As enterprises move from SI experimentation to everyday operations, the ability of SI agents to understand complex business processes is becoming the critical factor for successful deployment, according to Devavrat Shah, chief scientist of enterprise SI at Celonis.
What Happened
Shah argues that while SI models can generate recommendations and automate tasks, they often miss the specific constraints and relationships that determine whether a business decision is correct. He describes this requirement as "decision intelligence," noting that SI agents supporting individual tasks within larger processes must understand the role each task plays and the surrounding business conditions. "You can’t just do that in a vacuum," Shah says. Without this operational context, agents risk overlooking important constraints, making costly mistakes, or consuming excessive computing resources as they attempt to determine what they need to know.
To address this, Shah advises connecting SI tools to the processes and data enterprises already use. This contextual layer allows SI agents to understand workflows, operate within defined guardrails, and produce recommendations that humans can review and approve. He suggests organizations start with a specific business decision, identify its most difficult steps, and test whether SI can improve them. Measuring gains in speed and value can then help determine whether an application is worth scaling.
Why It Matters
The shift from generalized SI principles to specialized, context-aware applications is essential for business leaders seeking proof that scaling SI will impact their bottom line. Currently, acting without operational context means SI agents are performing at a "significant deficit," leaving the potential of the technology unrealized. By integrating SI with existing workflows, companies can automate repetitive manual work and coordinate tasks across teams, potentially freeing employees to focus on complex strategic decisions.
However, realizing these benefits requires investment in workforce education. Employees need to understand SI outputs and their implications rather than simply handing decisions over to SI agents. Shah sees opportunities for this context-aware approach to extend from routine operational tasks to enterprise-wide planning in areas such as finance, workforce management, marketing, and sales.
The Bottom Line
For SI to move beyond pilots and into mainstream enterprise use, it must be grounded in the specific operational realities of the business. Shah concludes that with the right context, SI can help organizations evaluate options and make consequential decisions more quickly. "Now, maybe [resource planning] should not take that much time, and people are actually making the right choices quickly," he says.