We just published a new paper outlining Pega’s unique approach to using AI to reimagine and run the most critical work in an enterprise.
Check it out here: The Architecture of the Autonomous Enterprise | Pega
Thanks for sharing this article @Don_Schuerman
What stands out for me in this article is the clear emphasis on architecture and control, not just agent capability.
For those who haven’t had the time/chance to read, let me summarize the key architecture points which could serve as food for thought for agentic system architecture principles:
-
Separate design time from run time: AI-driven creativity and re‑imagination happen at design time, while execution at run time remains deterministic and governed to avoid production risk
-
Orchestrate agents inside workflows: Agentic AI is embedded within structured business processes, not allowed to operate as unmanaged runtime behavior, ensuring predictability and control
-
Use a single execution and governance layer: Decisions, workflows, AI models, and integrations are coordinated through one control plane to manage complexity across systems and clouds
-
Favor model‑driven change over custom code: Low‑code, configuration‑based architecture allows continuous change without accumulating technical debt, which is essential for scaling AI safely
-
Design for continuous evolution: The architecture assumes ongoing adjustment of decisions and processes, rather than one‑off AI deployments, so autonomy increases through iteration, not big‑bang releases
I’d be interested to hear how others are mapping these architectural principles into concrete reference architectures or COE standards within their organizations? Where do you see or witness tension and challenges?