A cloud IVR is an interactive voice response system that runs on a provider's servers instead of your own. You build the menu in a browser, calls reach it over the provider's phone network, and you pay by the minute or by the agent seat instead of buying servers, port licenses and a maintenance contract. Callers can hear exactly the same menu. What changes is who runs the system and how you pay for it.
The comparison table comes first. After it, you'll find what three kinds of cloud IVR charge, read from each vendor's pricing page on November 18, 2026, and an inventory worksheet for moving the menu you have now without losing a branch of it along the way.
Cloud IVR vs on-premises IVR: what is different?
The call flow can be identical. The differences are in who owns the hardware and how capacity is bought and paid for. An on-premises IVR runs on servers you buy, is usually licensed by how many calls it can hold at once (often called ports), and is patched by your team or your reseller. A cloud IVR is rented, and the provider runs it.
| On-premises IVR | Cloud IVR | |
|---|---|---|
| Where it runs | Servers in your data center or a rack you rent | The provider's data centers, managed in a browser |
| How you pay | Servers and licenses up front, then a yearly maintenance contract | Monthly, per minute of call time or per agent seat |
| Capacity | The ports you licensed; calls past that number get a busy signal or overflow | Grows with your call volume, up to the account limits the provider sets |
| Upgrades and patches | Scheduled by your team or your reseller | Shipped by the provider to every customer |
| Changing a prompt or an option | Your team or integrator edits the flow on your servers | Edited in a flow designer and published, with no new hardware |
| Lookups in your systems | Direct connections inside your own network | Requests to your systems over the internet, through an API or a private link |
| What an outage depends on | Your power, network and servers | The provider's region, plus the network between callers, the provider and your agents |
| Where recordings and caller input are kept | Your storage | The provider's region, so check which region and which compliance reports cover it |
A spike in calls exposes the capacity row first. NIST's definition of cloud computing describes resources that "can be rapidly provisioned and released with minimal management effort or service provider interaction" (NIST SP 800-145). An IVR sized in ports can't do that. When a storm or a billing error sends twice the usual calls, the extra callers hear a busy signal before the menu ever answers.
The lookup row is often the most work in a migration. An IVR that has spent years reading account balances straight out of a database in the next rack needs every one of those lookups rebuilt as a request your systems accept from outside the building.
How does a cloud IVR work?
Your phone numbers point at the provider, either ported to it or forwarded by your current carrier. When a call arrives, the provider runs the flow you built in its designer: it plays prompts (recordings you upload or text-to-speech), collects keypresses or speech, looks up the caller in your systems through an API, and sends the call to a queue or a phone number.
Where that call goes next matters for both cost and design. If your agents work on the same cloud platform, the IVR hands the call to a queue there, and the menu choice and anything it looked up show on the agent's screen. If your agents are still on the old system, the cloud IVR transfers the call to a number on that system. On usage-priced platforms, that second leg is usually billed for as long as the call lasts, which the cost section works through below.
Prompts move without much trouble. Our IVR script examples cover the full prompt set (greeting, menu, errors, queue, after hours), with word counts and play times, which is a good checklist when you re-record or switch to text-to-speech on the new platform.
Can a cloud IVR understand speech?
Yes, and some platforms price it into the plan. Genesys Cloud CX 1 lists a speech-enabled IVR among its features. Amazon Connect Customer includes self-service with Amazon Lex in its voice rate, while its Customer Basic plan charges $0.008 a minute for self-service, plus Lex.
A menu that understands "I need to change my delivery" in place of "press 3" is usually called conversational IVR, and adding it doesn't depend on where the IVR runs. Our guide to what conversational IVR changes and what it leaves alone explains how it routes a spoken request and what the speech services charge, and includes a worksheet for deciding which menu options should be finished on the call instead of routed.
What does a cloud IVR cost?
