Restaurants, cafés and delivery
The voice agent that takes the restaurant phone at peak hours
Where is the order, how much longer, can the address be changed, what's in a platter. Connected to Syrve (iiko), the agent finds the order by the customer's number and answers from real data. The team stays in the kitchen and on whatever needs a person.
Interactive demo
Watch a restaurant call, step by step
Pick a scenario. On the left, the conversation; on the right, what gets written in Syrve and what the team receives after the call.
- AgentGood evening! You've reached the restaurant's virtual assistant.
- ClientI ordered 40 minutes ago. Where's my order?
- AgentOne moment, I'm looking up the order by your number.
- AgentIt's with the courier. The estimated delivery time is 20:40.
- ClientPerfect, thank you!
The order in Syrve
- Lookup
- Order found by the caller's number
- Status
- With the courier
- Estimated time
- 20:40
After the call
- Transcript and recording saved
- Summary written after the call
- Reason: order status
Illustrative simulation, without sound: the dialogue is an example, but every action in the “What the system sees” panel exists in the platform today.
Want to hear a real voice? Book a demo70+
languages the agent can speak, for customers from any country
Syrve (iiko)
Ready-made integration: orders, menu, confirmation calls
Phone + website
The same agent on the ordering line and on the voice button on your website
In short
What is an AI voice agent for restaurants?
An AI voice agent for restaurants is a virtual assistant that answers the restaurant's phone. With Kallina, the agent finds the order in Syrve (iiko) by the customer's number, answers questions about the menu, opening hours and delivery zones, confirms website orders, takes reservation requests and hands over to a person anything that needs a decision.
- Ready-made integration: Syrve (iiko). r_keeper and Oracle Micros are coming soon.
- Channels: phone (your existing number forwarded, or a new number) and the voice button on your website.
- Languages: more than 70, supported by the voice engine, tested on your case.
- After every call: recording, transcript, summary and the reason for the call.
- Price: monthly credit packages, plus implementation paid once.
The problem
Peak hour, seen from the phone
In the evenings and at weekends, the same questions come in exactly when the kitchen and the operators are busiest.
The phone rings over the orders
Whoever is packing or serving can't also hold the phone. Calls go unanswered or the customer waits on the line.
The same questions, all evening
Where is my order, how much longer, until what time do you deliver. The answer is already in the system, but someone has to look it up.
Changes get lost on the way
A corrected address or a note for the courier, said in a hurry on the phone, barely makes it onto the order and even less to the kitchen.
Call types
Which calls the agent handles
Six calls that repeat in every restaurant with delivery or a dining room. For each one, you decide what the agent may do and when it hands over to a person.
Where the order is and how much longer it will take
The agent recognises the caller's number, finds today's order in Syrve and says what stage it is at: received, being prepared, ready or with the courier, with the estimated time when there is one.
- Lookup by phone number, no codes to read out.
- Stage, address and estimated time, straight from Syrve.
- Transfer to an operator when a delay needs a decision.
Address, time, items, comment
On the confirmation call, the agent applies the changes you allow directly in Syrve: it adds an item, changes the address, the delivery time or the note for the courier.
- Removing an item is noted on the order and passed to an operator.
- Switching to pickup and cancelling go to a person, according to your rules.
- On other calls, the request reaches the operators as an alert, with the customer's words.
- Change
- Delivery time: 21:15
- Comment
- Ring the intercom twice
- Applied by
- Agent, during the call
- Notified
- Operators on shift
What's in it, what it costs, what's on the menu
The agent reads the menu synced from Syrve: names, categories, prices and descriptions. Allergens, portion weights and recommendations come from the knowledge base you approve.
- Products deleted from the Syrve item list are no longer offered by the agent.
- An allergy mentioned on the call is flagged as a priority for the team.
- Anything that isn't in the menu or the knowledge base goes to a person.
Website orders, confirmed by call
When a new website order appears in Syrve, the agent calls the customer, checks the items, the address and the time, and records the outcome. If nobody answers, it tries again; after that, the order gets a note that it must be called manually.
- It doesn't call outside the set hours, and it skips old orders.
- It skips orders marked "do not call".
- If the customer clearly declines, the agent can cancel the order in Syrve, with the reason noted.
- New order, 3 itemsConfirmedIn conversation
- New order, 1 itemNo answerIn conversation
- New order, 5 itemsConfirmedIn conversation
- New order, 2 itemsRetryIn conversation
Reservation requests, taken down in full
The agent asks for the day, time, number of guests, name and phone number, then sends the request to the team, to a spreadsheet or through a webhook. Confirming the table stays with you.
