Skip to content
Salesforce Dictionary - Free Salesforce GlossarySalesforce Dictionary
DictionaryIInteractive Voice Response (IVR)
ServiceAdvanced

Interactive Voice Response (IVR)

An Interactive Voice Response (IVR) is the automated phone menu that greets inbound callers and routes them before a human agent picks up.

§ 01

Definition

An Interactive Voice Response (IVR) is the automated phone menu that greets inbound callers and routes them before a human agent picks up. The caller hears a recorded prompt, presses keypad digits or speaks a request, and either resolves the issue through self-service or lands in the right agent queue. In Salesforce, IVR is not a standalone product. It is delivered through Salesforce Voice (formerly Service Cloud Voice), where the telephony layer runs the menu and Salesforce supplies the data, screen pop, and Omni-Channel routing.

Salesforce offers three telephony models, and each handles the IVR differently. With Amazon Connect, the IVR lives in Amazon Connect Flows. With Partner Telephony, a vendor such as Genesys or NICE runs the IVR through a managed package. With Partner Telephony from Amazon Connect, you bring your own Amazon Connect instance. Older deployments built IVR screen pops through Open CTI, but that framework reaches end of life on February 28, 2028, and Salesforce points customers to Salesforce Voice instead.

§ 02

How IVR fits into the Salesforce Voice stack

The three telephony models and where the IVR lives

Salesforce Voice supports three telephony models, and the model you pick decides who owns the IVR. With Salesforce Voice with Amazon Connect, Salesforce provisions an AWS account and an Amazon Connect instance for you, and the IVR menu is built as an Amazon Connect Flow. With Salesforce Voice with Partner Telephony, a provider like Genesys, NICE, or Cisco runs the IVR inside their own platform and connects to Salesforce through a managed package they publish. With Partner Telephony from Amazon Connect, you keep an Amazon Connect instance under your own AWS account and wire it in. In every case Salesforce handles the agent side: the screen pop, the Voice Call record, real-time transcription, and Omni-Channel routing. The telephony partner handles the phone numbers and the menu logic. Choosing a model is usually about existing investment. A team starting fresh often takes Amazon Connect, while an enterprise with a multi-year Genesys contract keeps Partner Telephony and treats Salesforce as the agent workspace on top.

Building the IVR in Amazon Connect Flows

When you run the Amazon Connect model, the IVR is an Amazon Connect Flow. A flow is a drag-and-drop diagram of the call from the moment it connects to the moment an agent answers or the caller hangs up. You drop blocks onto a canvas and link them in sequence. Common blocks include Set logging behavior, Set voice (which picks the text-to-speech voice), Play prompt, Store customer input (which captures keypad or spoken entry), and Invoke AWS Lambda function for a data lookup. Salesforce Voice expects specific flow types for inbound calls, outbound calls, and transfers, and the setup process installs starter versions you then customize. After you build a flow you publish it and assign it to a phone number so live calls use it. Because the flow is the IVR, every prompt wording change, every new menu branch, and every routing rule is an edit to the flow rather than a Salesforce config change. Admins who own Voice need comfort in both the Amazon Connect console and Salesforce Setup.

Data dips: pulling Salesforce context into the menu

The feature that turns a generic IVR into a personal one is the data dip. A data dip is a mid-call lookup where the flow queries an external system, gets an answer, and branches on it. In the Amazon Connect model this runs through an AWS Lambda function invoked from the contact flow. The Lambda authenticates to Salesforce, queries by the caller phone number or a digit-entered account number, and returns values like the caller name, open case count, or account tier. The flow then uses those values to personalize. A returning caller can hear their name and an offer to get an update on yesterday's case instead of starting cold. Salesforce publishes a Trailhead project, Build a Data Dip with AWS Lambda and Amazon Connect, that walks through dropping Store customer input and Invoke AWS Lambda function blocks, publishing the flow, and testing it against a phone number. One practical caution from that project: the lookup runs against a time limit (the sample uses an eight-second window), so a slow callout produces an error path the flow has to handle.

Handing the call to an agent through Omni-Channel

The IVR rarely resolves every call, so the last job of the menu is the handoff. When the flow decides a human is needed, it transfers the call into Salesforce Omni-Channel, the same routing engine that distributes cases and chats. The voice work item lands in a queue, Omni-Channel matches it to an available agent with the right skills, and the call rings through. Whatever the IVR collected travels with it. The agent sees a screen pop with the matched record and the reason the caller gave, so the conversation skips the May I have your account number opener. This is the payoff of doing the data dip earlier: the context is already attached. Routing can be skills-based, so a billing question reaches a billing-trained rep, or it can go direct to a named rep for relationship accounts. Because Omni-Channel is shared across channels, a supervisor sees voice, chat, and case load in one place rather than in a separate telephony console, which is one of the main reasons Salesforce steers teams off Open CTI.

