Certus AI Review: Price, Fit, and Limits
Certus AI sells multilingual phone handling on a quote. What it does well, what a five-day setup costs you, and the one question that decides whether it fits.
On this page
Three quotes land in one week and each is a single number on a page. No per-minute rate, no implementation line, no per-store breakdown. Certus AI is one of those quotes.
That is not a fault in the product. Certus sells multilingual phone handling — orders, deliveries, reservations, upselling, complaints, in English, Spanish and French — to operators running several units and, in the vendor's own framing, drive-thru lanes. Anything built for a rollout gets sold by a person, not a checkout page, and the vendor describes implementation at around five days.
The trouble starts when you set that number beside a published plan. At 900 calls of two and a half minutes you are buying 2,250 connected minutes a month, and a quote for that line is only judgeable per minute. A quote for twelve stores is a different purchase wearing the same label.
TL;DR
- Certus AI is quote-priced restaurant voice AI that publishes orders, deliveries, reservations, upselling and complaints in English, Spanish and French (October 2026).
- Its real advantage is treating Spanish and French as shipped product rather than a roadmap item, which is worth money to a group with mixed-language callers.
- The vendor puts setup at around five days, so decide in advance who covers the phone while menus are configured store by store.
- No price and no payment path are publicly documented, which makes both sales-call questions rather than website questions.
- Divide any quote by your connected minutes: 900 calls of 2.5 minutes is 2,250 minutes, and $99 buys 750 metered minutes against $299 for 5,000.
What Certus AI actually is
A voice agent that answers a restaurant's phone in more than one language. The scope published on Certus AI's own site is wide: orders, delivery questions, reservation requests, upselling and complaints, in English, Spanish and French. The buyer in the vendor's positioning is the multi-unit group and the drive-thru operator, not the corner café with one line. On reservations specifically, the open question is whether a request lands in the book or arrives as a message for the host, and OpenTable's voice AI partner list is the fastest check on whether a booking claim is live for your plan.
Two things are missing from that public description, and they are the two a buyer needs first. There is no published price, because Certus sells on a quote. And there is no published POS list, so how an order reaches your kitchen is a demo question rather than a reading question.
Everything below is read off the vendor's public pages and treated as what the vendor says, not as a measurement. That distinction matters, because every ambitious number on these sites comes from the vendor's own customers.
Where it is genuinely good
Language as a product, not a setting
Any phone agent can answer in one language. Making Spanish and French first-class means the same voice takes the order, confirms the sides, and hears the complaint without handing the caller to whoever happens to be free. The complaint is the part operators underestimate: an angry caller in a second language is the call that pulls a bilingual server off the floor at 7:40, a cost you feel in the dining room, not on the invoice.
Upselling that belongs in the same conversation
Certus publishes upselling as part of the call flow rather than a separate campaign. On a $38 ticket, one added side on one call in five is a real monthly figure, and it is the difference between an answering service and a salesperson. Ask to hear it happen on a demo call, unprompted, with your own menu.
A setup that is a schedule, not a delay
Five days sounds slow beside the under-24-hours promise published elsewhere in the category (Loman, October 2026). For one shop that needs the line covered tonight, it is slow. For a group configuring menus, modifiers and store hours across locations, five days with someone doing the configuration with you is closer to a plan than to a delay. The question is who is doing the configuring and whether they have seen your POS.
What the quote has to itemise
One number is not a price. These lines turn a quote into something comparable.
| Line in the quote | Why it moves the monthly bill |
|---|---|
| Per-location fee | A group total divided by stores is not the per-store cost, because every line answers its own calls |
| Per-minute or per-call component | 2,250 connected minutes turns a small-sounding rate into the largest line on the invoice |
| Implementation and configuration | Five days of someone's time is paid for somewhere, usually at the front |
| Telephony and number porting | Forwarding is a setting; porting the number is a project with a date and a risk |
| POS write path | A kitchen ticket and a summary email carry the same price and completely different labour |
| Term and exit | Ask what happens to your number on the last day of the contract |
Phone orders into Toast walks that write path. Then do one division. Your connected minutes are calls multiplied by average length: 900 calls at two and a half minutes is 2,250 minutes. Every $100 of monthly price on that line works out at about 4.4 cents a connected minute. Run the same division on the plan you are comparing it with and the shape of the deal becomes obvious.
Why the line matters at all: a QSR Magazine analysis estimates U.S. restaurants lose roughly $20 billion a year to unanswered calls, and that the average restaurant misses about 150 a month. What missed calls cost a restaurant does the local version of that arithmetic with your ticket.
The count that decides whether to buy it
Before the demo, spend ten minutes in your own call log. Take the last 100 calls and tally two things: how many switched language, and what those tickets were worth.