- No integration with any specific reservation system.
- Sent via webhook, to Google Sheets or as an alert for the team.
- For a reservation system with an API, the connection is built on request.
- Day and time
- Saturday, 19:30
- Guests
- 6, with a high chair
- Contact
- Caller's name and phone number
- Status
- To be confirmed by the team
Opening hours, delivery zones, payment
Until what time you're open, whether you deliver to a neighbourhood, what the minimum order is, how to pay. The agent answers from the knowledge base, in the customer's language.
- Answers are changed from the dashboard, no developer needed.
- It also works at night, when the restaurant is closed.
Restaurant types
Which kind of restaurant is it for?
The same tools, with different rules. Here is what the agent looks like in the most common setups.
Delivery restaurants
Order status by the customer's number, notes for the courier and confirmation of website orders.
Dine-in restaurants
Reservation requests taken down in full, opening hours, menu and allergens, with transfer to the floor team for groups and events.
Chains with several locations
The agent searches for the order across all restaurants in the Syrve account; each location's hours and zones live in the knowledge base.
Dark kitchens
No counter and no dining room: the agent holds the phone, the team holds the kitchen. Voice ordering into a cart can be built as a pilot.
Cafés and bistros
Opening hours, menu and quick questions, without pulling anyone away from the bar.
Peak hours and night-time
A schedule rule sends calls to the agent all day, only in the evening or only when the restaurant is closed.
Integrations
How does it connect to the restaurant's systems?
Syrve (iiko) works today. r_keeper and Oracle Micros are coming soon. For another system with a documented API, we build the connection through webhooks.
- Available
Syrve (iiko)
Searching for an order, the customer history and comments written on the order.
Details - Coming soon
r_keeper
Restaurant cash register: orders and menu.
- Coming soon
Oracle Micros
Restaurant and hotel cash register.
- Available
Webhook tools in the call
During the call, the agent calls your address and uses the response in the reply.
- Available
SMS via Infobip
Confirmations, reminders and messages after the call, sent by the agent or the campaign.
- Available
SIP / Asterisk telephony
Your numbers and lines connect via SIP, with inbound calls and data.
The path of a call
From the first ring to the note on the order, every step follows the rules you write.
The customer calls
On the restaurant's number, forwarded to Kallina all day or only when the operators are busy.
Notice and greeting
The customer hears that the call is recorded, then the agent asks how it can help.
Looks it up in Syrve
While talking to the customer, the agent finds the order by number or reads the menu.
Answers or hands over
It answers from the data, logs the request or transfers the call to an operator.
After the call
Transcript, summary and alerts for the team; on known orders, the note appears in Syrve.
What each person in the restaurant sees
The agent doesn't add a new screen for the kitchen. The work lands where the team already works.
- The operator
- Only the calls that need a person, plus alerts with customer requests: what they want, their exact words and the order.
- The kitchen
- Nothing new to learn: changes and notes appear on the order in Syrve, as before.
- The courier
- The corrected address and the delivery note, on the same order they already have.
- The manager
- In the Kallina dashboard, every call with recording, transcript, summary and reason, to see what drives the most calls.
Comparison
AI voice agent, voicemail or one more operator?
Differences in how they work, not measured performance. Most restaurants combine the agent with their existing team.
| Voicemail | One more operator | Kallina voice agent | |
|---|---|---|---|
| Peak hour | Records messages; someone calls back | One call at a time | Several calls at once, within the limits of your plan and phone line |
| Where the order is | Can't answer | Looks it up manually in the system | Looks it up in Syrve by the customer's number |
| Nights and days off | Yes, messages only | Only during a shift | Yes, according to the schedule rule |
| Languages | The recorded message | The languages they speak | More than 70, tested on your case |
| Complaints and compensation | No | Yes | Hands them to a person, with a summary of the request |
| What remains after the call | The message recording | What the operator writes down | Recording, transcript, summary, reason |
With your own numbers
How many orders can be lost on the phone?
We don't quote savings percentages. Enter your own numbers and see the formula; the result is yours, not a promise from us.
- Unanswered calls per month
- 180
- Orders that may be lost per month
- 54
- Their value per month
- 1,350
Formula
calls per day × % unanswered × % who would have ordered × days = lost orders; × average value = their value.
This is an estimate based on your numbers, not a promise from us. How many of these calls the agent takes, you only find out by testing it on your own calls.
Pilot
Voice ordering, built on your menu
For delivery, dark kitchens and delivery platforms, an agent can be built that listens to the order, adds items to the cart from the menu, reads back the total and places the order once the customer confirms. It is not a standard feature: we build it as a pilot, usually on the voice button on your website, and test it on real orders before launch.