Touch-tone menus versus conversational IVR with Agentforce

Classic IVR is touch-tone. The caller hears Press 1 for sales, 2 for service and walks a fixed tree, sometimes with narrow speech recognition bolted on. Conversational IVR replaces the tree with natural language. The caller states a request in plain words, the system reads the intent, and it routes or resolves without a numbered menu. In current Salesforce deployments this conversational layer is increasingly an Agentforce or Einstein Bot agent extended into voice through Salesforce Voice. The bot greets the caller, handles common self-service like an order status check or a password reset, and escalates to a person only when the request gets complex. The design goal is deflection: clear the routine share of calls without an agent and route the harder calls to a skilled rep with full context already gathered. Conversational IVR also sidesteps the classic menu complaint, the deep nested tree, because there is no tree to get lost in. Most teams still keep a touch-tone fallback for callers who prefer it or for noisy lines where speech recognition struggles.

Open CTI, retirement, and what replaces the old screen pop

Before Salesforce Voice, many orgs integrated an outside IVR (Cisco, Avaya, Genesys, or a custom build) using Open CTI. The external system ran the whole call flow, and Open CTI handled the browser side inside Salesforce: the screen pop, click-to-dial, and basic call controls. The IVR kept the menu and only told Salesforce when the call reached an agent. Salesforce has set Open CTI to reach end of life on February 28, 2028, and no longer lets new Agentforce Service orgs implement it. The recommended path is Salesforce Voice, which Salesforce says adds native Omni-Channel integration, real-time call transcription, Next Best Action, and Agentforce capabilities that Open CTI never offered. For a team on Open CTI today the migration is not a same-day switch. It means choosing a telephony model, rebuilding menu logic as flows or in a partner package, and re-pointing screen pops at the Voice toolkit, which is why the multi-year runway matters for planning.

§ 03

Stand up an IVR with Salesforce Voice and Amazon Connect

There is no single IVR object you create in Salesforce. You stand up IVR by configuring a Salesforce Voice contact center and then building the menu in the telephony layer. These steps describe the Amazon Connect model, the most common starting point.

  1. Turn on Salesforce Voice with Amazon Connect

    In Setup, enable Voice and complete the Amazon Connect model setup. Salesforce provisions an AWS account and an Amazon Connect instance tied to your contact center, including the tax and AWS root contact details the wizard asks for.

  2. Create the contact center and assign access

    Create a contact center record that links Salesforce to your Amazon Connect instance, then grant agents and supervisors access through the Voice permission sets so the softphone shows in the utility bar.

  3. Build the IVR as an Amazon Connect Flow

    In the Amazon Connect console, edit the inbound flow. Add Play prompt and Store customer input blocks for the menu, and add Invoke AWS Lambda function blocks where you want a Salesforce data dip to personalize or pre-route the call.

  4. Route the handoff into Omni-Channel

    Configure the flow to transfer to Salesforce when a caller needs an agent so the work item flows through Omni-Channel with the collected context, and the agent gets a screen pop instead of asking for details again.

  5. Publish, assign a number, and tune

    Publish the flow, assign it to a phone number, and place test calls. Review IVR analytics for abandonment points and top intents, then reword prompts or reorder branches based on what callers actually do.

Telephony modelremember

Amazon Connect, Partner Telephony, or Partner Telephony from Amazon Connect. The choice sets whether you build the IVR in Amazon Connect Flows or in a partner managed package.

Data dip via Lambdaremember

An optional AWS Lambda callout from the flow that queries Salesforce by phone number or account number and returns context the menu branches on.

Conversational layerremember

An optional Agentforce or Einstein Bot agent extended into voice to replace numbered menus with natural-language intent capture and self-service.

Routing typeremember

Skills-based routing through Omni-Channel for queue distribution, or direct-to-rep routing for named-account relationships.

Gotchas
  • Keep menus shallow. Every extra level drops completion, and three levels is the practical maximum for touch-tone trees.
  • Always offer an escape to a human at each level. Trapping callers in self-service when they want a person hurts satisfaction.
  • Data dips run against a time limit. The Trailhead sample uses an eight-second window, so build an error path for slow or failed callouts.
  • Open CTI reaches end of life on February 28, 2028. If your IVR screen pop still rides on Open CTI, plan the move to Salesforce Voice well ahead.

Prefer this walkthrough as its own page? How to Interactive Voice Response (IVR) in Salesforce, step by step

§

Trust & references

Official documentation

Straight from the source - Salesforce's reference material on Interactive Voice Response (IVR).

Was this entry helpful?
Help us write better definitions. Quick reactions or detailed edit suggestions.

About the Author

Dipojjal Chakrabarti is a B2C Solution Architect with 29 Salesforce certifications and over 13 years in the Salesforce ecosystem. He runs salesforcedictionary.com to help admins, developers, architects, and cert/interview candidates sharpen their fundamentals. More about Dipojjal.

§

Test your knowledge

Q1. What is an Interactive Voice Response system in Salesforce Service Cloud Voice telephony?

Q2. What is a healthy design principle for an Interactive Voice Response menu tree in production?

Q3. What does a modern Interactive Voice Response stack support beyond DTMF keypad input today?

§

Discussion

Loading…

Loading discussion…