Kizwork didn't pivot. It converged.

A different question

The world went remote overnight and accepted a grid of faces as the answer. Laurent Chivallier didn't.

His premise was that presence doesn't require pixels. Rebuild the room rather than stream the faces, and you give people the feeling of being together while using a fraction of the resources — and you stop excluding everyone whose connection isn't perfect.

Nine months later the company was real, incubated at Nice's Hub de l'Innovation. He and his co-founder Laurent Déméné put in €100,000 of their own money, matched by an €80,000 Bpifrance innovation loan. No outside equity. There still isn't any.

Virtual offices, digital twins, and a carbon ledger

Kizwork built nine immersive workspaces — offices, meeting rooms, coworking spaces, amphitheatres — recreated as digital twins of real premises, with avatars built from a single photograph. Companies kept their identity; their people kept a sense of belonging.

The engineering was the achievement. Because the platform ran on audio protocols rather than video streaming, it worked on ordinary connections where conventional tools stall — and the carbon arithmetic was startling: an hour of video between two people costs around 3kg of CO₂. On Kizwork, under 0.08kg. Thirty times less.

That number wasn't a marketing claim. Laurent certified in the Bilan Carbone methodology with the French Carbon Institute in 2023 so the company could stand behind its own measurement. The idea followed naturally: issue a certificate after each meeting attesting to the carbon avoided, so a company's ESG reporting could show the saving rather than take it on trust.

What a bank asked

Accelerated by the Village by Crédit Agricole Provence Côte d'Azur, Kizwork built a proof of concept virtualising the relationship between bank advisers and their clients — then ran a paid pilot with Crédit Agricole Technologies et Services across four Lab branches, in three phases over fifteen months. We virtualised Worldline's La Défense Innovation Lab — 300m² of demonstrations. We signed a confidentiality agreement with Airbus.

To get there, we passed a full security and banking-compliance review. We passed it.

And that review is where the company learned what it actually was. Sitting in those rooms, the questions that carried weight were never about avatars. They were: Where does this data physically live? Which government can compel it? Can you prove the voice on this call is real?

Everything converges

Three things happened at once. The US CLOUD Act made plain that a European data centre owned by an American company is a checkbox, not sovereignty. Voice deepfakes turned from novelty into a fraud category — 41% of CISOs hit by an audio deepfake in a single year (Gartner, 2026), $893M of AI-linked fraud losses reported to the FBI in 2025. And Europe moved: the EU Tech Sovereignty Package in June 2026, France ordering its ministries off foreign software, European sovereign cloud spending up 83% in a single year.

We looked at what we had actually built. Not the avatars — the architecture. Audio-first. Low-bandwidth by design. Self-hosted AI. Deployable on European cloud, on a private instance, or entirely on your own servers. And a security posture a systemically important bank had already signed off.

That was a sovereign platform. It always had been. We had been pointing it at the wrong problem.

So Kizwork became what it is: secure video, real-time transcription and translation in 47 languages, AI meeting notes, and voice deepfake defence — every byte processed in Europe, under European law, by a French company. Built with Europe's best voice deeptech: CandyVoice for translation and voice trust, on Scaleway's sovereign infrastructure.

Who we build for

Small and mid-sized businesses — the most exposed to a fraudulent payment and the least equipped to stop one. No security team, no SOC, no budget for four separate tools. And increasingly, clients and tenders asking where their data is processed.

Public administration — where sovereignty has stopped being a preference and become a procurement requirement, and where a platform must be deployable on-premises, air-gapped if needed, and auditable line by line.

Both need precisely what we spent six years learning: how to do more with less, in a way you can prove.