Salesforce’s journey to AI-native voice

For years, the voice channel has been the slightly awkward cousin at the customer experience family reunion.

Customer data lives in the CRM. Call routing lives in the contact center platform. Recordings, analytics, and knowledge tools scatter across different systems. Meanwhile, agents toggle between tabs while customers wait on hold.

Salesforce’s journey in voice has been about fixing that fragmentation. The goal isn't to reinvent the phone, but to bring conversations, context, and action into a single workspace.

Agentforce Contact Center represents the current milestone in that journey.

The evolution of Salesforce Voice integration: Connecting the phone to the CRM

Open CTI (computer telephony integration) was the critical first step. It gave organizations a practical way to connect telephony systems to Salesforce and enabled features like screen pops, click-to-dial, and embedded softphones.

At the time, this was a major step forward. Instead of treating the CRM and phone system as two entirely separate planets, Open CTI built a wormhole between them, with straightforward ROI.

An incoming call could trigger a screen pop. An agent could click a number in Salesforce to dial. A call could be associated with a case, contact, or account. Suddenly, the agent did not have to start every conversation by asking, “Can I have your name, date of birth, account number, and name of first pet?”

But Open CTI was still fundamentally a bridge. The call itself lived somewhere else.

With Open CTI now in maintenance mode and heading to retirement in 2028, companies face a hard deadline to update their infrastructure.

Open CTI is ultimately a historical marker. The more interesting story is what Salesforce learned from it, and how those lessons informed its next strategic moves.

Why voice must be treated as data

A phone call is not just an event.

It is not merely a start time, an end time, a queue, a disposition code, and a recording file with a mysterious name like CALL_2026_06_26_FINAL_FINAL_v3.wav.

A phone call is more than just an event recorded as a start time, an end time, and a raw audio file hidden in a database. Every customer call contains intent, emotion, history, commitments, exceptions, and frustrations. It contains the raw material for the next customer interaction and the exact context a future agent needs to deliver great service. And increasingly, it contains the information an AI agent needs to act intelligently.

That is why voice must be treated as data.

Salesforce Voice, originally launched as Service Cloud Voice, represented a meaningful shift in ambition. The objective was no longer just to place call controls next to the customer record, but to make the voice interaction itself a native part of the Salesforce ecosystem. In technical terms, Salesforce reduced the infrastructure mismatch between the contact center and the CRM. In simpler terms, the phone call finally unpacked its bags and moved into the house.

The real breakthrough was conversation intelligence

While a unified agent workspace dramatically improves training time and handling accuracy, the real breakthrough was conversation intelligence. Once spoken conversations become transcript data connected directly to the CRM, service capabilities multiply. Organizations can:

• Recommend relevant knowledge in real time.
• Suggest next-best actions.
• Create summaries and follow-up tasks.
• Improve case documentation.
• Identify coaching opportunities.
• Analyze customer sentiment and recurring issues.
• Support quality and compliance reviews.
• Preserve context across customer interactions.
• Give AI a reliable view of what actually happened.

This is the point where voice stops being a black box.

For decades, organizations have captured enormous volumes of call recordings. But recordings alone are more like a warehouse full of unlabeled boxes. Transcripts, structured interaction records, and connected CRM data turn those boxes into a searchable library, effectively transforming your historical data into an operating system for action.

The great CX convergence

For a long time, the customer experience industry treated CCaaS, CRM, quality management, workforce management, knowledge, analytics, and automation as separate categories. Each had its own buying cycle, data model, implementation team, and alphabet soup of acronyms.

But customers experience those systems as a single journey.

When something breaks, it is rarely clear or relevant to them why. A transfer, a repeated question, or missing context is enough to create frustration, regardless of which system is responsible.

This is why these categories are converging. While specialized disciplines remain deeply important, the rigid walls between them are organizational leftovers rather than customer-centered design. Voice sits right in the middle of this convergence, making it the most sensitive place for friction to appear if left unintegrated.

Why “adding AI” is not a CX strategy

There is a temptation in every technology cycle to take the existing workflow, attach an AI label, and call it transformation. That approach is the business equivalent of strapping a jetpack onto a filing cabinet. It may move faster, but it is still a filing cabinet.

The opportunity with AI-driven voice is not simply adding transcription, agent-assist, or a more conversational IVR to the old process. The true value lies in starting with the desired customer outcome and redesigning the entire journey around it.

