Questions and answers

Not upfront, and we would not ask you to. It starts by preparing replies for your team to approve, so for the first week or two you are reading what it would have said before any guest sees it. Once a type of reply has been tested and your team is happy with it, you can have that one sent automatically, one type at a time, for as much or as little as you are comfortable with. Every step is logged, so your team can see what happened and why. You can see every voice transcript and every chat agent message. We run judges across every AI activity and report on the outcomes, so you are alerted to anything that needs a closer look.

Before anything goes live, you decide what it can handle and what always comes back to your team. Allergies, complaints and anything involving money always go to a person. Anything new or unclear waits for your approval, and you can send it, change it or leave it. We set those categories with you before it goes live, and change them whenever you ask.

Your venue chooses whether guests are told up front that they are talking to AI, or only when they ask. Telling them up front is on by default: when the agent answers your line, its opening sentence says it is an AI assistant working with your team, and while that setting is on, an opening line that does not say so is refused and never spoken. In chat and email you make the same choice. Whichever you choose, if a guest asks, it says so every time, and on the phone a guest can always ask for a person and the call is transferred. It never pretends to be human.

The AI does not answer them. A dietary or allergen question is handed to your team with no AI answer given, on the phone and in writing, rather than paraphrased from something it has read. Serious complaints and anything with a legal edge are treated the same way.

With an hour. No slides, no pressure, and it is enough to see whether this fits your venue. If it does, day one is a sit-down at your venue where we take a baseline of the numbers as they stand today and pick one target worth moving.

About five hours of your total time over two to three weeks, in short slots, never during service. That is the sit-down on day one, two short check-ins while it runs, and 45 minutes at the end to walk through the report. You get your own locked-off login and can watch everything from day one.

Yes. We run a free 48-hour check of your venue's booking paths, links, search results and socials, from public data only, with no access to your systems needed. You get three to five findings in plain language, and the report is yours either way.

Six things are built today: answering the phone, keeping every enquiry together, responding to event enquiries faster, keeping on top of reviews, understanding where your bookings come from, and letting guests book directly. Beyond those there is a library of ideas across the kitchen and operations, guest experience and language, front of house and bookings, back office and finance, and your team. Use them individually, or combine them into the mix that suits your venue.

The phone agent is live and answering calls. Phone, chat, web and email enquiries come together in one view today, and a summary of every call lands beside them; WhatsApp is in early access and Instagram is coming. On reviews, what is running today is the watching: a daily check on your Google listing that tells your team something has changed. The grading and the drafted replies are a mock-up we would build with you. If you want the full list of what AI could do in your venue, that is our use-case library.

Yes, and not as a staged recording. We run the voice agent with you on a call so you can bring your own questions, including the awkward ones. There is no open demo line on the site: a number anyone can dial is a real cost and a real abuse risk, so we do it with you instead. For enquiries you can try the live chat on the site, which is our own account rather than a picture of what it could look like.

OrdersUp hosts its systems in Singapore, and your venue's data is walled off from every other venue. Some services we rely on process data outside Singapore, including the AI service, the voice service and the WhatsApp Business Platform; where that happens we require them to protect it to a standard comparable to the PDPA, as the PDPA requires for overseas transfers. Every sub-processor is named in the privacy policy, your keys are encrypted, and anything that fails, fails shut rather than guessing.

No. Under the commercial terms we use, the AI provider does not train its models on what we send or on what it returns, and the content we send is not retained by default. Where anything is held it is deleted within a short window, up to 30 days, except where the law requires otherwise. We send it only what it needs for the task in front of it, not a venue's whole guest history.

See it, export it or have it deleted whenever you ask. We hold it to a published retention schedule: enquiries 24 months, message contents 12 months, voice transcripts 6 months, quality-check data 90 days, and guest records up to 36 months of inactivity. It is a published schedule rather than a per-venue setting, so it is the same for every venue.

Pause it whenever you need to. Your enquiries still reach your team exactly as the guest sent them, and your team takes over immediately. You also get a kill switch you control, one action and under a minute, and a daily list of your bookings and open enquiries by email whether or not the system is running. If it ever stops, you carry on exactly as you do now.

Every night at 2am Singapore time your database backs itself up, and each backup is checked as it is written, so it cannot quietly save the wrong one. We keep 30 days of them on the machine that runs your service and 90 days in a separate Singapore location. We also take a fresh backup before we deploy any change, and we do not deploy without one. In plain terms, the most you could lose is what was entered since 2am that morning. We have restored real production data from one of these backups and confirmed it came back correctly. Being back inside four hours is what we aim for rather than something we guarantee: we have not yet rehearsed a full rebuild end to end, and we would rather tell you that than let you assume otherwise.

Two things are true, and one thing is not. Your enquiries do not disappear: email keeps arriving in your inbox and is picked up when we are back, so nothing is lost, it is just late. And you carry on the way you did before us, because your own booking process is the backstop, by design rather than as an excuse. The thing that is not true is the web: a guest trying to book online in that window gets an error page and cannot book, which is the part we like least and will not talk around. We do not run a public status page, and you will not find an uptime figure on this page. We do not offer a guaranteed service level, and what we do commit to is in your agreement rather than on this page.

Once you are live with us: routine questions go to email and we answer them the next business day. If something is service-critical you have a direct line to Todd. From Thursday to Sunday, 18:00 to 24:00 Singapore time, we aim to acknowledge within about an hour and to resolve it that evening; outside those hours it is the next morning. Those are the aims we work to, and the support schedule that applies to you is the one in your agreement rather than this page. You would hear from us before you had to chase us.

Yes, as part of your build. The phone agent is live and answering calls, and putting it on your own line is one of the things your build does. The technical set-up is ours and the settings are yours: the opening line, including whether it tells guests up front that it is an AI (it does by default), the booking terms it reads out, whether a phone booking waits for your team to confirm it, and how long it holds a late table. What we need from you is the number you want forwarded to it and the person who approves the opening line.

We are new, and we have no case studies yet. In return for your time, your feedback and an honest assessment of whether it works, we cover the cost of the build. What we get back is reference customers who can say out loud whether it worked. That is the whole trade.

Each build is a focused set of sessions over a few weeks between your team and ours: we start with a real operational challenge in your venue, establish a baseline, then build or configure a practical AI solution around it. A build ends with one thing built specifically for you, working in your business, and a before and after against the day one baseline, whichever way the numbers point. A design partnership is a series of short builds with an operator, letting us co-design how modules work together to add greater value and measure the impact. In both cases the build or design partnership period is not subject to subscription costs, and you lock in reduced fees for 12 months if you continue beyond it. Our goal is to provide solutions that work for operators, and the best way to do that is for us to design them together and prove them on real venues.

Angie looks after Singapore and is the person you will meet. Todd builds the product: twenty-five years in software, most recently leading a global team of 650 consultants building and testing AI systems, and he still writes the code himself. Not a form and not a queue. You get Angie or Todd.