Skip to content
World clockEU--:--UK--:--USA--:--CN--:--PLDEFRIT中文EN

portal about AI and technologyevents · analysis · interviews · technical background

Search
LIVE
›

Salesforce's reported $2.7B Cresta deal tightens the AI customer service field

Salesforce said on 2 November 2026 that it had signed a definitive agreement to buy Cresta, a customer service AI vendor. The Information reported the price at $2.7 billion, a figure Salesforce has not confirmed.

AI & modelsAnalysisGrace OkonkwoPublished: 4 November 20264 min readSources 27
Salesforce's reported $2.7B Cresta deal tightens the AI customer service field

The Information reported that the cloud giant is paying $2.7 billion for Cresta Intelligence Inc. Cresta is a venture-backed provider of customer service software. Closing is expected in the first quarter of Salesforce's fiscal 2028. The deal remains subject to regulatory approvals and customary closing conditions.

Salesforce previously agreed to buy Fin Inc., formerly Intercom Inc., for $3.6 billion. Its largest acquisition of the year is the $5.66 billion purchase of Braze Inc. Both targets sell AI software that answers or personalizes customer interactions. That makes Cresta the second multibillion-dollar customer service deal of the year for Salesforce.

What Cresta brings: voice agents and human assist

Salesforce said Cresta's voice agents resolve complex customer calls end to end. They hold natural real-time conversations and handle interruptions. Marriott, Comcast and Alaska Airlines are among the brands using the platform at enterprise scale, according to Salesforce. Cresta's documentation describes shared memory. It preserves context across channels and handoffs between human and AI agents.

Cresta's Agentic Assist gives human representatives real-time hints, reminders and workflows tailored to each conversation. A separate agent called Deputy targets complex tickets that need several employees to process. It collects the information and loops in the right workers. Opera lets organizations fine-tune AI models for specific service use cases. Synthetic Customers generates AI-written support conversations for testing new agents.

Why the deal matters beyond the price

Salesforce described Cresta as a voice-first operating system for customer experience. Kishan Chetan, executive vice president and general manager of Agentforce and Service Cloud, framed voice as the most difficult and consequential service channel. Ping Wu, Cresta's chief executive, said the combination would embed real-time conversation intelligence and agentic experiences directly into Salesforce's customer service platform.

For buyers, the near-term effect is a narrower field of independent vendors. They will also see more Salesforce bundling of voice agents into Service Cloud. Neither company has said how Cresta's tools will be priced or packaged once the deal closes.

Salesforce bought its way into marketing automation with Braze and into support automation with Fin. Cresta fills the voice gap between them. Whether customers gain depends on whether the pieces ship as one product or as three.

Rivals push integration, speed and scale

Observe.AI announced a Headless Companion Agent on 3 November 2026. It adds no new application window. The deployment model puts AI guidance inside existing CRM and agent desktop systems. The company said the system processes live interactions and sends structured data to customer-configured endpoints through an event-driven architecture. Swapnil Jain, Observe.AI's chief executive, said companies should not have to redesign the agent desktop to use AI.

Scale claims are getting specific.

KT Corp said its voicebot handles an average of 1.2 million customer inquiries a month. The company called that the highest volume in the industry. First-contact resolution rose 4 percentage points, from 77 percent to 81 percent, since the voicebot launched. Average call handling time fell by 31 seconds, or 11 percent, after KT Corp deployed its AI consultation assist tool.

Ubase said it filed patents for multi-agent technology. The approach cut response times to under 0.8 seconds. That compares with 2 to 3 seconds under a single large language model approach. It said its accuracy in distinguishing intent behind inquiries about similar products exceeds 97 percent.

Governance and human connection become the test

Forrester predicts that by 2027 AI alone will no longer be a competitive advantage for customer service organizations. The firm expects personal AI assistants to become the main service interface for 10 percent of consumers. Forrester's Retail Topic Insights 1 Survey found 35 percent of US online adults had used assistants like ChatGPT, Claude and Gemini to research or shop in the past year. The figures were 41 percent in the UK. They were 49 percent in Hong Kong.

