Enterprise AI adoption is stalling not because of model limitations, but because the underlying workflows were never designed for autonomous agents. Failed handoffs, undefined steps, and implicit human decisions are exposing a fundamental mismatch between modern AI systems and legacy processes. Salesforce is addressing this with Agentforce Operations, a new control plane that restructures workflows for deterministic agent execution.
Salesforce’s solution comes as organizations increasingly deploy AI agents to handle complex back-office tasks. However, many of these agents fail not due to algorithmic shortcomings, but because their operating environments lack the structured processes agents require. Sanjna Parulekar, Salesforce’s senior vice president of Product, emphasized this challenge in an interview with VentureBeat. “We’ve seen customers upload workflows that break in production because they were designed around human assumptions rather than machine execution,” she explained. “Our platform identifies these gaps and reengineers processes to make them agent-compatible.”
Turning chaotic workflows into agent-ready tasks
Agentforce Operations introduces a structured approach to workflow redesign. Enterprises can either upload existing processes or select from pre-built Salesforce blueprints that outline agent-ready task sequences. The platform then decomposes these workflows into discrete, executable steps that agents can follow without ambiguity.
This structural clarity solves a persistent problem in enterprise AI deployment: processes that rely on implicit knowledge or human intuition. “When agents encounter loosely defined steps or undocumented decision points, they either fail or require excessive human intervention,” Parulekar noted. “By forcing explicit task definitions, we make the process deterministic and observable.”
The system also incorporates session tracing, providing real-time visibility into agent workflows. This observability layer allows teams to monitor task progression, identify bottlenecks, and validate outcomes without relying solely on post hoc analysis.
Why deterministic workflows matter for AI agents
Most traditional automation tools rely on probabilistic routing—sending tasks to the next available agent or system based on incomplete context. Agentforce Operations takes a different approach: it enforces a pre-defined execution path where agents follow a clear sequence of steps. This deterministic structure eliminates the uncertainty that often derails AI deployments.
The distinction is critical for enterprises scaling AI across departments. Without such control, organizations risk deploying agents that increase operational costs by repeating failed tasks or requiring constant human intervention. “The real expense isn’t the AI model—it’s the overhead of fixing broken workflows at scale,” Parulekar observed.
This approach also shifts the governance burden from reactive troubleshooting to proactive process design. Teams must now ask: Which steps are essential? Where can automation reduce variability? And how do we ensure these workflows remain adaptable as business needs evolve?
The hidden risks of codifying flawed processes
While Agentforce Operations provides the technical scaffolding for agent-driven workflows, it doesn’t automatically fix underlying process issues. If enterprises encode broken workflows into a system, those flaws become entrenched at scale. The challenge then shifts from execution to governance: Who owns the process? How do we validate changes? And how do we ensure the workflow remains aligned with business objectives?
Brandon Metcalf, founder and CEO of workforce orchestration platform Asymbl, highlighted this shift in a separate interview with VentureBeat. “The bottleneck isn’t whether an agent can reason through a task—it’s whether the workflow beneath it is coherent enough to support execution,” he said. “Without a shared understanding of the end goal, neither humans nor agents can deliver consistent results.”
Metcalf emphasized that success requires clear ownership of outcomes, whether by a human manager or a designated agent supervisor. This governance layer becomes even more critical as enterprises distribute workflows across hybrid teams of humans and AI systems.
A preview of AI’s next frontier in enterprise operations
Salesforce’s Agentforce Operations signals a broader industry trend: the rise of workflow execution control planes as a prerequisite for scalable AI adoption. As organizations push agents deeper into core operations—from customer service to supply chain management—the need for structured, observable processes will only intensify.
The platform also reflects a maturing understanding of AI’s limitations in enterprise contexts. “We’re moving past the ‘AI will solve everything’ phase,” Parulekar noted. “The real opportunity lies in pairing capable models with well-designed workflows that set them up for success.”
For enterprises still grappling with broken processes, this shift offers both a wake-up call and a path forward. The question isn’t whether AI can transform workflows—it’s whether organizations are ready to build the infrastructure that will make that transformation possible.
AI summary
Şirketler, AI sistemlerini daha efektif kullanabilmeleri için workflow yönetimine ihtiyaç duyuyor. Salesforce'un Agentforce Operations adlı yeni workflow platformu, bu soruna çözüm getiriyor.
