Engineering for hospitality
Tools.Context.Control.
The engineering wrapped around the model that turns it into an agent.
Powering 2,500+ UK venues
AI needs foundations.
Everyone’s bolting a chatbot onto their product. But an agent is only as good as the data it can reach and the tools it can use. On siloed data, with nothing to actually do, it just makes things up.
Tools, context, control.
The three things an agent needs to be useful instead of a novelty, and the parts we’ve actually built.
It reads the graph
The agent sits on one connected graph of your guests, so it answers from what actually happened, not a generic guess.
It can actually do things
A real toolset: query the graph and your databases, browse the web, and build a report as a shareable HTML artifact.
On a leash
Runs that survive and pick up where they left off, work that carries on in the background, and guardrails on every step.
The graph.
The foundation the agent sits on: not rows in a table, but your guests and everything they do, joined by the relationships between them.
Bookings, spend, Wi-Fi, reviews and loyalty on one record, joined by relationships, not stored in nine separate databases.
Real relationships like who dined with whom and who referred whom, the social graph most tools never build.
Churn risk, lifetime-value tiers and influence worked out on top of the raw data, ready for the agent.
A question like “regulars referred by a VIP who haven’t been back” is a single traversal, not a week of SQL.
The agent’s tools.
Not just a chat box. A real set of tools the agent can pick up to actually get something done, with a leash on every one.
Graph queries
Traverse the connected guest graph to answer questions nothing else can.
HTML artifacts
Build a dashboard or report as a self-contained page, ready to share.
Query your data
Read across the service databases directly, safely and read-only.
Browse the web
Open a live browser session to look something up or check a source.
Heartbeats
Recurring check-ins on a schedule, delivered to chat, email or Slack.
Background jobs
Kick off heavier analysis that runs off to one side and reports back.
Memory & context
It remembers the thread and grounds every answer in your real data.
Guardrails
Read-only by default, writes need sign-off, and secrets stay secret.
Meet the agent.
All of this powers a general CRM agent, wired into the connected graph and the full toolset, so it can actually get work done rather than just chat.
- It sits on your data
Because it reads the connected graph, it answers from what actually happened, not a generic guess.
- It analyses and reports
Ask it to dig into your numbers and it comes back with the working, built as a report you can keep.
- It runs and doesn’t drop
Durable runs survive and resume, and heavier work carries on in the background instead of timing out.
- It stays on a leash
It reads and drafts; anything that changes your data waits for your sign-off. No surprises.
The proof.
The foundation is real and built, not a promise on a slide. Numbers, not adjectives.
The hard part, done first.
Connected data, a real toolset, and the foundation everything else compounds on.
A graph, not a guess
The agent sits on a graph that joins every guest, visit, booking and review by the relationships between them, so its answers are grounded in what actually happened.
It can do, not just chat
Query the graph and your data, browse the web, build an artifact, run in the background. The tools are the difference between a useful agent and a novelty.
Built to compound
We put the hard part in first, the connected data and the harness, so every new capability plugs into the same foundation instead of being another bolt-on.
Frequently asked questions.
Built on the right foundations.
See the connected graph and the agent that runs on it, and where AI is really headed for hospitality, in a live walkthrough with our team.
Book a demo