Reviews, emails and reservations handled with AI

For a steakhouse in Switzerland we built a platform where the restaurant replies to Google reviews, sorts its email and takes reservations in one place. Review replies are suggested by an AI model tuned to the restaurant’s voice, and a reservation from an email goes straight into the booking system.

Client
Swiss steakhouse
Industry
Hospitality
01 / The client

A restaurant whose guests expect top service online too.

The client is a steakhouse with a strong brand and two locations in Switzerland. Its guests expect the same level of service online as at the table.

They wanted one place for reviews, guest communication and reservations, with a German interface and as much automation as possible, but with a person always having the final word.

02 / The challenge

Three daily jobs that ate up the team’s time.

The team struggled with the same three problems every day.

  • 01Many Google reviews for two locations in several languages; replying by hand took hours, and the replies did not always sound the same.
  • 02One Outlook inbox for everything: reservations, group requests, supplier offers, job applications and complaints, all mixed together.
  • 03When a guest asked for a table by email, someone had to check availability and enter the reservation by hand.
03 / Our approach

Google, Outlook, OpenAI and the booking system in one flow.

01Reviews with AIReviews from both locations are imported automatically. A GPT-4.1 model fine-tuned on the restaurant’s own replies drafts an answer to what the guest actually wrote; a manager reviews it and publishes it in one click.
02AI inboxConnected to Microsoft 365. Every incoming email is sorted into one of six categories, from a single reservation to a complaint, and answered from the app with templates and the restaurant’s signature.
03Reservations from emailFor a table request, AI extracts the party size, date and time, also from phrases like “tomorrow” or “next Friday”, checks free slots in Aleno and books the table; staff only review the reply.
04DashboardViews on Google Maps and Search, calls, website clicks, direction requests and ratings, compared with the previous period and the same period last year.
04 / Results

One screen for both locations.

Reviews, guest communication and reservations for both locations are handled from one screen. A suggested reply is waiting for every new review, and a reservation that arrives by email is booked before anyone opens it. AI suggests, the team decides.

1screen for reviews, emails and reservations
6categories AI sorts every email into
2 hwindow around the requested time in which the system looks for a free table
05 / Technology

The technology behind the platform.

  • 01Backend: PHP 8.2, Symfony 7.3, Doctrine ORM, MySQL
  • 02AI: OpenAI: fine-tuned GPT-4.1 for reviews, GPT-4 Turbo for classification, extraction and replies
  • 03Google: Google Business Profile API and Business Profile Performance API
  • 04Microsoft and reservations: Microsoft Graph API (Outlook), Aleno GraphQL API
  • 05Frontend: Twig, Symfony UX (Turbo, Stimulus), ApexCharts, Summernote

Need a similar project?

Tell us briefly what you need, and we'll get back to you with a proposal and a free estimate.