Kea Review: Price, Fit, and Limits
Kea sells restaurant voice AI with order taking, built per restaurant and priced by quote. What it does well, what a build asks of your week, and who it fits.
On this page
The configuration form arrives before the demo. How do callers say "extra" versus "on the side", what should happen when the brisket is out at seven, and who picks up when the agent is stuck. Every restaurant-specific build needs those answers, and none of them live in the vendor's office.
Kea sells restaurant voice AI with order taking, built per restaurant and priced by quote (checked October 2026). A build is a genuine advantage over a template when your menu does not fit anyone's plan, and it is a project with your name on the homework.
That homework is the part no invoice shows. At a mid-volume shop, the owner's hours are the scarcest input in the whole exercise, which is why the honest question about Kea is not what it costs but what it costs you to configure.
TL;DR
- Kea sells order-taking voice AI with a restaurant-specific build, priced by quote, with no published figure as of October 2026.
- The advantage is fit: your modifier groups, your vocabulary, your exception rules, rather than a generic menu tree.
- The cost the invoice hides is your own time. Four kinds of knowledge come from your staff, and a build stalls without them.
- Normalise the quote to connected minutes: 750 minutes a month makes $250 a 33-cent-a-minute line, so ask for the number at 300 and at 900 calls.
- A published plan is the alternative shape: $99 buys 750 connected minutes with no build and no questionnaire.
What Kea publishes, and what it leaves to the call
Kea sells restaurant voice AI with order taking, built per restaurant and priced on a quote. The vendor's stated fit is the mid-volume shop, which is the honest place for a configured product: too much call variety for a template, not enough volume for an enterprise contract.
What is left to the call is most of what a buyer wants to know. There is no published monthly number, no published POS list, and no published payment path. None of that is suspicious for a quote-based vendor, and all of it means the review has to be about what to ask rather than what to read.
Treat the capability list as the vendor's description and test the pieces that matter on your own menu.
Where it is genuinely good
Order taking is the job, not a feature on a tier
A product built around orders starts with the parts that break orders: nested modifiers, half-and-half, a substitution the kitchen will accept, a caller who changes one line after the read-back. Kea's published position is order taking with a restaurant-specific build, which means the menu work happens with you instead of inside a template's limits.
A build is the right answer when your menu is not standard
A shop with a rotating specials board, a weekend-only item, or a modifier structure nobody else uses gets nothing from a generic agent. Someone sitting down with your menu and writing the rules is the difference between an agent that sounds right and one that tickets right.
Mid-volume is a real category
Shops in the 300-to-900-calls-a-month band fall between the published plans and the enterprise contracts. A vendor that quotes per shop can meet that band, and a vendor that publishes two tiers usually cannot.
The hour cost of the build
Four kinds of knowledge have to come from your side, and every one of them is time your staff does not have spare at 6 p.m.
| What the build needs | Who has it | What breaks without it |
|---|---|---|
| Menu and modifier structure | Whoever maintains the POS menu | Items ticket in the wrong group and the kitchen reads a puzzle |
| House vocabulary | The counter staff who hear it forty times a night | A regular's phrasing becomes a special instruction nobody recognises |
| Out-of-stock rules | Whoever works the line at 7:00 | The agent sells the item that ran out at six |
| Escalation path | The manager who answers the phone today | Calls land in an inbox nobody watches |
Put a number on the project before you sign: ask how many hours of your time the build takes, then look at the week you actually have. A shop whose owner is on the line five nights a week does not have ten spare hours, and a build that needs them will finish late, badly, or both.
Then normalise the price to the same units you use for anything else. At 300 calls of two and a half minutes you are buying 750 connected minutes, so a $250 quote is 33 cents a minute. The same $250 at 2,000 minutes is 12.5 cents. A quote that has not moved between those two volumes is a fixed price; a quote that doubled is a per-minute rate with a monthly label on it.
The question that decides it
Not "does it take an order". The deciding question is: who on your team owns this build after the demo call ends, and what happens to the agent when that person is on holiday?
A configured product needs an owner. That owner answers the vendor's questions, tests the read-back, watches the first week of tickets, and updates the menu when the specials change. In a shop with a manager who already owns the POS menu, that is a small addition. In a shop where the owner is also the buyer, the scheduler, and the Saturday host, it is a second job.
