Salesforce Buys Cresta for a Reported $2.7 Billion to Fix Voice Agents, but the Price Is Still a Secret
Salesforce signed a deal to buy Cresta and put its voice agents inside Service Cloud. Salesforce did not disclose the price. The Information reports $2.7 billion, and the deal is not expected to close until early 2027.
Salesforce said on 2 November 2026 that it signed a definitive agreement to acquire Cresta, the voice-first customer experience platform, and fold its voice agents into Service Cloud and Agentforce.
Salesforce did not disclose the price, the consideration, or Cresta’s revenue. The Information reported the price at $2.7 billion, citing a person familiar with the deal. Salesforce’s press release confirms the agreement and the target, not the number.
The deal is expected to close in the first quarter of Salesforce’s fiscal 2028, which runs from February to April 2027, subject to customary closing conditions and regulatory approvals. Cresta will keep supporting the contact center and telephony platforms its customers use today.
Quick answer: Salesforce is buying Cresta, but the only price is a $2.7 billion report from The Information, not a company confirmation. The logic is clear: voice is the messy part of customer service AI, and Cresta adds end-to-end voice agents, real-time coaching, and conversation analytics to Service Cloud. The catch is that the deal will not close until early 2027, Cresta’s revenue and retention are undisclosed, and most operational numbers here are vendor case studies, not audits.
We’ve signed a definitive agreement to acquire @Cresta, bringing advanced voice agents and conversation intelligence together with Service Cloud.
Connecting every customer conversation to what matters: resolution, retention and revenue.
More: https://t.co/qworzdrOpl
— Salesforce (@salesforce) November 2, 2026
What Salesforce is actually buying
Cresta sells a voice-first operating system for customer experience. Its agents take complex calls from start to finish, its software coaches human representatives while they are on the phone, and its analytics platform scores every interaction.
Cresta says it was founded at Stanford’s AI Lab and has more than 300 employees. Ping Wu, a co-founder of Google’s Contact Center AI, is CEO.
The customer list in Salesforce’s release is the useful part: Marriott, Comcast, and Alaska Airlines. Those are large support operations with high call volumes, not startups running a chatbot on a marketing site.
Cresta’s case studies claim Brinks Home cut transfers by 73% and Cox Communications grew revenue per chat by 20%. Those figures are vendor-selected and not independently audited, so treat them as directional.
Why voice is the hard part
Text agents can pause, retry, and hide mistakes behind a typing indicator. Voice cannot. A caller notices a two-second delay, talks over the bot, switches languages, or gives an account number in a noisy room.
That is why contact centers have been slower to automate than chat. The work is high stakes, the integration runs through old telephony systems, and a failed call costs more than a failed chat. Salesforce’s pitch is that Cresta connects the conversation to the customer record and the workflow that fixes the problem, instead of leaving a transcript in a separate tool.
“Voice is where service is hardest to get right, and where it matters most to customers,” Kishan Chetan, executive vice president and general manager of Agentforce and Service Cloud at Salesforce, said in the release.
The competitive set is not only AI startups. Cresta also sits alongside contact center incumbents such as NICE, Genesys, and Five9. Salesforce says Cresta will keep supporting those platforms after the close, a promise that matters to buyers who do not want a forced migration.
The price is reported, not confirmed
The $2.7 billion figure is plausible but unconfirmed. Cresta was valued at $1.6 billion in its March 2022 Series C, according to the company’s announcement. At $2.7 billion, Salesforce would be paying about 1.7 times that mark, four and a half years later.
That sounds modest until you see what is missing. Cresta’s revenue, net revenue retention, and gross margin are private. Without them, 1.7 times tells you little about whether Salesforce overpaid.
Salesforce also did not say how it will pay. The release mentions only a definitive agreement, not cash, stock, or a mix. The early 2027 close gives both companies months to reprice or walk away.
What the published prices show
Salesforce and Cresta do not sell the same unit, so these list prices are not directly comparable. They still show where the money goes in a voice deployment.
| Option | Published price | What the unit covers | Main catch |
|---|---|---|---|
| Salesforce Agentforce Conversations | $2 per conversation | A completed conversation handled by an Agentforce agent | Telephony, implementation, and custom work can sit outside the unit price |
| Salesforce Agentforce Flex Credits | $500 per 100,000 credits | Actions across Agentforce, including voice and chat workflows | Consumption is harder to forecast than a per-conversation plan |
| Cresta | Not published | Enterprise voice agents, agent assist, quality, and analytics | Buyers need a quote and a pilot before they can compare |
| Deepgram Voice Agent, standard | $0.075 per minute | Speech-to-text, text-to-speech, and agent orchestration | The language model and business logic are priced separately |
The gap between $2 per conversation and $0.075 per minute is the point. A five-minute call costs about $0.375 in Deepgram orchestration before the language model, telephony, and engineering time. Salesforce’s $2 buys a packaged agent inside its CRM.
