Skip to main content
HospitalityBirmingham, UK · 4 months in production
Miners Arms logo

Miners Arms

Two venues, one manager, and a booking system charging £500 a month that still let tables double-book.

£68,500

annual cost removed

Manager time on bookings
12 hrs a weekunder 1
£45,240a year of work running without a hire
  • Trigger.dev
  • n8n
  • Supabase
  • Telegram
  • Google Calendar
  • Google Business Profile
  • OpenAI
  • Gemini
  • Vercel

Where they were

Two Birmingham venues run by one operations manager. Bookings arrived by phone, by email, through a third-party platform and across the bar itself: four places to look, none of which knew what the others had promised.

What was broken

The booking platform cost £500 a month and still allowed the same table to be sold twice, because nothing reconciled the four sources. Reception cover ran eight hours a day, every day across both venues, simply to answer and record bookings, and the manager spent roughly 12 hours a week on top of that copying them between systems and phoning guests to apologise. Reviews went unanswered and neither venue posted anything.

What we built

  1. InPhone, web, email, walk-in
  2. BuiltOne booking queue
  3. OutNo slot promised twice
  1. A website for both venues, built as the front door the group did not previously have, with reservations taken on it directly rather than handed off to a third party.

  2. One booking queue. Phone, web and email requests land in a single stream and are written to a shared calendar, so every venue is looking at the same availability.

  3. Double-booking made structurally impossible rather than merely unlikely: table writes are queued through Trigger.dev, so two requests for the same table cannot both succeed.

  4. Reservations pushed straight to the team's Telegram as they arrive, so the venue sees a booking without anyone opening a dashboard.

  5. Guest records held in Supabase: visit history, preferences and repeat-guest recognition, so the second visit is not treated like the first.

  6. Confirmations and reminders sent automatically on a schedule, in the guest's language, with a one-tap cancel that returns the table to availability immediately.

  7. A daily covers summary to the manager each morning: what is booked, what changed overnight, and which tables are still open tonight.

  8. A content operation the group did not have before: posts planned, drafted, held for approval and published on a schedule, rather than remembered when someone had a quiet afternoon.

  9. Review handling that replies to guests automatically and escalates anything low-starred to the owners directly, so a bad night is seen the same day rather than a fortnight later.

  10. The £500-a-month platform retired, its data migrated, and its features replaced by the parts of it the venues actually used.

What it did

Measured after go-live, over 4 months.

Annual cost removed across both venues
£68,500
Of that: reception cover, 2,920 hrs at £15
£43,800
Of that: booking software fees, a year
£6,000
Manager time reclaimed, a year
624 hrs
A year of work running without a hire
£45,240
Of work handled a year, all sources
6,560 hrs
Booking accuracy over 4 months live
100%
From first call to live system
11 days

"We cut our software bill by £500 a month and stopped paying for reception cover on top of that. It now takes every booking, answers our reviews and runs our social, and that is work we were either paying for or simply not doing."

Yin, Operations Manager · Miners Arms

This build is our AI process automation proof point. See what we build for AI process automation.

Book a Free Strategy Call
Free Strategy Call

Book a free strategy call

Pick a time that suits you. Forty-five minutes to work out where automation would actually pay off, and what it would cost to build.

45 minutes · Free · No commitment

Pick a time that suits you

You will speak to Prakhar, who builds these systems, not a salesperson. Choose a slot and the calendar invite arrives straight away.

Loading the calendar…

Open the scheduler in a new tab if it does not load here.

What happens on the call

  • 45 minutes with Prakhar, our founder, not a salesperson.
  • You leave with a costed ROI forecast and a recommended build.
  • No commitment. If it would not pay for itself, we will say so.

No commitment on this call. Fixed price agreed before we start. Half up front, half when it is live and working.

Your records stay in your own systems: your calendar, your CRM, your database. AI processes what passes through and retains nothing between requests, and everything runs on commercial API contracts where your data is never used to train public models.