case. Model routing, caching and cost decided on systems I run myself.
Acceptance criteria and evidence handed to the teams who industrialise. Discovery and delivery in sync.
Projects
AI products shaped with people who know a trade from the inside, and agentic systems that work while I sleep.
Radiologists dictate, the report writes itself, structured. Guardrails against model inventions. In Word, the browser and on mobile.
Whisper on GPU · Claude · Python · FastAPI · Modal
A voice and chat concierge for luxury events, run by a private concierge company under its own brand. Seventeen tools, four languages, a clean handoff to a human.
Claude · real-time voice · Node · Postgres
The AI news that matters, read at the source and summarised every day. Paste a link or a YouTube video and get the full analysis. Free plan and team plans.
Ingestion · deterministic scoring · LLM commentary
Visit aisources.ioTwelve specialised agents with three human checkpoints. 70 to 99 % less production time.
Meeting capture, project memory, one-click deliverables, with model routing and prompt caching.
Specialised sub-agents, context compression and deterministic hooks: the workbench behind all of it.
The media
A show I created on YouTube about how we stay informed, learn and make up our minds when algorithms and AI sort things for us. In The Feed, researchers, teachers, journalists and creators open their phone and show their tools, their sources and their rules. Three seasons so far.
Writing
Research notes of 3,500 to 4,500 words on AI-native innovation models, read by senior peers in the field. Including the essay Le consultant et son double (2026), on what AI does to the business model of consulting.
Experience
Discovery and prototyping for strategic AI initiatives across 25 countries, in close work with the group's AI teams: counterpart to leadership on AI priorities, prototypes on real models tested with end users, operating model with business lines, IT, legal and security.
Generative AI workshops with partners and enterprise clients, diagnostics turned into repeatable offerings, the firm's VivaTech programme.
Strategy, innovation and transformation for large corporates and public institutions.
A campus for high-growth companies: programmes on rapid prototyping and customer discovery.
Innovation and R&D diagnostics and funding for SMEs and mid-caps.
MBA, ISC Paris · MSc Competitive Intelligence, EEIE · MA History, University of Tours · Certified Design Sprint Facilitator · French native, professional English
Questions
Alexandre Durand-Chabert is an AI-native builder based in Paris. He has sixteen years of experience in innovation, strategy and consulting, since 2010. He is AI Innovation Hub Lead at Devoteam, where he leads discovery and prototyping of strategic AI initiatives for the leasing arm of a leading European banking group, and he designs and builds AI products and agentic systems himself.
A radiology reporting product in production, designed with a radiologist; a voice and chat AI concierge for luxury events, run in white label by a private concierge company; AI Sources, a public AI news product; an editorial pipeline of twelve agents in production since 2023; a consultant digital twin; and a Claude Code plugin.
He designs the architecture, builds with Claude Code as his main development environment, and reviews and debugs the code. His work is AI-native: he decides what to build and why, and he ships it.
AI Sources is a public product created by Alexandre Durand-Chabert, at aisources.io. It reads the AI news that matters at the source, from labs and regulators, and summarises it every day in five languages. Anyone can paste a link or a YouTube video to get a full analysis. It has a free plan and team plans.
AirParty is a media created by Alexandre Durand-Chabert: a YouTube show, The Feed, about how we stay informed, learn and make up our minds when algorithms and AI sort things for us. In each episode, researchers, teachers, journalists and creators open their phone and show their tools, sources and rules. It has three seasons.
The Durand Chabert Letter is his newsletter on Substack. It publishes research notes of 3,500 to 4,500 words on AI-native innovation models, built on his “Better” framework: scanning, shaping, building, showing, scaling. It includes his 2026 essay Le consultant et son double, on what AI does to the business model of consulting.
Through LinkedIn, at linkedin.com/in/durandchabert.