Salesforce’s Agentforce pricing page lists the conversation and Flex Credit rates. Deepgram’s pricing page lists the voice agent rate. Neither includes the people who make a production voice agent reliable.
Confirmed vs Not Confirmed
| Claim | Status | What supports it |
|---|---|---|
| Salesforce signed a definitive agreement to acquire Cresta | Confirmed | Salesforce press release, 2 November 2026 |
| The price is $2.7 billion | Not confirmed by Salesforce | The Information report, citing a person familiar with the deal |
| Cresta will be folded into Service Cloud and Agentforce | Confirmed as intent | Salesforce release and executive quote |
| The deal closes in the first quarter of fiscal 2028 | Confirmed as expected | Salesforce release, subject to regulatory approval |
| Cresta will keep supporting existing contact center platforms | Confirmed as a statement | Salesforce release, with no detail on duration or terms |
| Cresta’s containment and revenue-lift figures are audited | Not confirmed | Figures appear in Cresta marketing material, not an audit |
| Indian language support will improve after the deal | Not confirmed | No official statement on Indian languages or India road map |
What it means for Indian developers
India is one of the world’s largest customer support markets, and much of that work is voice. The deal matters less because of the CRM logo and more because of what it does to the build-versus-buy calculation.
The build path is now credible at small scale. Open voice stacks can handle speech recognition, synthesis, and orchestration for a few paise per minute, and the Deepgram and Pipecat combination made the plumbing easier to assemble. A team can pilot in Indian English, Hinglish, or a regional language without a six-figure enterprise contract.
The buy path still wins for regulated, CRM-heavy support. Salesforce already holds the customer record, case history, and compliance workflow, and Cresta’s coaching tools sit on top. For a bank or an insurer, the integration is the product.
The catch is language and data handling. Cresta’s public material talks about operating in any language, but neither company has published a detailed Indian language road map, latency benchmark from an Indian region, or data residency commitment. Buyers should ask for those in writing.
There is also lock-in. If Cresta’s voice agents become the front end for Service Cloud, the cost of switching CRM grows. The promise to support other platforms softens the risk but does not remove it. For Indian AI startups, the signal is consolidation. The remaining room is in languages, accents, compliance, and vertical workflows that a global CRM vendor will not tune first.
The skeptical read
Salesforce has bought a strong product in a hard category, but it has not published the numbers that would prove the price. The strategic fit is real. The financial case is still a story told by people close to the deal.
The integration risk is also underrated. Cresta’s customers chose a specialist that works with their existing contact center, not a CRM giant that wants to own the conversation. Salesforce’s promise to keep supporting those platforms is the right answer today. The incentive to favor its own stack will grow after the close.
Voice demos are easy to make impressive. Production voice is not. The winners will publish real containment, escalation, and cost-per-resolved-call numbers, not stage demos.
Frequently asked questions
Is the Salesforce Cresta acquisition confirmed?
Yes. Salesforce signed a definitive agreement on 2 November 2026. The price is the part it has not confirmed.
How much is Salesforce paying for Cresta?
Salesforce has not disclosed the price. The Information reported $2.7 billion, citing a person familiar with the deal. Treat that as reported, not official.
When will the Cresta deal close?
Salesforce expects the first quarter of its fiscal 2028, which runs from February to April 2027, subject to regulatory approvals.
Will Cresta still work with Genesys, NICE, and Five9?
Salesforce says Cresta will keep supporting the leading contact center and telephony platforms. It has not said for how long, on what terms, or whether every integration gets the same investment.
What should Indian voice AI teams do now?
Price both paths before committing. Compare a packaged Salesforce and Cresta quote with an open stack using Deepgram or a similar provider, then add telephony, LLM tokens, compliance, and engineers. Test Indian English and code-mixed calls, not just clean US audio.
Sources
- Salesforce, “Salesforce Signs Definitive Agreement to Acquire Cresta,” 2 November 2026. salesforce.com
- The Information, “Exclusive: Salesforce Acquires Startup Cresta for $2.7 Billion,” 3 November 2026. theinformation.com
- Cresta, “About Cresta,” accessed 3 November 2026. cresta.com
- Cresta, customer results and product claims, accessed 3 November 2026. cresta.com
- Salesforce, “Agentforce Pricing,” accessed 3 November 2026. salesforce.com
- Deepgram, “Deepgram Pricing,” accessed 3 November 2026. deepgram.com
- PR Newswire, “Cresta Raises Series C, Quadruples Valuation to $1.6B,” 17 March 2022. prnewswire.com