On 2 November 2026 Salesforce signed a definitive agreement to acquire Cresta, a voice AI company whose agents resolve complex customer calls end to end and coach human reps in real time. Cresta will be connected to Service Cloud and Agentforce. Salesforce did not disclose the price; The Information reported about $2.7 billion. The deal is expected to close in Salesforce's first quarter of fiscal 2028, which runs February to April 2027.
- Cresta's voice agents already run calls for Marriott, Comcast and Alaska Airlines
- Salesforce says it will keep supporting the contact center and telephony platforms customers use today
- It is the latest in a run of 2026 CX deals that includes Fin and Braze
Salesforce announced the agreement in an official news post on 2 November 2026. CMSWire covered the deal the same day, citing The Information's report that Salesforce is paying about $2.7 billion and placing Cresta in a 2026 buying spree that has passed $15 billion in customer experience companies.
The short version for a sales or service lead: the phone is back on the CRM roadmap. Text chat agents arrived first because text is easier to log, search and audit. Voice is harder, and it is still where customers go when the problem is urgent, expensive or emotional. Salesforce is betting that an AI agent which can hold that call, and pass a clean record back to the CRM, is worth billions.
Most small teams will never buy Cresta, and they do not need to. The same pattern, a bot that answers routine questions and hands the rest to a person with full context, can be built on top of the CRM you already run. Companies such as YuSMP Group build those chat and voice assistants for mid-size businesses, which is usually cheaper than moving the whole service desk to a new platform.
What did Salesforce announce?
According to Salesforce, Cresta will join its service business and plug into two products: Service Cloud, its customer service application, and Agentforce, its platform for AI agents. The aim is to bring conversation intelligence from live calls straight into the customer record.
"Voice is where service is hardest to get right, and where it matters most to customers," said Kishan Chetan, executive vice president and general manager of Agentforce and Service Cloud. Cresta CEO Ping Wu said that joining Salesforce would let the company "realize our full vision at scale."
Two details matter beyond the headline:
- No forced migration, for now. Salesforce committed to supporting the contact center and telephony platforms customers already use. Analyst Blair Pleasant of COMMfusion told CMSWire that this "isn't necessarily about forcing customers into a single ecosystem."
- A long close. The deal still needs regulatory approval and is not expected to complete before spring 2027, so nothing changes in current Salesforce contracts this quarter.
What does Cresta actually do?
Cresta works on both sides of a call. Its voice agents can take a complex customer call from greeting to resolution without a person. When a human rep is on the line, the platform listens in real time, suggests the next step and flags compliance risks. After the call, it analyses every interaction so managers can see which behaviours close tickets and which ones lose customers.
Pleasant summed up the fit in one line: "Cresta understands the conversation, Salesforce understands the customer." That is the gap most CRMs still have. Notes from phone calls arrive late, short or not at all, which is why call data is usually the weakest part of a pipeline report. We saw the text version of this move earlier in the autumn, when Omnichat raised a Series A for AI chat agents, and in Salesforce's own purchase of Braze on the marketing side.
What does it mean for small and mid-size teams?
- Call notes are about to become automatic. Expect transcription, summaries and next-step suggestions to be a standard CRM feature within a year or two, not a premium add-on. Do not sign a long contract for a standalone call-recording tool today.
- Your CRM fields decide what the agent can do. A voice agent can only update records that exist. If deal stages, case reasons and contact roles are inconsistent, the AI will log garbage faster. Our guide to what a CRM database should contain is a useful checklist.
- Hand-off quality beats automation rate. Customers tolerate a bot that says "let me get a person" if the person already knows the problem. They do not tolerate repeating themselves.
- Integration is the real project. Telephony, CRM and ticketing usually come from different vendors. The CRM integration guide covers how to connect them before you add an AI layer.
What should you do now?
- Measure your call mix. Pull a month of inbound calls and tag them: order status, billing, how-to, complaint, sales. The repetitive half is your automation candidate.
- Fix call logging first. Make sure every call creates or updates a CRM activity with an outcome field. Without that baseline you cannot tell whether an AI agent helps.
- Pilot on one narrow journey. Appointment changes or order tracking are safer first tests than complaints or renewals. Track resolution rate and how often callers ask for a person.
- Keep a human in the loop. Route anything involving refunds, contracts or angry customers to a rep, and review a sample of AI-handled calls every week.