It depends on how you buy it. Contact center platforms sell the IVR inside a plan priced per agent seat or per minute, and programmable voice platforms sell the building blocks for you to put together yourself. These are list prices from each vendor's own pricing page, read on November 18, 2026.
| How you buy it | What comes with it | Published price |
|---|---|---|
| A contact center platform priced per seat | IVR, queues, the agent desktop and reporting in one plan | Genesys Cloud CX 1: $75 per user a month, billed annually, speech-enabled IVR included. Carrier service is separate: bring your own carrier or buy Genesys Cloud Voice |
| A contact center platform priced by usage | The same pieces, billed by the minute with no seat licenses | Amazon Connect Customer: $0.038 per voice minute, self-service with Amazon Lex included. Customer Basic: $0.018 a minute. Standard telephony rates apply on top of both |
| A programmable voice platform | A flow builder and phone numbers; you build and host the rest | Twilio Studio: 1,000 flow executions a month free, then $0.0025 per execution. Receiving a call on a US local number costs $0.0085 a minute, plus $1.15 a month for the number |
Sources: Genesys Cloud pricing, Amazon Connect Customer pricing and its pricing appendix, Twilio Serverless pricing and Twilio Voice pricing.
Seat price or usage price?
Per seat and per minute meet at one number: the minutes each seat carries. At $75 a seat, the money for one Genesys Cloud CX 1 seat buys about 1,974 minutes on Amazon Connect Customer at $0.038 a minute. Amazon bills the customer's whole call, menu and queue time included, so divide your monthly call minutes by your number of agent seats to find your own figure. Above roughly 1,974 minutes a seat (about 33 hours), the seat price costs less at these two list prices. Below it, the usage price does.
That only settles the price question. The two plans don't include identical features, telephony is extra on both, and the Genesys price assumes an annual commitment. Our AI call center software comparison lays out more vendors' published prices side by side.
The cost of keeping agents on the old system
One way to start is to put a cloud IVR in front of the old phone system and leave the agents where they are. On Twilio's US rates, a call that comes in on a local number and is then dialed out to your existing queue is billed on both legs while it stays connected: $0.0085 a minute to receive plus $0.014 a minute to call out, or $0.0225 for every minute after the transfer.
Take 20,000 calls a month, each spending one minute in the menu and six minutes with an agent after the transfer. The inbound leg comes to $1,190, the outbound leg to $1,680, and the flow executions past the free 1,000 to $47.50. The second leg is more than half of the roughly $2,920 total, which includes $1.15 for one local number. The call count and times are made up for the example, so put your own in. Whatever your numbers, that second leg is a reason to plan the agent move alongside the menu move. What the menu minutes themselves cost on Amazon Connect is worked through in our call center IVR guide, along with the six numbers that show whether a menu is working.
How do you move an IVR to the cloud?
Inventory every path the old IVR can take, rebuild and test each one on the new platform, then move one phone number at a time while the old system can still take calls. Most of the work sits in the inventory: holiday branches, closure messages and lookups that nobody hears on an ordinary day are the ones that get missed.
- Fill in the worksheet below for every phone number that reaches the IVR.
- Pull a month of menu reports so you know where calls go and where they end. You'll compare against these after the move.
- Rebuild the flow and the prompts in the new designer.
- Rebuild each lookup against an API your systems expose, and test it with test records.
- Test every branch: each key and each spoken option, every hours and holiday condition, no input and wrong input, and each transfer destination.
- Send one low-volume number first by forwarding it from your current carrier to the new IVR's number. Rolling back is then a change to the forward.
- Port or forward the remaining numbers once the first one holds up, and keep the old system able to take calls until the last one has moved.
