Enterprise AI can automate tasks and speed up decisions, but without understanding a business’s operations, it risks missing critical constraints and relationships. As companies aim to integrate AI into daily operations, the challenge is providing it with the context needed to function effectively within complex processes.
Devavrat Shah, chief scientist of enterprise AI at Celonis, emphasizes the importance of what he calls “decision intelligence.” AI agents, he argues, must understand their role within larger decision-making processes and the surrounding business conditions. Without this context, they may overlook constraints, make costly errors, or waste computing resources.
Connecting AI to existing processes
The solution involves linking AI to the processes and data already in use within enterprises. A contextual layer can help agents understand workflows, operate within defined boundaries, and generate recommendations for human review and approval.
Shah advises organizations to start by focusing on a specific business decision, identifying its most challenging steps, and testing whether AI can improve them. Measuring improvements in speed and value can then guide decisions about scaling the application.
This approach could also transform how employees allocate their time. By automating repetitive, manual tasks and coordinating work across teams, AI could free people to focus on more complex, strategic decisions. However, realizing this benefit will require investing in workforce education, so employees can understand AI outputs and their implications rather than simply handing decisions over to agents.
Looking ahead, Shah sees opportunities for AI to move beyond routine operational tasks to support enterprise-wide planning in areas like finance, workforce management, marketing, and sales. With the right context, AI could help organizations evaluate options and make critical decisions more quickly.
As Shah notes, “Now, maybe [resource planning] should not take that much time, and people are actually making the right choices quickly.”
The role of context in AI decision-making
Without operational context, AI agents may apply general principles to specific scenarios, leading to suboptimal results. Shah emphasizes the need for “decision intelligence”—AI that understands its role within larger decision-making processes and the surrounding business conditions.
To address this, Celonis focuses on process intelligence, helping organizations adhere to ideal processes, build better ones, and enable people to work with multiple processes in an intertwined manner. By digitizing processes and understanding where issues arise, organizations can re-engineer their operations for better outcomes.
Implementing the context layer
The context layer sits between the system of record layer (workflow tools and event logs) and the agentic layer (AI agents and human experts). Celonis’s context model is designed to:
- Abstract the system of record layer to identify the right process pieces for agentification.
- Bring role-based AI agents into the process in a structured, guardrailed manner.
- Enable agent-first interactions, allowing existing agentic infrastructure to communicate with the context layer.
While the system of record layer often exists within organizations, implementing the context layer requires careful consideration of information security.
The future of work with AI
Shah envisions three types of AI use cases:
- Automating repetitive, labor-intensive tasks.
- Improving coordination across teams, reducing time spent on emails and spreadsheets.
- Providing strategic support to human decision-makers, enabling quicker and more informed choices.
In areas like financial planning, workforce management, and marketing, AI could significantly reduce the time spent on annual planning processes, allowing organizations to make critical decisions more rapidly.
This episode of Business Lab is produced in partnership with Celonis.
