Langdock as usage-based AI platform

This reveals that Langdock is monetizing two different things at once, software access and underlying AI compute. A normal SaaS app mostly adds another user at near zero cost, but Langdock pays third party model providers when customers run chats, agents, and recurring workflows. That means revenue can scale faster than seats as automation spreads, but gross margin lands closer to an infrastructure tinted software business than a pure application layer.

• The product shape explains the margin profile. Langdock is not just a chat window for employees, it is a workspace where companies deploy reusable agents and workflow automations. Every extra task run can trigger model and infrastructure spend underneath, so cost of goods rises with usage.
  1 sacra 2 sacra
• That is a different model from classic enterprise SaaS and closer to AI platforms like OpenAI, where revenue is tied to inference volume, while still keeping a SaaS style control layer on top. The upside is larger contracts because spend can grow with process automation, not just employee count.
  1 sacra 3 sacra
• The closest internal comparable is Glean’s shift from search seats into agent building. Glean grew from $10M ARR in 2022 to $208M in 2025 as it expanded from search into no code agents and internal tools. Langdock is following a similar orchestration path, but with a stronger dependence on third party model costs.
  2 sacra 4 sacra

Going forward, the winners in enterprise AI will look less like 85% gross margin seat vendors and more like control layers sitting on top of expensive model usage. If Langdock keeps owning deployment, governance, and workflow design inside European enterprises, it can trade some margin for deeper spend per customer and much stronger product lock in.