Four calls tell you whether the build landed.
- Order your most complicated item, with two changes, and read the ticket.
- Change one line after the read-back and see whether it re-reads.
- Name an item that is 86'd and see what the agent does with it.
- Ask for a person and time how long it takes to reach one.
The wider field, if you want to see how a published plan compares on the same four, is in best AI order taking for restaurants.
Who should buy it
A mid-volume shop with an unusual menu and a manager who owns the menu, which is the buyer this shape is built for. A shop that has already tried a template and found it misreads the modifier structure. A group of two or three sites that wants the same build applied across them with local menu differences.
It also fits a shop that wants a conversation about the product before a contract, and is willing to spend the hours to make the conversation useful.
Who should not
A shop that needs the phone answered this week should not buy it. A build takes a build, and the fastest path to a working line is a product that imports your menu and prints a test ticket the same day.
A shop with nobody to own the configuration should not buy it either. The Kea alternatives page walks the options that need less of you, and Kovo vs Kea puts the published price beside the quote line by line.
And a shop whose calls are mostly questions rather than orders should not buy an order-taking build at all. You would be paying for menu structure work that never pays for itself, when a QSR Magazine analysis estimates around 60% of missed calls are someone trying to order or book.
Where Kovo fits, disclosed
We sell Kovo, so weigh this section with that in mind. Kovo takes one store's order the way the caller says it, reads it back, and prints it to Toast, Square, Clover, Olo, Epson Server Direct Print or Star CloudPRNT. Starter is $99 a month for 750 connected minutes, Unlimited is $299 for 5,000, the trial is 14 days with no card, and a call under 10 seconds is never billed.
What it does not do is the build. Kovo imports a menu and runs a test connection that prints a $0.00 ticket, and the menu merge keeps your edits when the source refreshes — including a sold-out toggle, which outranks a source that still says available — but nobody sits down with your specials board. The order limit is yours to set rather than the vendor's: defaults of 10 items in total and 8 of one item, an optional value limit, and zero turns any of them off. Above a limit the call rings your handset if you turned that on, and otherwise the agent says the order is too large and keeps taking what fits. Payment runs two paths: the card the shop already holds is charged while the caller is on the line, and a first-time caller gets a Stripe link with the line held open.
FAQ
How much does Kea cost?
No price is published, so the only number that counts is the one in writing (checked October 2026). Normalise it before you compare anything: 300 calls of two and a half minutes is 750 connected minutes, so a $250 monthly quote is 33 cents a minute, and the same $250 on 2,000 minutes is 12.5 cents. Ask for the quote at 300 calls and again at 900, because a price that doubles with volume is a per-minute rate in a monthly costume.
What does a restaurant-specific build actually require from me?
Four things, and they all come from your staff: how the menu and its modifier groups are structured, how your regulars phrase an order, what the agent should do when an item has run out, and who picks up when it is stuck. Ask the vendor how many hours of your time the build takes and which of those four they write. Anything they need from you is a cost the invoice does not show.
Is Kea a good fit for one location?
The vendor positions itself at mid-volume shops, which is where a configured build makes sense over a template. The test is the all-in monthly number at your own volume, per location, with setup and support separated. If that number needs a two-year term to look reasonable at 300 calls a month, the shape is wrong for a single shop even if the product is good.
Does Kea take a card during the call?
No payment path is publicly documented, so treat it as an open question and test it live. Order two items, say you want to pay by card, and see where the number goes. Then ask what holds the kitchen ticket until the payment clears, since a ticket printed on an unpaid order is the failure that costs a Friday.
What happens on a call the agent cannot finish?
Ask who answers, how fast, and whether the context travels with the call. Send a complaint and a refund request through the demo line and watch where both land. A transfer that drops the caller into a general inbox at 8:40 on a Saturday is worse than a phone that rings twice and then finds a person.
Final Thoughts
Kea is the right shape for a shop whose menu does not fit a template and whose manager has the hours to sit through a build. For that buyer, the quote is worth the effort, and the four questions above are the whole due-diligence list.
If the honest answer is that nobody has the hours, buy the smaller thing and keep the phone answered this week. Count the calls you missed last month first, then take a test call with your own menu and see whether the ticket the kitchen receives is good enough before you commission a build.