The 80% Context Rebuild: A Hidden Velocity Tax
Salesforce environments are increasingly defined by what one vendor calls a “Velocity Tax”: a structural cost baked into everyday change work. Analysis of 12,290 interactions with an AI agent in Salesforce shows that enterprise teams spend about 80% of their time reconstructing system context before they can safely execute any modification. Only 1.2% of interactions actually culminate in a change being made. Instead of a streamlined flow from idea to implementation, work is split into understanding, planning, and only then action. Administrators must trace dependencies, inspect automations, and verify permissions to answer basic questions such as what exists, where it is used, and what might break. This extensive system forensics is rarely visible on roadmaps or sprint metrics, yet it dominates day-to-day Salesforce context switching and quietly erodes enterprise developer productivity.
Why Salesforce Context Switching Dominates Admin Work
The research highlights a sobering pattern: among the most active users, 89% of activity occurs in the understand and plan stages. Instead of launching into configurations or code, teams start by untangling years of accumulated metadata and logic. They review flows, fields, automations, and access rules to reconstruct how the system actually behaves. Labels like “DEPRECATED,” “DO NOT MODIFY,” and “DO NOT DELETE” appear in 7.1% of all interactions, signalling an informal governance layer built on fear of unintended consequences. As organizations grow, staff churn and one-off projects layer on new customisations without a coherent map. The result is a CRM system overhead where each small change demands extensive investigation. Planning work more than doubles after 9 p.m., suggesting that context rebuilding spills beyond normal hours and puts sustained pressure on teams tasked with keeping Salesforce both stable and adaptable.
AI Agent Execution: Just 1.2% of Interactions End in Change
Despite the promise of automation, AI agent execution in Salesforce remains a tiny sliver of activity. Only 1.2% of the observed interactions end in an actual change to the environment. Most AI usage clusters around investigation and planning rather than direct modification. Tools can generate new flows, fields, and code quickly, but they often operate without a complete dependency graph. That gap raises the risk of downstream breakages discovered only after deployment. Every AI-generated object adds to the complexity that future admins must untangle, creating a new strain of technical debt: time saved upfront, paid back later in diagnosis and repair. Early Agentforce deployments show a similar trajectory, with initial emphasis on implementation followed by rising investigation rates as systems mature. Without deeper system understanding, AI agents risk amplifying Salesforce context switching instead of reducing it.
The Cost of Complexity and the Rise of System Forensics
The cost of this context-heavy model is more than theoretical. The report estimates that individual administrators spend between 620 and 1,040 hours a year reconstructing system context, equating to USD 42,000 to USD 70,000 (approx. RM193,000 to RM322,000) annually per admin. For a 10-person team, that Velocity Tax can reach USD 700,000 (approx. RM3,220,000), absorbed largely in work that happens before any visible delivery. As complexity compounds, teams increasingly shift into what is effectively system forensics: tracing lineage, reverse-engineering intent, and relying on ad-hoc warnings embedded in labels or comments. CIOs frustrated with multi-year modernisation projects see AI as a potential escape hatch, shrinking timelines from months to days. But without addressing the underlying complexity, AI simply moves faster in a maze that remains just as opaque, limiting both enterprise developer productivity and sustainable change velocity.
New Abstraction Layers: Linking Understanding, Planning, and Action
The findings suggest that the core problem is not planning itself but the lack of a unified path from understanding to action. Today, admins and developers juggle dashboards, documentation, and tribal knowledge to assemble the context they need. Each handoff—between tools, teams, and phases—adds friction and risk. To break the cycle, organizations need better abstraction layers that surface dependencies, usage patterns, and governance in one place. Automation must focus not only on generating new artifacts but on maintaining a live, coherent map of the Salesforce landscape. When understanding, planning, and execution are stitched into a single workflow, AI-agent interactions can safely assume more responsibility for actual change execution. Until then, AI risks being confined to investigative support while humans continue to shoulder the CRM system overhead of manual untangling and repetitive Salesforce context switching.