Salesforce is paying $5.66 billion to act on customer behavior in real time, not for another AI model
Salesforce agreed to buy Braze, whose event-streaming platform turns live app and web behavior into messages across channels. For agentic marketing, the gap was never ideas. It was execution.
Listen to this article · 5 min
AI-generated narration of the full article.
Marketing AI has had a strange gap for years. Models got very good at deciding what a customer should hear next: which offer, which product, which tone. The systems that actually send the message often learned about the customer’s behavior hours later, in a nightly batch, by which time the customer had bought elsewhere, churned or already received three contradictory emails.
Salesforce’s announcement on October 20 is a bet on closing that gap. It has signed a definitive agreement to acquire Braze for about $5.66 billion in enterprise value, paying $47.50 in cash per Class A share.
What Salesforce is buying
Braze is a customer engagement platform built on an event-streaming architecture: it takes behavioral events from mobile apps and websites as they happen and coordinates personalized messages across apps, web, email, SMS and WhatsApp. Its AI feature, BrazeAI, optimizes content, offers, channels and timing. It has about 2,770 customers, mostly in high-volume consumer businesses.
Salesforce says Braze would strengthen cross-channel personalization in Marketing Cloud within Customer 360, and complement Data 360, its customer data platform, and Agentforce, its agent platform.
The deal has not closed. Salesforce expects completion in its fiscal year 2028, which begins in February 2027, subject to Braze shareholder approval and regulatory clearances. Until then, Braze and Salesforce remain separate products, and every integration described is a plan.
Decision is cheap; timely execution is not
The reason this matters for enterprise AI is architectural. A customer-facing agent is only as good as the latency of its loop: observe what the customer did, decide what to do, act in a channel, record the result. Most marketing stacks are fast in the middle and slow at both ends. The decision is a model call measured in milliseconds; the observation arrives in a batch file and the action waits for a campaign schedule.
Event streaming fixes both ends. That is why the acquisition reads less like a marketing feature and more like infrastructure: Salesforce is buying the part of the loop that its agents cannot reason their way around.
It also follows a pattern we have seen in other systems. The value of an AI agent increasingly depends on who controls the system that commits the action, the point we made about the contest for the transaction layer. In customer engagement, the transaction is the message actually sent.
What enterprises should watch
Identity resolution. Real-time personalization requires that an anonymous app event, a logged-in web session and a CRM contact are recognized as the same person, quickly and correctly. Two platforms with two identity models do not merge themselves. Ask how identities will be resolved between Braze and Data 360 before planning around the combination.
Consent at the speed of events. A faster loop means more messages decided automatically. Every one needs a valid basis: opt-in for SMS and WhatsApp, the channel’s own rules, and in the U.S., state privacy laws that give consumers rights to opt out of targeted advertising and the sale or sharing of their data. Consent state has to travel with the event, not be checked in a nightly job.
Who decides what the agent may send. When an agent chooses offer, channel and timing, marketing loses the review step that a campaign calendar used to provide. Brand, legal and compliance rules have to become constraints the system enforces, and frequency caps across channels become a control, not a courtesy.
Overlap and migration. Many enterprises already run Marketing Cloud, Braze, or both. Roadmaps after acquisitions tend to consolidate. Customers of either product should ask for written commitments on support and migration paths, and avoid new multi-year dependencies on integration features that do not exist yet.
The deal says something clear about where Salesforce thinks the bottleneck is. It is not the intelligence of the next best action. It is getting that action to the customer while it is still the best one.