en.Wedoany.com Reported - Enterprise AI teams are hitting bottlenecks because underlying workflows were not designed for agents. Salesforce has launched a new platform, Agentforce Operations, which transforms backend workflows into tasks that agents can complete. Users can upload their own processes or use blueprints provided by Salesforce, and the platform will break them down into specific tasks for agents.

Sanjna Parulekar, Senior Vice President of Product at Salesforce, stated, that many enterprise workflows were not built for agents. "What we observe with customers is that failures in processes often exist within the product requirements document; they don't work when uploaded. We can optimize and trim content, replacing it with agents." The platform provides a deterministic structure for workflows, reducing failures caused by loose steps or implicit decisions.

Parulekar also mentioned that focusing on how a process operates and breaking it down into clear steps can make the system more deterministic. Agents know their tasks upon joining, and the system allows for human checkpoints to enhance transparency. This approach differs from traditional automation tools; it is not the agent that decides the next step, but rather the system that enforces a predefined structure.

However, encoding a workflow does not fix a flawed process. If the steps are wrong, coding merely solidifies the problem at scale. Once a workflow is distributed across multiple agents, governance challenges arise. Brandon Metcalf, CEO of workforce orchestration company Asymbl, pointed out that the key for humans and agents following a workflow is a shared goal. "You must understand the goal, otherwise you cannot successfully complete the task. Someone must manage the deliverables, whether a human or an agent."

The bottleneck has shifted: the question is no longer whether agents can reason through tasks, but whether the underlying workflow is coherent enough. For enterprises that have built processes around human judgment, this is harder to fix than switching to a smarter model.

This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com