AI for restaurants, pubs and hotels: what to automate first in 2026

Booking replies are the first build for most venues, because the same questions arrive by phone, email, Instagram and WhatsApp at all hours, and the booking system already holds the answers. The build reads each enquiry, checks availability, offers a time and books it. It passes private hire, groups above the venue's limit and anything it cannot answer to a person, with the conversation attached.

For the phone, an AI receptionist can take bookings during service and put everything else on a callback list. The AI receptionist cost guide compares UK plans and custom voice agents.

In the government's AI adoption survey of 3,503 UK businesses, published 29 October 2026, 23% of hotel and catering businesses had adopted AI, up from 11% in 2025. Hotel and catering also had the highest share of businesses neither using nor planning to use AI, at 73%.

Which booking and EPOS tools an AI build works with

An AI build works best when it can read and write bookings directly. As of 7 November 2026:

• ResDiary lists a Consumer API for custom booking journeys, a Data Extract API for customer and booking data, an EPOS API and webhooks that fire when a booking changes. Venues request access through ResDiary's support team.
• OpenTable says API use must follow "your applicable agreement with OpenTable", and partners apply through a form. Confirm access before a build is scoped.
• SevenRooms says it has more than 100 integrations and an API, and lists about 40 EPOS systems it connects to, including Lightspeed, Toast and Zonal.
• Lightspeed publishes a K-Series REST API that covers orders, menus, daily sales data and reservations for booking platforms.

Where a tool has no open API, a build can still start from the booking confirmation emails each system sends.

An AI can share a venue's written allergen information, but it should never decide whether a dish is safe for a guest. The business is legally responsible for accurate information on 14 allergens, and the FSA's best practice is written information backed by a conversation with trained staff.

What the law and the FSA say

• Non-prepacked food. The FSA's allergen guidance, updated 17 July 2026, says a restaurant or takeaway "must supply allergen information for every item that contains any of the 14 allergens". Staff can give it verbally, with a written notice in a clearly visible position.
• Online and phone orders. The same guidance says allergen information must be provided "before the purchase of the food is completed" and again "when the food is delivered".
• Staff training. "Food businesses must make sure that staff receive training on allergens."
• Natasha's Law. Since 1 October 2021, food prepacked for direct sale, such as sandwiches made and wrapped on the premises, needs a label with a full ingredients list and the allergens emphasised. The law is named after Natasha Ednan-Laperouse, who died after eating a pre-packaged baguette that did not then need ingredients labelling.
• FSA best practice. The FSA's best practice guidance for England, Wales and Northern Ireland says "You are not required by law to follow best practice guidance", but "Businesses must ensure that all mandatory food allergen information is accurate." It tells staff to refer to the written allergen information, and says that for pre-ordered group bookings allergen information "should also be discussed with individuals on the day".

Rules for an AI allergen flow

• Answer only from the venue's current allergen matrix, never from the model's general knowledge or a guessed recipe.
• Quote the matrix entry and the venue's cross-contamination statement, then hand the guest to a trained member of staff.
• Ask whether anyone has an allergy before confirming any phone or online pre-order, as the FSA's best practice suggests.
• Write every allergy on the booking so it reaches the kitchen, and flag it for a check on the day.
• Update the matrix first when a recipe or supplier changes, so the bot reads the new version.

Rota and ordering are back-office jobs where AI drafts and a manager approves, because both commit money.

Rota admin

The build collects staff availability by WhatsApp or the rota tool, drafts next week's rota from forecast covers in the booking system and chases swaps and sickness cover. The manager approves the rota and keeps control of hours and pay.

Supplier orders and invoices

The build reads daily sales from the EPOS, such as Lightspeed or Square, and the latest stock count. It drafts each supplier's order against the venue's par levels and sends it only after the chef or manager approves. When deliveries arrive, it reads the delivery note and the invoice, flags short deliveries and price changes, and queues approved invoices for the accounts system. The AI document processing guide covers how invoice reading works and where people still check.

Yarigai is a done-for-you AI agency in London. It builds lead capture and booking flows, support agents that answer from a client's own documents, inbox triage, invoice reading and filing, and receipts into Xero or QuickBooks. For a venue, that means building inside its own booking system, inboxes and accounts. The venue owns the code, the data and the accounts from day one, and there is nothing to sign up for.

A project starts with a free 30-minute scope call on Google Meet, and the venue keeps the notes whether it hires Yarigai or not. A blueprint of about a week, credited to the build, sets out a written plan, a clickable prototype, the monthly running cost, a fixed price and the payback maths. The build takes 2 to 6 weeks with a demo of working software every week. A single workflow, such as booking replies, can go live in about two weeks.

A person approves anything that sends money, emails a customer or deletes data. AI and hosting bills are paid by the venue at cost, and the monthly running cost is on the quote. Yarigai's guarantee is one sentence: "If the build isn't profitable for you, we won't build it."

Yarigai has no published hospitality case study. Its customer support and booking flow work is described on the services pages.