IVR migration inventory worksheet
| What to collect | Where it is on the old system | What to check on the new one |
|---|---|---|
| Every number that reaches the IVR, and the flow each one starts | Carrier records and the IVR's number-to-application mapping | Each number lands on the right flow after it is forwarded or ported |
| The menu tree, option by option | The flow export or the original build documents | Every option reaches the same destination as before |
| Prompt recordings and their wording | The prompt or media folder on the IVR server | Each prompt plays at the right volume, in every language you offer |
| Business hours, holidays and emergency closures | Schedule tables in the IVR or the call distributor | Each closure message plays on the right dates, in your time zone |
| Lookups such as account, order, appointment or balance | Database queries or the integration settings | The same test record returns the same answer, and the flow still works when your system is down |
| Data passed to the agent with the call | Call-attached data settings in the IVR and the agent desktop | The agent's screen shows the same fields |
| Payment or identity check steps | The secure part of the flow and its compliance scope | The new provider's compliance reports cover that step (for card payments, its PCI DSS Attestation of Compliance) |
| Callback and queue position offers | Queue settings in the call distributor | Offers trigger at the same thresholds |
| The reports you track | Report definitions | Menu completion, menu abandonment and transfers are counted the same way, so before and after compare |
Don't skip the last row. If the new platform counts a call that leaves the menu for a queue differently than the old one did, your numbers will change on cutover day without anything happening to callers. Settle the definitions before the first number moves. Our call routing guide helps with the routing rules themselves, including what splitting one queue into several costs in agents.
How Telvana fits into either kind of IVR
The AI step doesn't have to wait for the migration. Telvana adds an AI agent as one more option in the menu, "Press 1 to talk to our AI assistant now," set up as an external transfer to the agent's own number. That works the same on an on-premises IVR and on any cloud IVR that can send a call to an outside number, so the option moves with the menu when you migrate (connecting your phone system).
Callers who take the option are answered right away, and callers who'd rather wait for a person keep their place in line (hold queue relief). The agent answers from the knowledge base you give it, looks up customers, checks availability and books, reschedules or cancels appointments in the tools you connect. When a call needs a person, it transfers to the number you choose under rules you set, and every call is recorded, transcribed, summarized and tagged by outcome. Our team builds the agent from your scripts and transfer rules and tests it on your call types, and you approve it before it takes a live call. Pricing starts at $399 a month based on call volume, plus a one-time setup fee (pricing).
Frequently asked questions
Is a cloud IVR the same as a hosted IVR?
Usually, yes: both mean the provider runs the IVR and you don't own servers. Some vendors use "hosted" for a dedicated system in their data center, sized for you. NIST counts resource pooling across many customers and rapid elasticity among the essential characteristics of cloud computing, and a dedicated hosted system may have neither, so ask how its capacity grows.
What is a conversational cloud IVR?
It is a cloud IVR that lets callers say why they are calling in their own words instead of pressing a key, then routes the call from what they said. The speech recognition usually comes from the platform or from a cloud speech service it connects to. Our conversational IVR guide covers how it works and what it costs.
Do we have to port our phone numbers to use a cloud IVR?
No. Your current carrier can forward each number to the number the cloud provider gives you, which lets you test with real calls and roll back by changing the forward. Port the numbers once the new flow has held up, or keep forwarding if your carrier contract makes that the better deal.
Is a cloud IVR secure enough to take card payments?
It can be, if the provider's compliance reports cover the payment step. Ask for its PCI DSS Attestation of Compliance and check which services it names. Keep card numbers out of recordings and transcripts, either with keypad entry that the platform masks or by pausing recording during that step.
How much downtime does a 99.95 percent uptime SLA allow?
About 22 minutes in a 30-day month, since 0.05 percent of 43,200 minutes is 21.6. Twilio, for example, advertises a 99.95 percent uptime SLA on its pricing page. Ask each provider how its SLA defines an outage and what it credits, and decide where calls go if the IVR is unreachable.
Can we keep our on-premises IVR and still add AI?
Yes. An AI agent can sit behind one option on the menu you already have, as an external transfer to the agent's number, so you can add it before, during or after a move to the cloud.
Map your menu before you move it
Book a demo and bring the menu you run today. We'll show you which of its options an AI agent could finish on the call, and you can hear a Telvana agent take one.