Consider a customer calling about a delayed shipment. The old model would route the call, authenticate the caller, and give the agent a script to read. A redesigned, AI-native model recognizes the caller instantly, understands their order status, proactively explains the delay, and offers approved compensation options in the same conversation. A human agent only steps in when complex exceptions or human empathy are required, and they do so without making the customer repeat themselves.

That is not just adding AI to voice; that is voice redesigned around resolution.

The role of partner telephony is shifting, not disappearing

None of this means telephony providers are going away. They continue to offer deep expertise in areas like global carrier infrastructure, compliance, and operational resilience. Many organizations will continue to rely on those capabilities for the foreseeable future.

What is changing is where interaction intelligence lives.

As more of the interaction moves closer to the CRM and AI layers, the value shifts from pure connectivity to how effectively systems work together to deliver a cohesive experience.

Salesforce Voice created a way to bring those elements closer together while preserving meaningful provider choice.

That is a useful distinction because the future is not necessarily native voice versus partner telephony but rather about where the center of gravity for customer context, automation, and AI should live.

Agentforce Voice and the next step

Until recently, voice AI mostly worked behind the scenes by summarizing calls, digging up data, and suggesting next steps for human agents. While that remains incredibly valuable, AI is now joining the conversation directly.

An AI voice agent can understand spoken requests, answer questions, and execute tasks securely. But this isn't about replacing humans with robots; it’s about using AI to boost speed and efficiency while saving human judgment for where it matters most.

Now, AI is actually joining the conversation. An AI voice agent can understand spoken requests, answer questions, and execute tasks securely. But this isn't about replacing humans with robots; it’s about using AI to boost speed and efficiency while saving human judgment for where it matters most.

That is the promise of a connected platform.

Agentforce Contact Center: The market test begins

By adding native voice to its broader service story, Salesforce isn't trying to match specialized CCaaS providers feature-for-feature overnight. Instead, it is removing the traditional boundaries between voice conversations and CRM data, bringing routing, automation, and AI agents into a single, cohesive loop.

The question now is what Salesforce wants to become in voice.

This is the first year of Agentforce Contact Center, so it’s too early to declare victory or failure. The real market test is whether customers reward Salesforce for a more integrated, AI-native customer experience.

My expectation is that Salesforce will not try to win by becoming the best traditional telephony company in every category. Instead, it will likely focus on becoming the best platform for building AI-driven service experiences that happen to use voice.

That is a different kind of ambition.

If businesses adopt Agentforce Contact Center for its speed, simplicity, and tight integration, it becomes a highly capable option for any organization looking to reduce a fragmented CX stack.

If customers trust it with complex, high-value voice journeys, it will give Salesforce a strong incentive to double down on enterprise voice. Either way, the center of gravity in customer service is shifting.

The partner opportunity is moving up the stack

For years, the value in the Salesforce voice ecosystem was focused entirely on the plumbing. Partners made their mark by connecting telephony platforms, embedding softphones, syncing data, and simply keeping the wires from catching fire. While that infrastructure work still matters in complex environments, the next wave of value is much more strategic.

The real opportunity now lies in designing and operating Interactive Voice Assistants (IVAs). Building an AI agent that understands intent, accesses trusted data, takes secure action, and hands off to humans seamlessly is no longer just a telephony integration. Instead, it requires customer experience design, AI orchestration, data governance, and operational change management.

In other words, the fun stuff starts after the phone rings.

However, launching an AI voice assistant is only the first lap, not the finish line. AI-driven voice experiences need continuous care because knowledge changes, policies evolve, and new edge cases multiply constantly. This is why managed services are essential in an AI future. Someone needs to monitor outcomes, tune prompts, validate quality, and maintain integrations over time.

The best organizations will not deploy an IVA and admire it from a distance. They will treat it as a living service capability that is measured, governed, and expanded based on evidence. The practical path is straightforward: start with a meaningful customer outcome, pick a well-bounded journey, launch quickly, and learn from real interactions. By moving fast and demonstrating outcomes before expanding, AI voice becomes far more than just a demo.

A new era of voice innovation

Salesforce’s voice journey has followed a clear direction. First, connect the call to the CRM. Then, bring the agent experience into one workspace. Next, capture the conversation as usable data, and use that data to guide people and automate workflows. Now, the platform enables AI agents and human agents to operate from the exact same customer context.

The opportunity ahead is not just about making contact centers more efficient. It is about designing customer experiences that were previously impractical or impossible. Voice is no longer just an isolated channel for handling calls; it is becoming a platform for resolving, anticipating, and strengthening customer relationships, one conversation at a time.