BestHub
Sign in
Data Bricklaying Diary
Author

Data Bricklaying Diary

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.

165
Articles
0
Likes
383
Views
0
Comments
Recent Articles

Latest from Data Bricklaying Diary

100 recent articles max
Data Bricklaying Diary
Data Bricklaying Diary
Oct 19, 2026 · Operations

Before Your AI Agent Troubleshoots, Make It Show the Evidence

This article outlines a rigorous framework for AI-assisted incident troubleshooting: verify alerts before trusting agents, gather structured evidence from tasks and call chains, treat root causes as testable hypotheses, separate fixed queries from model analysis, enforce read-only budgets, require authorized runbooks for remediation, and measure effectiveness by diagnostic accuracy and recovery time rather than suggestion adoption rates.

AI troubleshootingauthorization controlseffectiveness evaluation
0 likes · 16 min read
Before Your AI Agent Troubleshoots, Make It Show the Evidence
Data Bricklaying Diary
Data Bricklaying Diary
Oct 19, 2026 · R&D Management

Who Confirms What in Your Ontology Pilot? A Responsibility Matrix That Tracks Expert Time and Handover Readiness

The article details a responsibility assignment framework for ontology pilot projects, using an after-sales service example to show how to map deliverables to specific preparers and approvers, break expert involvement into measurable tasks, distinguish processing time from lead time, and validate handover readiness through a simulated policy-change drill.

RACIbusiness rule confirmationexpert time tracking
0 likes · 17 min read
Who Confirms What in Your Ontology Pilot? A Responsibility Matrix That Tracks Expert Time and Handover Readiness
Data Bricklaying Diary
Data Bricklaying Diary
Oct 17, 2026 · Artificial Intelligence

Agent Calls Succeed But Answers Fail: Tracing Observability to Quality Evaluation

The article explains how successful Agent execution traces can still produce wrong answers, using a procurement case where an Agent cited an outdated contract attachment. It details a systematic approach: trace errors through observability across indexing, retrieval, context assembly, and generation; convert feedback into reproducible test samples with versioned inputs and expected behaviors; evaluate using rule-based checks, model-assisted judging, and human review; separate critical error categories instead of relying on a single score; maintain regression suites with both failure and success cases; and rerun tests in isolated environments while recording versions to distinguish new results from original incidents.

agent evaluationerror tracingmodel-assisted judging
0 likes · 15 min read
Agent Calls Succeed But Answers Fail: Tracing Observability to Quality Evaluation
Data Bricklaying Diary
Data Bricklaying Diary
Oct 17, 2026 · Artificial Intelligence

Why AI Demos Succeed But Production Fails: The Business Capability Contract Gap

The article explains why AI models that impress in demos often fail in real business processes, introducing the Business Capability Contract that defines service objects, availability timing, output nature, evidence sources, responsibility roles, and closed-loop acceptance to bridge the gap between model outputs and executable business actions.

AI acceptance criteriaAI business integrationAI governance
0 likes · 21 min read
Why AI Demos Succeed But Production Fails: The Business Capability Contract Gap
Data Bricklaying Diary
Data Bricklaying Diary
Oct 16, 2026 · Operations

What's Missing in Agent Observability with Jaeger, Prometheus, Graylog

The article details how to extend existing observability stacks for AI agents by adding task-centric correlations, verifying signal routing to Jaeger, Prometheus, and Graylog, building query entry points from tasks, protecting sensitive data, ensuring collector resilience, and validating the integration with a test procurement task.

Agent ObservabilityCollector ResilienceGraylog
0 likes · 16 min read
What's Missing in Agent Observability with Jaeger, Prometheus, Graylog
Data Bricklaying Diary
Data Bricklaying Diary
Oct 15, 2026 · R&D Management

Why Your AI ROI Calculation Is Wrong: Three Value Paths You Must Separate

This article explains why AI project ROI calculations often fail, proposing a framework that separates value into process efficiency, knowledge enhancement, and product innovation, distinguishes realized cash benefits from released capacity and future opportunities, requires comparable baselines covering failures, includes full lifecycle costs, and uses a value evidence card for continuous re-evaluation.

AI ROIBaseline SettingContinuous Re-evaluation
0 likes · 28 min read
Why Your AI ROI Calculation Is Wrong: Three Value Paths You Must Separate
Data Bricklaying Diary
Data Bricklaying Diary
Oct 14, 2026 · Backend Development

One Task, Multiple Traces: Connecting Agent Sessions, Tools & Async Execution

This article explains how to correlate multiple distributed traces for a single long-running agent task by distinguishing task, session, execution, action, attempt, and trace identifiers, propagating context across queues and pauses, instrumenting at the right layers, and handling missing trace segments without losing business continuity.

Agent ObservabilityLong-running TasksOpenTelemetry
0 likes · 15 min read
One Task, Multiple Traces: Connecting Agent Sessions, Tools & Async Execution
Data Bricklaying Diary
Data Bricklaying Diary
Oct 14, 2026 · R&D Management

FDE #07: Why Launch ≠ Value — Production Gates & Metrics That Prevent False Success

This article distinguishes production readiness gates from value validation metrics in AI agent deployments, detailing four hard gates (quality, security, operations, governance), a six-dimensional metrics framework with seven required fields per metric, attribution boundaries for business results, and decision triggers (continue, expand, contract, exit) to avoid false success signals.

AI agent deploymentFDEattribution analysis
0 likes · 27 min read
FDE #07: Why Launch ≠ Value — Production Gates & Metrics That Prevent False Success
Data Bricklaying Diary
Data Bricklaying Diary
Oct 13, 2026 · Backend Development

Why Your Agent Stuck at 'Processing': Designing Observable Task Status & Business Results

This article explains how to design observable Agent task statuses by distinguishing tasks, actions, and sessions; tracking execution phases, wait reasons, and action results; separating activity from real progress; and building task observation cards with evidence links for debugging and human handoff.

AgentBusiness ResultsHuman Handoff
0 likes · 16 min read
Why Your Agent Stuck at 'Processing': Designing Observable Task Status & Business Results