Salesforce is making a calculated bet that the future of enterprise AI won’t belong to any single vendor. It belongs to whoever controls the connective tissue between them.
The San Francisco-based software giant is building a new tool that will allow third-party AI agents — built by competitors, startups, or customers themselves — to tap directly into the vast reservoirs of customer data housed inside Salesforce’s platform. The move, first reported by The Information, signals a significant strategic pivot: rather than insisting companies use only Salesforce’s own Agentforce AI products, the company is positioning itself as the central hub through which all manner of autonomous software agents can access the structured customer records that enterprises have spent years building.
It’s a shrewd play. And a necessary one.
The enterprise AI market is fracturing fast. Companies aren’t waiting around for a single vendor to deliver every AI capability they need. They’re stitching together agents from OpenAI, Anthropic, Google, Microsoft, and dozens of smaller firms, each optimized for different tasks — scheduling, customer support, sales outreach, supply chain management. The result is a messy, multi-agent world where the most valuable asset isn’t the agent itself but the data it can act on.
Salesforce, which stores customer relationship data for more than 150,000 companies worldwide, is sitting on exactly that asset. The new tool effectively turns Salesforce into a data access layer for the broader AI agent economy, regardless of who built the agent making the request.
From Walled Garden to Open Interchange
For years, Salesforce’s competitive strategy followed a familiar enterprise software playbook: build a comprehensive platform, make it sticky, and keep customers inside the walls. The company’s own AI efforts — from Einstein AI features introduced years ago to the more recent Agentforce product line — were designed to give customers reasons to stay within the Salesforce orbit.
That approach worked when AI was a feature bolted onto existing software. It’s less convincing now that AI agents are becoming standalone actors capable of executing complex, multi-step business processes on their own.
CEO Marc Benioff has been vocal about Salesforce’s AI ambitions. During the company’s most recent earnings call, he described Agentforce as the company’s fastest-growing product initiative, with thousands of enterprise deals already signed. But the competitive pressure is real. Microsoft’s Copilot agents are deeply integrated with Office 365 and Dynamics. Google is pushing its Gemini-based agents into Workspace and Cloud. Startups like Sierra, co-founded by former Salesforce executive Bret Taylor, are building AI agents specifically for customer interactions — the same turf Salesforce considers home.
Opening up its data layer to outside agents is an acknowledgment that Salesforce can’t win the agent war alone. But it can win the data war.
Think of it this way. An AI agent built by a third party might be better at handling a specific customer service workflow. Fine. But that agent still needs to know who the customer is, what they’ve purchased, when they last called, and what their contract terms look like. That information lives in Salesforce. By making it accessible through a standardized interface, Salesforce ensures it remains indispensable even when its own agents aren’t the ones doing the work.
The strategy carries echoes of what Amazon Web Services did with cloud infrastructure a generation ago. AWS didn’t need to build every application. It just needed to be the platform those applications ran on. Salesforce appears to be applying the same logic to customer data in the age of AI agents.
The Technical and Competitive Stakes
Details on the new tool’s architecture remain thin. According to The Information, the capability is designed to let external agents connect to Salesforce’s customer data stores in a controlled, permissioned way — meaning enterprises would retain governance over what data gets shared, with whom, and under what circumstances. That governance layer will be critical. Enterprises won’t hand over their CRM data to outside AI agents without strict access controls, audit trails, and compliance guarantees.
This is where Salesforce’s existing infrastructure gives it an edge. The company has spent years building out its data platform, including the acquisitions of MuleSoft (for integration), Tableau (for analytics), and more recently its Data Cloud product, which unifies customer data across multiple Salesforce applications. A tool that extends Data Cloud’s reach to third-party agents would be a logical extension of work already underway.
But execution risk is substantial. Interoperability between AI agents from different vendors is still an unsolved problem across the industry. There’s no universally accepted standard for how agents authenticate, request data, or handle errors when interacting with external systems. Anthropic has proposed its Model Context Protocol (MCP) as one approach. OpenAI has its own function-calling conventions. Google has its own agent frameworks. Salesforce will need to either support multiple protocols or push its own — neither of which is simple.
And then there’s the trust question. Will competitors actually want their agents connecting to Salesforce’s data layer? Microsoft, for instance, has every incentive to pull CRM data into its own Dynamics 365 platform rather than route its Copilot agents through a Salesforce-controlled pipeline. Google faces similar incentive conflicts. The tool may prove most useful for smaller agent developers and enterprises building custom agents internally — a large but less glamorous market.
The competitive dynamics here are layered. Salesforce isn’t just competing with other CRM vendors anymore. It’s competing with hyperscalers for control of the enterprise AI stack. Every major cloud provider wants to be the place where AI agents live, think, and access data. Salesforce’s advantage is specificity: it doesn’t have general-purpose data. It has customer data — the most commercially valuable data type in most enterprises.
Wall Street has noticed the shifting terrain. Salesforce shares have been volatile over the past year as investors try to price in the company’s AI potential against the risk that generative AI could commoditize traditional CRM software. The stock is up roughly 13% year to date, but it has underperformed the broader tech sector. Analysts at firms including Morgan Stanley and Goldman Sachs have flagged the AI agent opportunity as both Salesforce’s biggest growth lever and its biggest existential risk, depending on whether the company can position itself at the center of the emerging multi-agent architecture or gets relegated to a data source that smarter agents simply query and move on from.
What This Means for Enterprise Buyers
For CIOs and technology leaders evaluating their AI agent strategies, Salesforce’s move carries immediate practical implications. The promise of a single integration point for multiple AI agents — all drawing from the same trusted customer data — could simplify what is currently a nightmarish integration challenge. Most large enterprises run dozens of AI experiments simultaneously, many of which need CRM data but lack a clean way to get it.
A standardized data access tool from Salesforce could reduce the custom integration work required for each new agent deployment. It could also give IT departments a single control plane for managing which agents can see what data — a governance capability that barely exists today in most organizations.
But buyers should be clear-eyed about the trade-offs. Routing all agent data access through Salesforce deepens dependency on the platform. It makes Salesforce harder to replace, which is precisely the point. Companies that are already considering alternatives to Salesforce — and there are many, given the platform’s cost — should think carefully about whether this new capability locks them in further.
There’s also the question of pricing. Salesforce hasn’t disclosed how the new tool will be monetized, but the company’s recent moves suggest it will tie agent-related capabilities to premium tiers or consumption-based pricing. Agentforce already carries its own pricing structure separate from core CRM licenses. A data access tool for third-party agents could become another revenue line — or another line item on an already expensive Salesforce bill.
So where does this leave the market? In a state of rapid, somewhat chaotic evolution. The AI agent space is moving faster than any single vendor can control. Salesforce’s decision to open its data layer is less an act of generosity than a recognition that openness, in this particular moment, is the best defensive strategy available. If every AI agent needs customer data, and Salesforce has the customer data, then Salesforce wins — no matter whose agent is asking.
That’s the theory, anyway. The execution will determine whether Salesforce becomes the indispensable backbone of the multi-agent enterprise or just another API endpoint in an increasingly crowded field.
WebProNews is an iEntry Publication