Salesforce has agreed to acquire Cresta. The product question is what happens to a customer’s request after the call ends. The announcement provides no repeat-contact rate for the combined offering. Calls answered by a voice agent therefore cannot yet establish an improvement in service quality.Announcement
The official announcement is dated 2026-11-02 without a publication time or timezone. This article uses a November 3, 15:45 Taipei cutoff. New reporting that Taipei day puts the price at $2.7B; Salesforce has not disclosed it. Closing is expected in the first quarter of fiscal 2028, meaning February through April 2027, subject to regulatory approval and other conditions. This is an agreement, not a completed acquisition.Price reporting Closing conditions
Cresta already offers voice agents and real-time assistance for human representatives. Salesforce names Marriott, Comcast and Alaska Airlines among its customer brands, without detailing each deployment or its results. Those names establish enterprise use cases, not evidence that agents already complete every call.Products and customers
The plan is to combine Cresta with Service Cloud, connecting calls to customer records and subsequent work. Existing voice capabilities should be distinguished from future integration results: the announcement gives no integrated-product price, general availability date or comparable independent test results. The buyer gains an opportunity to combine capabilities; customers still await the actual offering.Integration plan
Commitments made on calls need a follow-up status
Imagine a traveler calling to change an itinerary. An agent confirms the request, but a fee waiver still needs a supervisor’s approval. If the system saves only a fluent call summary, the next representative must reconstruct what was promised and what remains unauthorized. This hypothetical illustrates the work that product design must address between understanding speech and completing a service.
I would locate the integration’s value in that handoff. When the customer contacts the company again, a human representative should be able to continue the same request, see the pending approval and identify its owner, rather than ask the customer to start over. The agent should also distinguish an application submitted from a change completed. Preserving these states could make the CRM record useful for subsequent action. Adding a transcript alone may not reduce duplicated work.
That design also changes how an enterprise should assess automation. Suppose the share of calls ended without a human increases, but customers repeatedly call back because their problems remain unresolved. Service costs may simply occur later. I would put more weight on whether the same case is completed and how much information a human needs to ask for again after a transfer. These are proposed evaluation measures, not improvements the companies have reported.
Salesforce promises continued support for customers’ existing major telephony and contact-center platforms after closing.Compatibility commitment Preserving that choice could let companies separate the decision to replace their telephone system from the decision to adopt voice agents. The promise does not yet explain which case states each connection can exchange. If only some connections preserve complete handoff information, customers will receive different levels of integration value.
The commercial rationale is to bring call-handling capabilities closer to the system that executes service work. Its success still depends on whether the next person or agent can continue that work. Once the combined product arrives, stronger evidence than automated call volume would be fewer repeat contacts for comparable cases, without an increase in unfinished requests.
The cover reuses the site’s historical Open Secure AI Alliance graphic containing the Salesforce logo. It is not a Cresta product image or an image of this acquisition. Downloading the official event image failed because of DNS resolution.
Sources: