Salesforce agreed to acquire voice AI firm Cresta for $2.7 billion on November 2, 2026. Here is what the deal means for service businesses that live and die by the phone.
Ido Cohen · Published 2026-11-03 · CRM
Salesforce just agreed to pay $2.7 billion for Cresta, a voice AI platform that coaches phone reps in real time and can handle entire customer calls autonomously — and this is the twelfth acquisition Salesforce has closed or signed in 2026 alone. If your service business runs on inbound and outbound phone calls — and most do — this deal is a signal you cannot ignore.
Cresta is not a simple call-recording tool. According to Futurum Research's analysis of the deal, Cresta brings three capabilities that most voice AI vendors sell separately: autonomous voice agents that handle complex calls end to end, real-time coaching for human representatives during live calls, and conversation analytics across every interaction.
The real-time coaching piece is what sets Cresta apart from rivals. During a live call, Cresta's platform surfaces relevant knowledge, suggested responses, compliance reminders, and behavioral coaching cues in real time — not in a post-call report you review the next morning. G2 reviewers have consistently called the real-time coaching the standout capability, with one contact center director noting that representatives were getting prompts mid-conversation "without waiting for call reviews with a manager."
The autonomous voice agent side of the platform holds natural, real-time conversations, handles interruptions, and runs at enterprise scale. Salesforce named Marriott, Comcast, and Alaska Airlines as existing Cresta customers — companies running millions of calls per month.
Salesforce is on a buying spree that would make even Marc Benioff's most aggressive investors raise an eyebrow. According to Axios's reporting, Cresta is the twelfth deal Salesforce has signed this year, coming just days after the company announced a $5.66 billion deal to acquire marketing platform Braze and a separate acquisition of logistics supplier Rune Technologies.
The pattern is obvious: Salesforce is trying to own the full customer interaction layer — before the sale (marketing, via Braze), during the sale (CRM, Agentforce), and after the sale (service, via Cresta). Kishan Chetan, EVP and GM of Agentforce and Service Cloud at Salesforce, framed the rationale bluntly: "Voice is where service is hardest to get right, and where it matters most to customers."
That framing is strategically important. Every other channel — chat, email, social — has been relatively easy to automate. Voice is harder because it requires real-time response, context retention across a conversation, and the ability to handle interruptions. By acquiring Cresta, Salesforce is betting it can close that gap inside its Service Cloud platform, which already sits at the center of how mid-market and enterprise businesses manage customer relationships.
The transaction is expected to close in the first quarter of Salesforce's fiscal year 2028, subject to regulatory approval, and Cresta will continue supporting existing contact center and telephony integrations in the meantime.
Here is the honest read: most service businesses — a 12-person plumbing company, a dental group with four locations, a personal injury law firm — are not Cresta's target customer today. Cresta's pricing is enterprise-grade (roughly $150,000 per year on some marketplace listings), and it is engineered for high-volume call centers.
But the acquisition matters to you for three reasons.
First, it accelerates the voice AI timeline for everyone. When Salesforce bakes Cresta's technology into Service Cloud, the capabilities eventually trickle down to smaller businesses through Salesforce's SMB tiers, AppExchange partners, and third-party CRM integrations. Competitor CRMs — HubSpot, Zoho, ServiceTitan, Jobber — will respond. Voice AI coaching that costs $150K per year today will cost $200 per month in 36 months. That is how this market has worked with every prior technology wave.
Second, your phone calls are already your biggest data blind spot. Most service businesses track web leads, form fills, and email opens obsessively — but have no structured data on what happens on the phone. How many calls convert to booked jobs? What objections come up most often? Which staff member's close rate is 60% and which is 30%? Right now, you probably have no idea. Voice AI — whether Cresta inside Salesforce or a standalone tool — answers all of those questions.
Third, your competitors will move faster than you expect. Franchise groups, private equity-backed home services networks, and multi-location dental and legal groups are already piloting voice AI. If you operate in a market where any of your competitors have institutional capital behind them, assume they are testing this now.
You do not need Cresta. You need the same outcomes — real-time coaching, call analytics, and eventually autonomous booking — at a price that makes sense for a business doing $1M–$10M in revenue. Here is the current realistic stack:
The SMB voice AI tools are maturing fast. Dialpad Ai surfaces real-time coaching prompts during calls on a plan priced for small teams. Balto provides live call guidance specifically tuned for sales and service calls. Air.ai and Synthflow can handle inbound calls autonomously — booking appointments, answering FAQs, and routing complex inquiries — for a fraction of Cresta's enterprise pricing.
None of them are as sophisticated as Cresta. But sophisticated is the wrong benchmark for most service businesses. Good enough + actually deployed beats perfect but never implemented every time.
Here is the part most AI marketing coverage glosses over: voice AI creates a data concentration problem.
When you route your calls through any of these platforms — Cresta, Air.ai, CallRail, or anything else — you are handing a third party a recording of every conversation your business has. That includes pricing discussions, customer complaints, negotiation details, and any sensitive client information (health conditions for med spas, legal matters for attorneys, financial situations for financial advisors).
Before you deploy any voice AI tool, answer these questions:
This is not a reason to avoid voice AI. It is a reason to read the terms before you click "Start free trial."
The Salesforce-Cresta deal closes in roughly 12–15 months. That gives you a window to get ahead of this shift rather than scramble to catch up.
This week, do three things:
1. Audit your current call data. Pull the last 30 days of inbound call volume from your phone system. If you cannot answer "what percentage of calls converted to booked appointments," you have a measurement gap that needs to close before any AI tool can help you.
2. Start a free trial of one call intelligence tool. CallRail offers a 14-day trial and integrates with most CRMs. Fathom is free for call recording and AI summaries on video calls. Pick one, connect it to your busiest line or booking team, and use it for two weeks. You will immediately see which call topics come up most and where your team loses prospects.
3. Check your call consent disclosures. Add a "this call may be recorded" disclosure to your phone greeting if you do not already have one. It takes five minutes and protects you legally before you add any recording software.
The businesses that will benefit most from the voice AI wave are not the ones who wait for Salesforce to productize Cresta's tech in 2028. They are the ones who understand their call data now, so they know exactly what they want an AI to improve.
---
What is Cresta and why did Salesforce buy it?
Cresta is a voice AI platform that handles customer calls autonomously, coaches human representatives in real time during live calls, and analyzes every interaction for performance insights. Salesforce agreed to acquire Cresta for a reported $2.7 billion on November 2, 2026, to add voice AI capabilities to its Service Cloud product — completing its ambition to own the full customer interaction lifecycle from marketing through post-sale service.
Does the Salesforce-Cresta deal affect small service businesses directly?
Not immediately. Cresta is an enterprise product priced for high-volume contact centers, and the deal won't close until early fiscal 2028. But the acquisition signals that voice AI is moving from a novelty to a standard feature inside mainstream CRMs, which means SMB-priced alternatives will accelerate in capability and decline in cost over the next 12–24 months. The indirect effect on your competitive landscape is real and starts now.
What voice AI tools can a service business realistically use today?
Several SMB-accessible options exist. CallRail and Fathom handle call recording and transcription at low cost. Dialpad Ai and Balto provide real-time coaching prompts for live calls. Air.ai, Synthflow, and Bland.ai can handle inbound calls autonomously — answering questions, booking appointments, and routing calls — at pricing that starts under $500 per month. None are as sophisticated as Cresta, but all provide enough value to materially improve how a service business handles phone leads.
Is there a legal risk to recording customer calls with AI tools?
Yes, and it is state-specific. Twelve U.S. states — including California, Florida, Illinois, and Washington — require all-party consent before recording a phone call. If your business operates in or takes calls from customers in those states, you must disclose that the call is being recorded and obtain consent. Additionally, businesses serving HIPAA-covered clients (medical practices, med spas, etc.) need to ensure any voice AI vendor will sign a Business Associate Agreement before handling protected health information.
How should I think about first-party call data before adopting voice AI?
Voice AI tools are only as useful as the calls they analyze. Before adopting any platform, establish a baseline: track your call volume, conversion rate from call to booked appointment, and the most common reasons calls do not convert. Even a simple call log in a spreadsheet beats starting from zero when you deploy an AI tool. Businesses with clean call data get dramatically faster results from voice AI because the system has meaningful patterns to learn from, rather than starting blind.
---
Sources: