Deloitte's global chief architect argues enterprises must shift to 'humans as governors, agents as operators' or lose ground to competitors redesigning operating models.
Scott Rodgers, global chief architect and U.S. CTO of Deloitte's Microsoft Technology Practice, outlined the requirements for agent-first enterprise transformation in a sponsored piece by MIT Technology Review Insights. He argues that legacy processes aren't built for autonomous AI systems and require machine-readable definitions, explicit policy constraints, and structured data flows. With AI technology budgets expected to grow over 70% in two years, the piece warns that companies still running agent pilots risk structural disadvantage versus competitors who commit to full operating model redesign.
This piece is framework-level strategy, not technical instruction. The actionable kernel for developers: enterprise AI agents require machine-readable process definitions and structured data flows — meaning your team needs to instrument business logic in ways legacy systems never demanded. If you're building agent pipelines, the real gap isn't the model — it's that upstream business processes produce unstructured, ambiguous inputs that agents can't reliably act on.
Audit one high-volume internal workflow your team supports: map every input and output, then flag which ones are unstructured strings vs. schema-defined data. If more than 40% are unstructured, that's your refactoring target before any agent layer.
Go to claude.ai and open a new conversation
Tags
Also today
Signals by role
Also today
Tools mentioned