For decades, enterprises have poured their most valuable business knowledge into CRMs: customer histories, deal patterns, pricing logic, territory structures, and workflows refined over thousands of iterations. Most of that value stayed trapped behind logins, tabs, and manual navigation, accessible only to people willing to search for it. This year’s Dreamforce signaled the break: agentic workflows allow that accumulated intelligence to move beyond the platform and travel directly into daily operations, from active messaging threads to meeting prep and pipeline reviews. With more than 1,400 sessions delivered at the event, the underlying theme remained consistent. Here are three practical operational changes that illustrate how the Agentic Enterprise brings decades of stored data into active use.
1. Ambient CRM: Software Integrated into Daily Workflows
The adoption of ambient workflows, highlighted by announcements such as AIforce, Claudeforce, and Slackforce, reflects a transition in where software operates. Rather than requiring employees to open Salesforce to check deal health, prepare for client meetings, or update pipeline metrics, CRM logic and data permissions now run inside platforms like Claude and Slack.
“The shift is clear: the conversation has moved past what agents can do. It’s now about getting them into production and trusting what they do. AIforce made it clear that Salesforce is coming to wherever people already work, with its data, logic, and permissions intact.”
– Roland Berthelot, Global Head of Salesforce Studio at Globant.
This highlight addresses the constant context switching that slows internal operations. Instead of stopping tasks to feed data into a separate platform, teams interact with enterprise intelligence directly inside active work environments.
2. Unlocking Enterprise Data Through CRM Reasoning Models
Companies have spent decades building complex data models, metadata, and custom business logic inside legacy platforms. Historically, these core investments remained confined within rigid system boundaries.
Through Model Context Protocol (MCP) servers, open APIs, and specialized reasoning engines, such as Salesforce’s Koa model, developed in partnership with NVIDIA, enterprise context is becoming fully portable. Business logic can now power autonomous agents across an open ecosystem that includes AWS, Google Cloud, Anthropic, and OpenAI. This paves the way for how specialized models handle complex enterprise workflows:
“The evolution of Agentforce takes agents in Sales, Service, or Commerce to another level. They no longer just answer queries: they reason through complex processes, pursue long-term goals, and collaborate with each other. Specialized models like Koa demonstrate the massive value of having AI tailored specifically for enterprise logic.”
– Sergio Alvarez Alonso, Solutions Sales Specialist, Salesforce Studio at Globant.
The technical focus has moved from data retrieval to enabling agents to reason through multi-step operations while staying compliant with enterprise rules.
3. Practical Operating Models: Delegating Execution While Retaining Oversight
While autonomous agents demonstrate broad execution capabilities, large-scale implementation depends on strict governance, clear context boundaries, and direct human oversight. Frameworks like the Trusted Enterprise AI Harness provide the parameters needed to run these systems safely. Some keynotes emphasized the importance of having a strong ethical principle across every sales agent:
“Across every demo, the same core principle stood out: Human in the Loop. No agentic process proceeds without human approval, underscoring the value of strategic decision-makers. Take Salesforce’s agents, Piper and Hunter. Hunter alone generated $500 million in pipeline in a single quarter, proving where the industry is heading when autonomous execution is paired with human strategy.”
– Agostina Ardenghi, Media, Monetization, and Growth Director at GUT.
Under this setup, autonomous systems handle task execution, reserving strategic oversight for human teams. Putting this operating model into practice requires delivery structures that integrate intelligent agents into everyday workflows without losing visibility or oversight.
From Keynote Stage to Real Production
Dreamforce 2026 outlined the ecosystem’s technical direction, but converting platform announcements into governed enterprise adoption requires deliberate effort. Moving into production requires modern data foundations built on integration layers such as Data 360, MuleSoft, and Informatica, as well as a clear operational framework for human-agent collaboration. Organizations that are modernizing their technical foundations and deploying agentic workflows are establishing the standards for modern operational performance.
Planning the next stage of Salesforce architecture? Learn how Globant’s Salesforce Studio works with global organizations to modernize legacy systems, connect Data Cloud, and deploy production-grade agentic workflows. Explore Globant’s Salesforce capabilities here.
