Salesforce Acquires Cresta for $2.7 Billion to Put Voice AI on Customer Calls

Salesforce bets $2.7B on voice AI

Salesforce said on November 2, 2026 that it has signed a definitive agreement to acquire Cresta, a voice AI startup, in a deal reported at approximately $2.7 billion. The Salesforce Cresta acquisition is one of the largest AI acquisitions of the year, and it continues a buying spree that already included Fin, Braze, Listen Labs, and Contentful in 2026. Salesforce has not disclosed the price itself; the $2.7 billion figure comes from CMSWire and Silicon Republic reports, not from either company, so treat it as reported rather than confirmed.

Cresta builds what it calls a voice-first operating system for customer experience: AI agents that handle complex, high-volume customer service calls from start to finish, paired with real-time coaching software for the human agents who still take calls. The deal adds that capability directly into Salesforce's Service Cloud and Agentforce products, which is the point. Salesforce does not just want to sell software that helps human agents. It wants to sell the agent that takes the call.

The deal is expected to close in the first quarter of Salesforce's fiscal 2028, roughly February to April 2027, pending regulatory approval. That long window means Cresta keeps operating independently for months, and customers should not expect immediate integration. It also gives regulators and competitors time to react, which matters in a market where every large software company is buying AI capabilities at speed.

What the Salesforce Cresta acquisition buys

Voice is the hardest channel in customer service, and that is why this deal is interesting. Text chatbots can pause, can show options, and can hand off gracefully. Voice calls happen in real time, with accents, interruptions, background noise, and customers who are often already frustrated. An AI agent that can handle a complex service call end to end, without a human stepping in, is a much harder product than a chatbot, and it is also a much more valuable one, because voice calls are where support costs concentrate.

Cresta's coaching software is the other half of the value. Even in a future where AI agents take many calls, human agents will handle the hardest cases, and real-time coaching, suggesting what to say, flagging compliance risks, and summarizing after the call, makes those humans more effective. Salesforce is buying both halves: automation for the routine, and assistance for the exception. That combination is easier to sell to a cautious enterprise buyer than full automation alone.

The reported numbers give a sense of scale. Sacra estimates Cresta was generating roughly $100 million in annual recurring revenue as of April 2026, which would put a $2.7 billion price at a high multiple, though not an unusual one for a fast-growing AI company in 2026. Cresta's funding history includes a $50 million Series B in 2021, a Series C in 2022, and a $125 million Series D in November 2024 co-led by World Innovation Lab and Qatar Investment Authority, with Tiger Global, Qualcomm Ventures, and Workday Ventures among the backers.

The risks in a $2.7 billion voice bet

The first risk is integration. Salesforce's 2026 acquisitions, including the $5.66 billion Braze deal, create a large portfolio to stitch together, and customers have seen big software rollups produce confusing product lines before. Service Cloud buyers will want to know whether Cresta remains a distinct product, becomes a feature inside Agentforce, or gets split across both. The answer will shape pricing, migration, and support for years.

The second risk is trust. Voice AI agents that handle customer calls will record, transcribe, and analyze conversations at scale, which raises privacy and compliance questions that text chatbots mostly avoid. Enterprises in regulated industries will ask hard questions about data retention, consent, and audit trails before they let an AI agent take a call about a bank account or a health plan. Salesforce's enterprise credibility helps here, but it does not answer the questions by itself.

The third risk is the market itself. Voice AI for customer service is crowded, with startups and incumbents all claiming similar capabilities, and differentiation often comes down to deployment results rather than demo quality. A $2.7 billion acquisition only pays off if Cresta's agents handle real calls, with real customers, measurably better than the alternatives. That is a claim that takes quarters of customer data to prove, not a press release.

Even so, the direction is clear. Salesforce is betting that customer service is moving from software that assists humans toward agents that do the work, with humans coaching and handling exceptions. If that bet is right, owning a proven voice platform is worth a premium. If it is early, Salesforce will have paid a lot to learn that customers still want a human on the other end of a hard call. Either way, the next year of Service Cloud roadmaps will show how serious this bet really is. For the market context, see our coverage of Alphabet's cloud quarter and the Marvell Lumentum deal.