Salesforce is stepping deeper into the generative AI race with the introduction of Agentforce Voice, a new feature that lets its AI agents speak to customers in natural-sounding conversations.

The move expands the company’s AI-powered customer service platform beyond text-based chat, allowing businesses to customize tone, pacing, and pronunciation — and even let customers interrupt the agent mid-conversation, much like in real human interactions.

Announced ahead of Salesforce’s annual Dreamforce conference in San Francisco, Agentforce Voice marks the company’s most ambitious step yet to modernize enterprise customer service with AI. The feature integrates with major telephony systems, including Amazon, Five9, Genesys, NICE, and Ericsson’s Vonage — platforms already widely used across corporate call centers.

With this, Salesforce is aiming to position itself at the center of an emerging era of AI-enabled voice interaction that promises to redefine how businesses handle customer engagement.

The development signals how rapidly voice technology has become the next competitive frontier in artificial intelligence. What began as a text-driven revolution with OpenAI’s ChatGPT in late 2022 has evolved into a more immersive experience where machines can hold conversations that sound increasingly human. Both OpenAI and Anthropic have launched spoken versions of their chatbots that can converse fluidly without the flat, robotic tone typical of earlier voice assistants. Now, that same capability is finding its way into the backbone of corporate software — the systems that handle everything from customer inquiries to tech support.

Salesforce launched Agentforce in 2024 as part of its sweeping push to embed generative AI into its suite of customer relationship management (CRM) tools. The platform uses large language models to help companies automate and personalize responses to client requests over chat, email, and now, phone. According to Salesforce, Agentforce has been deployed in more than 12,000 business environments — a notable milestone that demonstrates its appeal, even if adoption has not matched early investor expectations.

Analysts at RBC Capital Markets noted last week that “investor enthusiasm around Agentforce has moderated as adoption has lagged expectations,” maintaining a “hold” rating on Salesforce’s stock. The cautious sentiment reflects broader market unease about how enterprise software firms will adapt as AI becomes more capable — and more autonomous.

Salesforce’s share price has slid about 28% since the start of 2025, a sharp contrast to the Nasdaq Composite, which has gained around 15% over the same period. The decline marks growing investor anxiety that AI could erode the long-term value of traditional software. In its latest earnings report, Salesforce itself warned that “new AI products may disrupt workforce needs and negatively impact demand for our offerings,” an admission that even industry leaders are not immune to the transformative, and potentially destabilizing, power of AI.

Those fears were amplified in September when Anthropic demonstrated that its Claude Sonnet 4.5 model was able to build a working chat application — similar to Salesforce’s Slack — in just 30 hours. The demonstration showcased how AI could potentially automate complex coding tasks and build full-scale software systems at a fraction of the time and cost, raising questions about the future relevance of traditional enterprise software development.

Yet Salesforce CEO Marc Benioff has sought to project calm. In a recent interview with CNBC’s Morgan Brennan, he rejected the notion that AI poses an existential threat to the business software industry.

“When we get into this kind of zero-sum game — that all this is going to get wiped out or change overnight — you’re not dealing with somebody who actually runs a company,” Benioff said. “Business is incremental, it’s evolutionary. We don’t see that kind of change.”

Benioff’s comments pinpoint a more measured view of how technology disruption unfolds — not as an immediate revolution, but as a process of adaptation. Salesforce has weathered multiple technological transitions over its two-decade history, from the rise of cloud computing to the era of mobile-first enterprise software. For the company, AI represents both a challenge and an opportunity: a chance to reinvent itself while maintaining the scale and reliability that large businesses require.

The new Agentforce Voice offering could become a key part of that reinvention. Salesforce is entering a race to make AI interactions more natural and contextually aware — a vital advantage for customer service operations that depend on human-like empathy and responsiveness. Voice, after all, remains the dominant mode of communication for many customers who prefer calling over typing.

The company says that with Agentforce Voice, users will be able to set different personalities for their AI agents — adjusting tone, accent, and delivery speed to align with brand identity or cultural context. The ability for callers to interrupt the AI mid-sentence adds a touch of realism, acknowledging one of the most common elements of human conversation: interjection.

Salesforce also plans to roll out Agent Script in November, a companion product that allows businesses to customize what AI agents say and do. Organizations will be able to fine-tune dialogues, craft guided responses, and set behavioral parameters for AI-driven customer service — effectively building a digital workforce tailored to their communication style.

However, a growing roster of startups is racing to build similar products, led by Sierra, a company co-founded in 2023 by former Salesforce co-CEO Bret Taylor. Sierra, now valued at around $10 billion, has been attracting attention with its AI agents that can “pick up the phone” to handle customer calls. Its clients include major names like ADT, SiriusXM, and SoFi, positioning it as a credible challenger to its former parent company.

For Salesforce, that competition carries symbolic weight: a sign that even its former executives are betting that AI-driven automation could define the next era of enterprise software. Sierra’s rapid ascent shows how new entrants are capitalizing on the market’s hunger for practical AI applications — tools that not only demonstrate novelty but also save costs and streamline operations.

Still, Salesforce’s advantage lies in its scale and integration. Few companies can match its reach across global enterprises, where its CRM products are already deeply embedded in daily workflows. Salesforce is betting that its existing relationships, combined with the trust it has built over decades, will help it retain an edge, even as nimbler competitors emerge.

The introduction of Agentforce Voice comes at a moment when the AI landscape is undergoing a shift from experimentation to execution. Businesses are no longer merely testing generative AI; they are deploying it at scale. Yet, as enthusiasm rises, skepticism lingers about whether AI systems can maintain reliability, handle complex customer emotions, and comply with privacy laws in live interactions.

Salesforce’s decision to spotlight Agentforce Voice at Dreamforce is believed to be a strategy to reassure investors that the company remains not only relevant but essential in the new AI-driven economy. Its challenge now will be to prove that these AI innovations can generate tangible revenue growth — and not merely serve as buzzwords to placate a restless market.