Phone orders are still a real slice of revenue for most restaurants, and if you run Epos Now, the only question that matters about a voice agent is simple: does the order end up in Epos Now correctly? A great-sounding voice that drops a modifier or sends the kitchen the wrong ticket is worse than a missed call.
Here's how X1 Voice works with Epos Now, where these setups get bumpy, and what to check before you buy.
The order has to land in Epos Now, cleanly
X1 Voice connects to Epos Now through our Deliverect integration. It reads your live menu, prices each phone order against it, and routes the finished ticket into Epos Now so it lands the same way an in-person or online order would. Nobody re-keys anything. The value is in the round trip: reading the live menu and firing a correctly-modified ticket, not just sounding human on the call.
The one thing to test with any vendor, Epos Now or otherwise, is a deliberately awkward order. Try "no onions, sub fries, add a drink, apply the coupon" and confirm it arrives in the POS right, not just that the voice repeated it back. Reading an order back and writing it down correctly are two different jobs, and only one of them shows up at the pass.
What changes on a real shift
The visible change isn't the call. It's what stops happening after the call.
Right now a phone order at 7pm costs you a person for ninety seconds of talking and another thirty of typing at the till, and both of those happen while someone is standing at the counter waiting. With the connection running, the ticket is already in Epos Now before the caller has hung up, and the person who used to walk to the till stays where they were. On a night with forty phone orders, that is close to an hour of counter time that was previously spent transcribing.
The quieter benefit is that the ticket text is your own. It came out of your product list, not from someone's handwriting, so the kitchen reads the same words it reads on every other order.
Where Epos Now setups actually break
Rarely at the connection. Usually in the product list.
Epos Now catalogs grow by addition. Someone adds a product for a special, it never gets archived, and two years later the list has three versions of the same burger at three prices. On a till that's an annoyance because staff know which button to press. On a phone call it's genuinely ambiguous, because the caller says "cheeseburger" and there are three of them.
Variants are the second trap. If your sizes or options are built as separate products rather than as variants of one, the agent has to treat them as separate items, and the spoken order gets longer than it needs to be. Worth checking before launch:
- Archive products that are no longer sold, rather than leaving them inactive but present.
- Give each item a name a caller would actually say, keeping the till abbreviation as an alias.
- Collapse near-duplicate products into variants where your setup allows it.
More on the naming half of that in menu naming for voice clarity, and on the option-handling half in how voice AI handles menu modifiers. A clean product list makes this a same-day project. A messy one makes the cleanup the project, and it was overdue anyway.
Payments and reporting
Callers can pay by card on the phone for pickup and delivery. That payment reconciles alongside your other tender types, so your end-of-day numbers stay in one place. Staff never read a card number out loud or write one down, which removes the notepad by the phone that was always the weak point.
You can mix the policies. Most operators end up taking payment on the call for delivery and for anything over a set amount, and leaving small collection orders as pay-on-arrival, because that is where no-shows actually cost kitchen time. How the card data is handled is covered in collecting payment over the phone safely.
Discounts deserve a decision of their own. If your Epos Now setup has staff discounts or promo buttons that a manager applies with judgment attached, decide before launch which of those a caller can trigger on the phone and which need a person. The safe default is a short allow list rather than access to everything in the catalog, and it is much easier to widen later than to claw back.
Setup, and what actually takes the time
Connecting Epos Now takes minutes, not a six-week project: link your account, import the menu, set your hours and any phone-only rules, run a test call, and go live. Most single-location restaurants are live the same day, and our team can handle white-glove setup if you'd rather.
The slower part is the decision list, and none of it is technical. Which calls always reach a person. Whether you quote a real pickup window or a flat time. What happens to a caller asking about a booking. What the agent does at 30 minutes to close. Ten minutes of thought here saves an awkward first week. The onboarding checklist is the version we work through with new sites.
If you run more than Epos Now across locations, that's fine too. X1 Voice reaches 80+ POS systems, directly or through Deliverect, so a mixed estate isn't a blocker.
The numbers to check in your first two weeks
Four, all from your own reporting rather than a vendor dashboard.
Answered-call rate against the same weekdays before the switch. Order accuracy, measured by pulling twenty phone tickets and checking them against their recordings instead of trusting a percentage. Average phone ticket, segmented by channel so you're reading the phone specifically. And the number of tickets that landed in Epos Now but needed a correction before they fired, which tells you whether the product-list cleanup went deep enough.
That fourth number should approach zero inside a fortnight. When it doesn't, the cause is nearly always one or two named products, and fixing them is a ten-minute job once you know which. The accuracy measurement method explains how to sample tickets so the figure means something.
The decision rule
Run one full week, then compare two counts: calls answered, and phone tickets that reached the kitchen with nobody retyping or correcting them. If both sit close to your total inbound phone volume, the integration is doing its job and the remaining questions are about pricing and call policy rather than about Epos Now. If either lags, you have a specific, named list of products to fix, and you can decide whether to keep going with real evidence instead of a hunch.