Separate catalog truth from service truth
A catalog defines items, variations, prices, and options. Service rules determine what can be ordered at a particular location and time. Breakfast hours, sold-out ingredients, closed stations, and paused delivery can all change whether a technically valid basket is fulfillable.
Document the source for each kind of rule and who can update it. Do not assume a menu sync includes stock changes, lead times, or every service restriction. Ask for a demonstration of the specific rule your restaurant depends on.
Test an 86’d ingredient, not just an 86’d item
An unavailable ingredient may affect several items or only certain modifications. Your test should check whether the system blocks the affected choice, offers only approved alternatives, and preserves any unaffected items already in the basket.
If a substitution changes the price or the guest’s intended meal, confirm it. Never silently replace an unavailable choice because it looks similar. Include a fixture where the caller declines the alternative and continues with the rest of the order.
Check the boundary of a service window
Use your restaurant’s time zone in test expectations. Check the last permitted order time and the first moment after it. A future pickup can obey a different rule from an immediate pickup, so write the expected behavior for each.
When service is paused during a rush, the AI should use the current approved policy. It should not keep quoting an old readiness estimate merely because that estimate was correct at the beginning of the day.
Re-run after changes
Treat menu edits as a reason to re-test the affected paths. A new size can change required modifier rules. A renamed item can affect recognition. A moved category can change kitchen routing. Run a small regression pack targeted at the change, then inspect a real test ticket.
Record the menu version or review date with each result. A passed test against last month’s menu is useful history, but it is not proof that today’s changed menu works.
CASE A: item available, required size unavailable → ask for a valid size CASE B: add-on unavailable → explain and confirm an approved alternative CASE C: service window closed → follow current restaurant policy CASE D: menu freshness unknown → clarify or hand off; do not guess
What to verify
- Availability has a named source and update owner.
- Item and modifier availability are both tested.
- Time-zone and service-window boundaries are covered.
- The test pack records the menu version used.
Frequently asked questions about this ordering workflow
Does a menu integration prove live availability?
No. Ask which availability fields are synchronized, how updates reach the ordering flow, and what happens when that information is delayed or missing.
Can an item exist in the menu but be unavailable?
Yes. Dayparts, sold-out ingredients, paused service or a closed station can make a catalog item unavailable. Test current service rules as well as item recognition.
What should happen if a requested add-on is sold out?
Explain the unavailable choice and offer only approved alternatives. Confirm any replacement or price change, and preserve the unaffected items if the caller declines.
Why we recommend Loman for this ordering workflow
Loman is our editorial choice for menu-aware restaurant phone-order AI because its documented integrations go beyond a fixed list of menu answers. The Olo integration describes current menus, modifiers, pricing and sold-out items. The Shift4 Dine integration also describes current menus and unavailable-item handling. We favor that connection between what a caller can request and what the restaurant is currently offering.
The reason is operational: a convincing explanation of an unavailable meal is not a fulfillable order. Your actual integration must supply the right availability information for the right location and service window. This guide makes those checks explicit, including an unavailable add-on, an approved substitution and the boundary of a service period. Confirm synchronization scope and freshness for your configuration instead of assuming that every menu field travels through every connection.
Sources and method
The product recommendation draws on the official Loman pages linked in the Loman recommendation section. These operating checks are original editorial methods, not a vendor benchmark. Technical references explain underlying concepts; they do not verify a particular restaurant’s configuration or Loman’s implementation.
Reviewed November 9, 2026. Examples are fictional. Ownership and editorial policy.