From the menu, not from memory
Products and prices come from the synced catalogue, not from what the model remembers.
The cart, read back
The customer hears what they ordered and the total before confirming.
A person, one step away
Any doubt about zones, payment or products goes to an operator.
How we get started, step by step
We set the exact timeline after the demo, once we know what the agent must do and which integrations it has.
A demo on your own calls
You show us the real reasons customers call; you talk to an agent built on them.
The connection and the rules
We connect Syrve and write the house rules and the knowledge base: menu, hours, zones, allergens.
Tests on real conversations
Unclear questions, corrections, language switches, customers in a hurry, before the agent answers a real customer.
Launch and follow-up
We forward your number or switch on the website button, then read the first calls together and adjust.
What needs to be prepared
Access to Syrve (iiko)
The API key and the restaurants in the account that we connect.
House rules
What the agent may change and until when, who approves a cancellation, what to say about a delay.
Knowledge base
Opening hours, delivery zones and terms, allergens, answers to frequent questions.
Phone line
Forwarding of the ordering number or a new number, through SIP carriers.
Languages
In the customer's language, whatever it is
The agent can speak more than 70 languages, supported by the voice engine. We test any new language on your case before we promise it.
Customers from other countries
Tourists and customers who don't speak the local language get answers in their own, on the same number.
The language can change mid-call
The customer starts in one language and continues in another; the agent follows.
Starting language
Set per agent or, for confirmations, based on the order data.
- RomânăNative
- РусскийNative
From the voice engine catalogue
Each language is tested on your case before we promise it.
Data and security
What happens to call data?
Recording announcement
On calls through our telephony, the customer hears that the call is recorded, before recording starts.
Where the data is stored
On our servers and with the providers listed on the Security and data page, under the data processing agreement.
Export and deletion on request
A customer's data can be exported or deleted at their request, following the procedure on the Security and data page.
Pricing
What does it cost?
Kallina is paid for with monthly credit packages, used up according to call duration and the chosen model. Implementation, meaning the script, the Syrve connection and the tests, is paid once. Telephony is paid separately.
Demo on your phone
Schedule a demo
The fastest answer is obtained by speaking with a Kallina agent about your case.
Frequently asked questions
What is an AI voice agent for restaurants?
It is a virtual assistant that answers the restaurant's phone: it says where the order is, answers questions about the menu, opening hours and delivery zones, takes reservation requests and confirms website orders. With Kallina, the agent reads the data from Syrve (iiko) and hands over to a person anything that needs a decision.
Can the agent take a new order by phone?
It can be built as a pilot: the agent adds menu items to the cart, reads back the total and sends the order after confirmation. We recommend starting with order status and confirmations, then testing new orders on your menu and delivery zones.
Do you work with a system other than Syrve?
The ready-made integration for restaurants is Syrve (iiko). r_keeper and Oracle Micros are coming soon. For another system with a documented API, we build the connection through webhooks.
Does the agent know what has sold out today?
The agent doesn't know automatically what has sold out today. The menu is synced from Syrve with names, prices and descriptions, and products removed from the item list are no longer offered; today's stop list is not read yet. If you keep it in a system with an API, let's talk about connecting it.
Can it cancel an order?
Only according to your rules. On the confirmation call, if the customer clearly declines, the agent can cancel the order in Syrve with the reason noted. Otherwise, cancellations and switches to pickup go to an operator.
What does the agent do if the client is angry because of the delay?
It does not promise discounts it cannot give. It says what it sees in the order and, if the person asks, transfers them to an operator.
What happens with allergies?
The agent answers only from the menu and the knowledge base. When a customer mentions an allergy, the call is flagged as a priority for the team, and on known orders the note appears in Syrve. The decision stays with the kitchen.
Can you also make table reservations?
The agent takes down the reservation request and sends it to the team, to a spreadsheet or through a webhook. We have no integration with any specific reservation system; confirming the table stays with you.
Does it work for a chain with several restaurants?
Yes, the agent works for a chain too: we connect every restaurant in the Syrve account, and the agent searches for the order across all of them. Each location's hours and zones go into the knowledge base.
Does the client realise they are speaking to an AI agent?
Yes. Before recording, a short notice says that the call is recorded. We recommend that the agent introduces itself as a virtual assistant.
Can you name the restaurants that use Kallina?
No, not without their written consent. In the demo, we show you the workflows, not client names.
Show us an ordinary evening of calls
We start from the real reasons your customers call and show you the agent on your case.