Grounded over generative
Every answer cites a source — a menu line, a calendar entry, a known fact. If we don't know, we say so.
Production AI applications,
built on large language models.
We build reliable, grounded AI assistants for industries where saying the wrong thing has real consequences. One product at a time, in production, with operators.
Principles
What every product we ship has in common.
Operator-first
The business runs the business, not the AI. Our products earn trust by staying out of the way.
Cheap to try, fast to ship
Ten-minute setups. Two-week pilots. We pick proven primitives so we can spend our risk budget on the user experience.
Products
In production today.
More on the way.
Tableside
LiveAn always-on host for restaurants. Reads your menu, holds the table against your live calendar, and replies in the diner's language — 24/7.
- Menu intelligence
- Reservations
- 30+ languages
- 10-minute embed
Next
In designAn operator-first AI for the next industry where saying the wrong thing has real consequences.
We're hiring operators and engineers to figure out what ships in 2027. If you know a domain that's tired of being underserved by enterprise SaaS, write to us.
Team
Engineers who ship daily.
We're a small team based in London. Our founders have shipped production software at scale — payments, infrastructure, developer tools — and we're bringing the same operational discipline to AI assistants.
We pick boring infrastructure (Postgres, Qdrant, Qwen3 embeddings) so the product gets our risk budget. We talk to operators every week. We ship every week.