A concierge trained on your menu, your hours and your policies
Most guest calls ask the same handful of things: what time you open, whether there's parking, what's on the menu without gluten. A model trained on your own data can answer those directly, and your team spends the call on the guest in front of them instead.
Browse the guides and tools for venue ownersWhat to let it answer, and what to keep off it
A concierge trained on your own venue answers the questions that come in every night by phone and message: what time the kitchen closes, whether there's parking nearby, which dishes on tonight's menu are vegan, what the cancellation policy says. Those are questions with a fixed answer sitting in your own hours, your own menu and your own written policies, so a model trained on that material can give the answer back accurately and repeatedly, without a member of staff putting a table on hold to reach the phone.
There are two things it should never be asked to answer.
- Allergens as a guarantee. The Food Information Regulations 2014, enforced by local authority environmental health, require a venue to be able to tell a guest which of the fourteen allergens are in a dish, and that duty sits with a trained member of staff checking a current allergen matrix. A concierge can point a guest to the allergen information you publish and tell them to confirm with the venue before ordering; it should never state, on its own authority, that a dish is free from an allergen.
- Availability it cannot see. If the model has no live connection to your booking system, it does not know whether Saturday at eight is free, and it should say so. A wrong table offered with confidence does more damage than no answer at all.
The line is not about how fluent the model's answer sounds. It is about what it was actually given to know.
How it works, and what you own
Built on your own material
The concierge is trained on the venue's own information: the menu as it's written that week, the opening hours, and the policies you've actually decided on, such as whether dogs are welcome or how a deposit works for a table of ten. It doesn't draw on general knowledge about restaurants or hotels to fill gaps, because a guest asking about your venue wants an answer about your venue, not an average one stitched together from other people's websites.
Delivered through finetuning.uk
The technical build sits with finetuning.uk, the practice's sister service for trained language models, while the design and content work that feeds it stays with Vinrose. That split matters for a practical reason: the tool that talks to your guests and the menu or site that carries it are built by different specialists, and a venue owner shouldn't have to manage that handoff themselves.
Kept current
A concierge trained once and left alone drifts out of date as fast as any menu that never gets reprinted. When the kitchen changes a dish, a hotel adjusts its cancellation policy, or Christmas hours go up on the door, that change needs to reach the model's source data. Ask whoever runs the update for you how a change reaches the concierge and how long that takes. If the answer is vague, treat it as a gap to close before you rely on the tool for anything a guest will hold you to.
What you own
You own the source files: the menu text, the policy documents, the hours, and a record of what was fed into the model and when. That matters if you change agency, add a second venue, or simply want to check what the concierge was actually told to say. A tool you can't inspect or export isn't one you control, whatever it says to your guests in the meantime.
To talk through what a concierge would need from your venue before it went live, start a conversation, or see how it would sit alongside a website build and a menu it can actually read from.
Start a conversation about an AI concierge for your venue
Send us your menu, your opening hours and the policies you've already written down, and we'll tell you what a concierge trained on them could actually answer for your guests.