Governance is the other half.

Karthik Suri, chief product and corporate strategy officer at Sprinklr, said: "Trust has become the foundation that enables scale. Without governance, organizations hesitate to expand AI adoption." He described a governance frontier. It sets where AI can act independently, where human approval is required and where escalation is needed.

Forrester also expects twenty global B2B companies to build autonomous, cross-system service networks. They would partly answer bot-initiated requests.

Comments 0

Sources

27
  1. 01Salesforce to buy customer service automation startup Cresta for reported $2.7B - SiliconANGLEEN
  2. 02Salesforce Agrees to Buy Cresta for Voice-Led Customer Service AI - Unite.AIEN
  3. 03Observe.AI Introduces Headless Companion Agent for Customer ServiceEN
  4. 04Scale and Trust: The Two Tests for AI-Powered Customer Service - CX TodayEN
  5. 05Predictions 2027: Human Connection Will Be The Next Battleground For Customer Success And Service - ForresterEN
  6. 06KT Corp accelerates AI overhaul of customer service, handling 1.2 million calls a monthEN
  7. 07Ubase Files 2 Patents to Fix AI Customer Service "Endless Loops" — Response Time 0.8 Seconds - BigGo FinanceEN
  8. 08Android-Smartphones sichern Fotos jetzt automatisch auf Windows-PCsEN
  9. 09 Nvidia tells data center operators to limit business from Chinese customers with new global restrictions, report claims EN
  10. 10Nvidia pushes cloud customers to cut back on China businessEN
  11. 11Huawei tells customers to find their own AI chip memory, effectively outsourcing procurement to the black marketEN
  12. 12CyberKord raises €2.7M to automate IT and cybersecurity operations with AIEN
  13. 13Google Launches LiteRT Tensor API for Building, Customizing, and Optimizing ML Models for On-DeviceEN
  14. 14Aidbase – AI Customer Support Platform, Knowledge-Based Chat & Assisted Ticketing - Quasa.ioEN
  15. 15Retail’s Next AI Imperative: Customer Experience ObservabilityEN
  16. 16Smells like AI: Why customer trust breaks when automation shows up in the wrong momentEN
  17. 17AI Customer Service Success Takes More Than Deflecting TicketsEN
  18. 18 Nvidia Brev security breach may have exposed SSH keys used to access customer VMs EN
  19. 19Chinese automakers reportedly head to Beijing for talks on EU hybrid export quotasEN
  20. 20Cities Worldwide Build and Scale AI Services With NVIDIA Physical and Agentic AIEN
  21. 21SUSE to Cut Roughly 10% of Its Workforce, Citing AI and Automation - XenoSpectrumEN
  22. 22From 1D Benchmarks to Complex 3D Geometries: Can Agents Reliably Specify Problems for Automated PDE Simulation?EN
  23. 23Launch HN: Momentic (YC W24) – Agents that catch bugs by using your appEN
  24. 24Chainalysis Confirms Intrusion Into Internal Tools and Exposure of Customer Identity DataEN
  25. 25Management - Ensuring trust in public service AI - Business ReporterEN
  26. 26AI Behind Critical Moments: How Automation and Humans Work Together - CX TodayEN
  27. 27Global Satcom Services Revenues to Reach $145 Billion by 2035EN

All figures and quotations in this text come from the sources listed below.

Content prepared by the editorial team with AI assistance.

Grace Okonkwo

Grace Okonkwo

AI, models and technology

Grace Okonkwo covers AI, models and technology for FLASH24, working from primary sources such as model cards, API documentation and benchmark papers rather than vendor summaries. She checks training data provenance, evaluation conditions and reported scores against the underlying datasets before any figure reaches print. She interviews researchers and engineers directly, tracks release calendars from major labs, and compares successive model versions on the same tests. Her own self-hosting, home-network and documentation-reading habits feed straight into that desk, since she tests tools on her own hardware first. She does not publish benchmark claims without a reproducible method.

Newsroom →

Comments

0
  1. No comments yet — be the first.

Write a comment

Comments are public. We do not publish abuse, spam or advertising.