Salesforce’s Agentic AI Ambition Meets Data Access Backlash

Salesforce wants to automate the modern marketing team while keeping tighter control over the data that fuels it.

In its latest push into generative AI, the cloud software giant has rolled out Marketing Cloud Next, a suite of tools powered by autonomous AI agents that handle everything from campaign strategy to execution. These agents create audience segments, draft email and SMS copy, personalize website content in real time, and monitor ad performance—tweaking spend or messaging as needed, without human intervention.

The pitch is simple: let AI do the grunt work, so marketing teams can move faster and focus on strategy. “We’re going to see cost savings if we’re putting more of the segmentation decisions on AI and being more narrow in our targeting,” said Ariel Kelman, President and CMO of Salesforce. “If we can speed up the process of going from an idea to executing it, then the CMOs can say no less often.”

But the company’s simultaneous clampdown on data usage—especially inside Slack—is raising concerns across the enterprise software industry. Just weeks before launching its flagship AI product, Salesforce quietly revised Slack’s data usage policies to prohibit customers from training external AI models on their own Slack data. It also restricted third-party applications to only temporary access windows, limiting how external tools like Glean, an enterprise search startup can retrieve and work with customer-owned data.

Glean warned customers that Salesforce’s move essentially “locks your data within Slack.” Benchmark general partner Bill Gurley called it out more bluntly on X: “Enterprise customers need to declare if they support ‘open data’ or ‘closed data.’ And customers should move away from ‘closed data’ vendors as fast as possible.”

For a company publicly preaching AI openness and collaboration, the optics are uneasy: Salesforce is building AI agents designed to work across platforms while simultaneously closing its doors to AI competition.

Inside Marketing Cloud Next

Marketing Cloud Next is the most expansive application of Salesforce’s AI agent framework to date. Built natively on the Salesforce Platform, it introduces agents that can operate autonomously across campaign creation, engagement, lead management, and optimization. It runs on Agentforce, the company’s internal agent infrastructure that debuted last year with Ask Astro, a scheduling assistant for Dreamforce attendees.

According to Salesforce, marketers can give agents high-level instructions like “build a campaign to increase loyalty among at-risk customers,” and the system will execute—creating the customer journey in Flow, segmenting audiences, writing content, and setting channel preferences. The idea is to compress what once took weeks into hours.

“Agentforce transforms how work gets done, but it doesn’t replace marketers—it empowers them,” said Steve Hammond, EVP and GM of Marketing Cloud. “With Marketing Cloud Next, we’re giving marketers more than AI-powered features—we’re delivering a complete agentic marketing solution.”

The platform introduces several new capabilities:

• Create: Agents collaborate with marketers to generate campaign briefs, assemble content, and build journeys directly in Flow. They also pull from unstructured data in Google Drive, SharePoint, Zendesk, and customer support content via Salesforce Data Cloud.
  
• Engage: “Do-not-reply” emails and static landing pages are replaced by agents that hold live, multi-turn conversations via SMS, email, and web. Agentforce Personalisation Decisioning uses real-time behavioral signals to recommend offers, while Agentforce Web Curation dynamically adapts web experiences per visitor.
  
• Qualify: Agents qualify and nurture leads autonomously—initiating contact, drafting outreach, scoring intent, and escalating to reps when needed. Multi-language support enables global campaigns.
  
• Optimise: Agents monitor paid media in real time, pausing underperforming ads, reallocating spend, and recommending changes. Through Segment Intelligence, marketers can identify high-ROI audiences and adjust mid-flight. A new Salesforce–LinkedIn integration allows syncing customer lists, tracking conversions, and accessing insights within Marketing Cloud Next.
  

The product works alongside existing Salesforce implementations with no migrations or platform overhauls required. Features are available to current customers through new Marketing Cloud+ SKUs beginning this summer.

The Bigger Push Into Agentic AI

The timing of Marketing Cloud Next aligns with a broader enterprise trend: increased reliance on AI agents to manage internal operations. According to Accenture, the percentage of companies giving full operational control to AI rose from 9% in 2023 to 16% in 2024, with those companies reporting 2.4x higher productivity. The global AI agent market, estimated at $3.7 billion in 2023, is forecast to top $100 billion by 2032, according to S&S Insider.

Salesforce sees Agentforce as its competitive answer to agentic initiatives from Microsoft, Google, and OpenAI. Kelman said the technology has “progressed a lot” since its Dreamforce debut, especially in reasoning capabilities. Agents can now make dynamic business decisions such as which customer should get a discount, or when to trigger a follow-up email without step-by-step human instruction.

But Kelman was also clear that agentic marketing has a foundational requirement: clean, unified data. “Without well-structured data—both structured and unstructured—you’re never going to get good accuracy of responses and less effectiveness of whatever you’re trying to do,” he said.

A Growing Rift Over Platform Control

The innovation narrative around Marketing Cloud Next has been complicated by Salesforce’s policy moves on data access particularly within Slack. In a blog post published in late May, the company stated that the new restrictions are intended to “better protect customer data, uphold the integrity of Slack’s platform, and ensure its ecosystem continues to grow in a way that’s open and responsible.”

But for many third-party vendors building AI-enabled tools on top of Salesforce’s products, the changes are anything but open. Multiple startups have said privately that the new restrictions are forcing them to scale back key features or abandon Slack integrations entirely. “It’s really a bad look if they start preventing interoperability and integrated systems,” said one executive.

The debate over customer data rights isn’t new. Phil Fernandez, former CEO of marketing automation platform Marketo, shared on LinkedIn that Salesforce once asked Marketo to pay “a huge amount” for a shared customer to connect their CRM data. After he rallied their joint customers, Salesforce dropped the demand.

At a recent customer conference, Kelman showed a keynote slide with the phrase: “Your data is not our product,” emphasising Salesforce’s public stance on trust and data ethics. But that same data when it lives inside Slack can no longer be freely accessed, even by the customers who created it.