Frontline architect. Sharing knowledge and insights in enterprise architecture.
This article distinguishes business capabilities (what an enterprise can do) from business processes (how work gets done) using TOGAF 10's metamodel, showing capabilities as stable abstractions while processes operationalize them in variable workflows.
Gartner predicts 30% of generative AI projects will be abandoned post-PoC, rising to 50-60% without AI-ready data; this analysis reveals why PoCs succeed but production fails due to unclear business value, underestimated costs, data quality gaps, and poor business integration.
The author describes how using Codex for PPT creation revealed a shift from prompt-driven AI to loop-driven autonomous workflows, where agents leverage context, memory, and tools to continuously execute tasks, requiring clear goals, stable context, integrated tools, and verification mechanisms, redefining the engineer's role from operator to system designer.
The article explains that while Skills suffice for fixed tasks like document generation, ontologies become essential when AI must reason across complex business objects and relationships — such as organizational changes affecting roles, permissions, and processes — but cannot replace data governance, rule governance, or eliminate hallucinations entirely.
OpenClaw is a self-hosted AI assistant gateway that connects user entry points to LLMs, memory, and tools, enabling task execution beyond chat; its four-layer architecture processes messages through skills for real-world actions, while security requires trusted entry, least privilege, tool classification, execution isolation, and auditability.
The article explains how AI adoption is driving enterprises to rebuild semantic layers as shared business logic infrastructure, detailing what semantic layers are, the four problems they solve, industry endorsements from Thoughtworks, SAP, and Gartner, practical building approaches starting with consensus on core entities, and the boundaries where semantic layers excel versus where they fall short.
This article explores how AI-accelerated code generation creates hidden cognitive, verification, and architectural debts that erode team understanding, ultimately slowing integration, maintenance, and refactoring despite faster initial output, arguing true efficiency requires deliberate comprehension over raw generation speed.
This article details SAP's four-layer process architecture — Value Chain, Process Areas, End-to-End Processes, and Sub-Processes — explaining how each layer connects strategy to execution with definitions, characteristics, and examples like O2C and P2P.
The article argues that process hierarchy levels should be determined by the granularity of control needed for stable operations, not by a fixed structural model, and outlines three conditions for effective hierarchy: clear roles, distinct management purposes, and embedding in real operational systems.
This article argues that business processes are not merely cost-reduction tools but the operational pathways that transform resources into value, explaining why scaling enterprises must treat process transformation as a core business problem to drive revenue, cost efficiency, asset utilization, and digital readiness.