Most operators guess the first number and guess it wrong in both directions: they remember the three calls that were hard work and forget the fifty that came in English and ended in a two-item ticket. If eight calls out of 100 switched language and six of those were group orders, the case for paying for multilingual handling is strong, because the calls you cannot staff are the calls with the most items on them. If it is two calls and two pizzas, multilingual is a comfort rather than a purchase, and a quote-based product is an expensive way to buy comfort.
Do the second half of the count too. The cost of a language gap is not the call. It is the order that never got written, which lands between $35 and $85 by the industry estimate.
Who it fits
Buy Certus if the phone arrives in more than one language across a group of stores and the language mix differs by location, which is the case where a single-vendor, single-language agent quietly fails one site out of five. It also fits a drive-thru or QSR operation where callers switch language mid-order and the staff on the headset cannot always follow.
It fits a group that wants implementation done with them. That is a real preference, not an apology: someone configures the menu, someone tests the ticket, and someone is accountable for the first live call.
Who should not
A single shop whose calls arrive in one language and end in one ticket should not buy this. You would be paying a quote-based price for coverage, an upsell script and a five-day build, and the money buys nothing the kitchen can see. The Certus AI alternatives list is the shorter path for that shop.
A shop that needs a published price before it will take a sales call should not buy it either. Quote-based is a legitimate way to sell to a group and a bad way to sell to somebody comparing three numbers on a Friday.
And a shop that needs the card charged before the caller hangs up has an unanswered question here: no payment path is publicly documented, so verify it on a live call rather than in a deck.
Where Kovo fits, disclosed
We sell Kovo, so weigh this section with that in mind. Kovo answers the phone, takes the order the way the caller says it, reads it back, and prints the ticket to Toast, Square, Clover, Olo, Epson Server Direct Print or Star CloudPRNT. It answers a website chat on the same engine, and books tables in its own reservation book rather than writing into OpenTable or Resy. It is $99 a month for 750 connected minutes or $299 for 5,000, with 14 days free and no card at signup.
Two limits worth stating plainly. Kovo does not publish a language list, so if language coverage is why you are reading about Certus, test Kovo on recordings of your own callers before you decide anything. And the card is a trade rather than a copy: Kovo charges a card the shop already holds, off-session, while the caller is still on the line, and a first-time caller gets a Stripe link with the line held open. No card number is spoken to the agent, so a stranger pays the link before the ticket is paid. The head-to-head lives in Kovo vs Certus AI, and there is a wider field in the Loman review if you want the flat-fee shape of this market too.
FAQ
Is Certus AI expensive?
Nobody outside a sales call can answer that, because Certus publishes no price and sells on a quote (checked October 2026). Judge it on connected minutes: 900 calls of two and a half minutes is 2,250 minutes a month, so a $400 quote is about 18 cents a minute and a $900 quote is twice that. Ask for the per-store number in writing before implementation starts, not after.
What happens to my phone line during the five-day setup?
The phone stays on your current arrangement until you cut over, so the vendor's stated five days is a build window rather than five days of downtime. Put the cutover on a Monday or a Tuesday and keep the old forwarding rule for a week as a fallback. Ask who tests the first live call and on which store's menu.
Does Certus AI take a card on the call?
No payment path is publicly documented, which is not the same as saying none exists. Put it in the demo script: order two items, say you want to pay by card, and watch where the number goes. If the answer is a link by text, ask what holds the kitchen ticket until the payment clears, because a printed ticket on an unpaid order is the failure that costs a Friday.
Can Certus AI book a table?
The vendor publishes reservation handling, so the next question is whether a request is written into your book or left as a message for the host. OpenTable keeps a list of voice AI partners for its Core and Pro products, and that list is the fastest check on whether a booking claim is live for your plan. Taking a reservation and writing it back into the system are two different capabilities.
Should a single-location restaurant buy it?
Only if the phone genuinely arrives in more than one language and a hand-held implementation is what you want. A shop taking 300 one-language calls a month is buying coverage and a five-day build to solve a problem it does not have. Count the language switches in your last 100 calls first: if the number is under 10, the case is thin.
Final Thoughts
The question that decides Certus AI is not "is it good". It is: can you hear it take your worst call in your second language, and can you get the monthly number in writing at your volume and per store before implementation begins? Two yes answers means the product fits you. A stall on either one means you are talking to the wrong shape of vendor, whatever the voice sounds like.
Count the language switches in your last 100 calls before you sit through a demo. Then name the purchase: a multilingual line with an implemented rollout, or a ticket printer at a metered price. If it is the second, take a test call with your own menu, run those same 100 calls through it, and compare tickets rather than transcripts.