Records practices, thoughts, and pitfalls on the data grunt-work journey, sharing content on data platforms, data analysis, data processing, data governance, knowledge graphs, and more. Less theory, more hands‑on, making complex data technologies simple.
The article argues that AI programming efficiency should be measured by task acceptance, rework, and defects—not code volume or tokens—and proposes a practical framework tracking human interventions, shared improvements, and maintenance costs to assess true team productivity gains.
This article introduces Mission Engineering through a coffee-making analogy, emphasizing that completing steps doesn't guarantee mission success; it outlines a structured approach using a Mission Description Card to define goals, stakeholders, constraints, and success criteria, and advocates iterative experimentation over rigid processes.
The article presents a systematic framework for converting AI agent errors into reusable team capabilities by tracing root causes across requirements, design, context, implementation, and testing, then codifying fixes as templates, executable test cases, and maintained tools with defined scope and validation steps.
The article argues that suitability for ontology intelligence pilots depends on specific readiness conditions—concrete tasks, accessible data, verifiable results, clear authority for rule confirmation, and maintenance plans—not company size, illustrating with a comparison of two enterprises and providing a six-item checklist for project preparation.
The article argues that individual AI proficiency does not guarantee team delivery sustainability, advocating for clear role responsibilities, handover exercises to validate shared capabilities, hiring for AI usage plus engineering judgment and collaboration, and risk-based automation decisions to build organizational capability in the Agent era.
This article presents a rigorous framework for evaluating ontology project value through concrete business task improvements—such as reduced verification time, misrouting, and rework—rather than graph metrics, using a broadband cancellation case study to demonstrate marginal contribution analysis, end-to-end acceptance criteria, TCO/ROI calculation, and sunk-cost-aware investment decisions.
As large language models improve, enterprises must distinguish between obsolete manual ontology tasks and enduring business semantics—definitions, rules, mappings, evidence, and test cases—that should be decoupled from platforms and validated through disengagement drills to avoid vendor lock-in and ensure portable, verifiable knowledge assets.
The article explains that AI Agents can analyze manufacturing anomalies and create work orders, but real resolution requires distinct ownership for containment, diagnosis, disposal, and quality closure, with clear cross-shift handover and verification criteria, not just automated analysis.
Even with accurate RAG retrieval, business logic errors persist due to field mapping mismatches, missing facts, batch rules, and implementation flaws; ontologies unify reusable definitions but cannot replace data, computation, or permissions, so teams should first diagnose the exact failure point before investing in ontology modeling.
This article analyzes why employees abandon enterprise AI tools after launch, presenting a three-level adoption framework and a six-barrier diagnostic table, then details practical steps—embedding AI in workflows, building verifiable trust, clarifying accountability, training for independent judgment, and tracking a metrics scorecard—to iteratively validate real productivity gains before scaling.
