About Hyaline Labs
Turning implicit structure into infrastructure
“Ask Maria, she’ll know.” That sentence — or some version of it — is the informal system of record for most organisations. When Maria leaves, the record goes with her.
This has always been true. Organisations have tolerated it because the cost was diffuse — a slower decision here, a new hire spending their first month mapping the org by hand there — and the friction, while real, never reached the threshold where fixing it felt worth the effort. That changed when enterprises began deploying AI agents at scale.
The moment
Today, enterprises are trying to deploy agents at scale — but what they find is that fragmented architectures and disconnected tech stacks mean agents become partial automations rather than truly agentic workflows. Humans can operate around these gaps because they understand the operating model: they know who to call, who can say yes, and how to manually bridge the steps that no system has encoded. Agents have none of that. They cannot read the room, navigate the organisation, or make the judgment calls that complete the workflow. Where humans fill the gap instinctively, agents stall.
The question isn’t who needs a better org map. It’s what happens when agents know everything on the internet — and nothing about how your organisation actually operates.
Atlas
Atlas is the system of record for organisational decision-making authority. A structured, always-current graph of roles and responsibilities — queryable by humans through a visual org canvas, and by AI agents through an MCP server.
We did not build another wiki or org chart tool. Those exist, and share the same flaw: they capture what someone wrote at the time, not what is true right now. Atlas is different in kind. Infrastructure has to be reliable — and agents, unlike humans, cannot fill the gaps.
Hyaline Labs
Hyaline means glass-clear, transparent all the way through — and that’s what we’re trying to make organisations: clear about how they deliver value, as a structured, queryable fact that persists when people leave, teams reorganise, and agents go to work.
We’re a team of European founders who believe the next decade of work gets decided by one question: can the organisations deploying AI actually make it work end-to-end? Right now, the answer is mostly no — not because the agents aren’t capable, but because the human organisation they operate inside is illegible to them.
Our mission is to change that. We’re building the layer that makes organisations readable — for the humans navigating them and the agents working within them. When that layer exists, AI stops being a partial automation and starts being a genuine collaborator.
We’re not building for the org chart problem. We’re building for the decade where human and machine workforces have to function as one.
