The dust has settled on Dreamforce, Salesforce’s well-attended San Francisco fete. Our takeaway? Salesforce is setting the pace with a platform for AI agents embedded in business applications, and it’s made huge progress on its products and partnerships in the last year. Salesforce claims that its (newly renamed) Agentforce 360 product has 6,000 paying customers. But in customer conversations and sessions, we saw little adoption or impact from AI agents — lots of potential but a long way to go for a meaningful ROI.

Salesforce threw around the term agentic enterprise but failed to define it. We think this is referring to a Salesforce customer who ubiquitously uses AI agents, defined by Forrester as AI applications tuned to act on behalf of an enterprise or individual, performing tasks, making decisions, and interacting with data or other systems autonomously. Agentic AI, according to Forrester, entails systems of foundation models, rules, architectures, and tools which enable software programs to flexibly plan and adapt to resolve goals by taking action in their environment, with increasing levels of autonomy.

Here’s our take on Dreamforce 2025 for existing and potential enterprise customers. Salesforce has:

  • Fixed its branding (thank goodness). Salesforce, with its “digital labor” drumbeat, was one of the original sinners in telling CEOs and boards of directors that AI equals people, so who needs people? Regardless of CEOs pounding their shoes on the table, managers and employees actively resisted using work-provided agents. Who would use a thing designed to take away their job? It’s also balderdash: AI agents are power drills, not people, and only people know where to drill and why. In an impressive feat of company transformation, Salesforce CEO Marc Benioff fixed the branding snafu and reenergized his company by walking away from the job-destroying sound of AI as “digital labor” to the more inspiring statement that “AI elevates human potential.” We think this positioning frames agents for what they are: power tools that make your employees and customers more successful.
  • Made a sales pitch for your enterprise — not just your CRM — agents. Salesforce has doubled down to build out the Agentforce product to be an enterprise-grade agentic platform. We like this for your CRM applications, but Salesforce is not stopping there and aspires to be your enterprise agent platform. It’s not the only one: Business software providers, hyperscalers, data platforms, iPaaS players, search engines, AI platforms, ITSM providers, and ERP vendors all aspire to joining this land grab. We like Salesforce’s security model (which masks PII and other protected data when making a call to a language model), Amazon hosting for scale (as well as a deal with Google for hosting coming), Slack providing a coordination layer (to connect and activate “data and dialogue”), and a consumption-based pricing model option. We advise enterprises to start with targeted sales, marketing, or customer service agents with Agentforce before attempting full-scale automation projects across functions such as supply chain. It’s no surprise that the Dreamforce keynote’s case studies focused on narrow use cases and lacked details on metrics, ROI, or measurement of reduced burden on HITL cognitive load.
  • Showcased commitment to AI agents, but core product innovation was shown to be incremental. There’s no doubt where the “no software” company sees its (and your) future: enterprise AI, not just software. It renamed its entire portfolio from Cloud to Agentforce (e.g., Marketing Cloud to Agentforce Marketing). It goes beyond marketing: From what we saw and heard, Salesforce has been focused on expanding AI capabilities for each of its products and unifying products on the platform to make them AI agent-ready. Many other software companies are following a similar strategy, but Salesforce’s advantage is the hundreds of packaged AI agents for its core and industry clouds. Enterprise users will need to choose between leveraging those out-of-the-box agents but give up on the breadth beyond Salesforce — or go with a vendor that has limited prebuilt agents but spans into other apps to avoid lock-in. Innovation on core products was not center stage.
  • Doubled down on the data layer. AI agents without proprietary data and knowledge to ground them in your enterprise reality are just consumer-grade tools used at work. The key difference is grounding your AI agents in your data (which you know and already fund) and your knowledge assets (which lie scattered in file systems, repositories, and documents with ambiguous stewardship). Salesforce aims to solve that by asking you to move all your knowledge and data into its newly renamed Data 360 product. This strategy will likely bring much more accuracy and capability for the enterprises that do begin to shift their data gravity — but in the agentic world, Agentforce must play well with others or be left out of the game. We don’t see Data 360 replacing Databricks or Snowflake vector databases. But we do like the informal positioning of Data 360 by a senior product strategist as the agent grounding card catalog rather than the knowledge repository.
  • Offered mixed announcements on how MuleSoft will enable the agentic enterprise. Agent Script is a good step toward bringing deterministic outcomes through rules and outcomes, but it appears limited for complex tasks. We were disappointed that there was no similar announcement leveraging MuleSoft iPaaS to handle more complex scenarios. In contrast, MuleSoft’s Agent Fabric announcement was quite interesting. This leverages the API management side of MuleSoft to bring cataloging, governance, agent orchestration, and runtime visibility across all agents — even those outside of Agentforce. A control tower to govern your portfolio of agents across vendors will be necessary. To be a truly agentic enterprise, Saleforce must recognize that enterprise AI strategy will be multivendor and that any orchestration platform must address multivendor agent sprawl.

Salesforce’s vision for an agentic enterprise is ambitious and timely. But ambition alone won’t guarantee success. Customers need clarity on architecture, governance, and cost, not just pricing and a pragmatic roadmap to success. Most of all, they need evidence that AI agents pay off in a proper return on investment.

Interested in connecting with one of us about Salesforce’s strategy? Clients can reach out to schedule a call with our analysts at inquiry@forrester.com.