/ DESK 03
Applied Systems & Tools
Field notes
Each page earns its place through a distinct decision, credible evidence, and a practical next move.
Automation, Copilot, Workflow, or Agent? Choose the Simplest System That Fits
A practical decision guide for choosing rules automation, an AI copilot, a defined AI workflow, or an agent based on uncertainty, consequence, and control.
Read the field noteAI Readiness Is a Workflow Property, Not a Company Personality Test
Assess AI readiness around a named workflow, decision, and outcome across problem, process, data, technology, people, governance, and measurement.
Read the field noteThe dotSuper 0→1 Method: From a Real Constraint to the First Measurable System
dotSuper’s evidence-led method for moving from a real workflow constraint to the smallest useful intervention, a measurable result, and a defensible next decision.
Read the field noteThe RAG Readiness Checklist: Fix the Knowledge Operation Before the Chatbot
A pre-build assessment for source ownership, permissions, document quality, update paths, question coverage, evaluation, and accountable use.
Read the field noteHuman in the Loop Is Not a Control Until the Human Can Actually Intervene
A practical design guide for review authority, evidence, time, competence, escalation, override, and learning in human-AI workflows.
Read the field noteAI Agent vs Deterministic Automation: Use Agency Only Where It Earns Its Risk
A decision framework for choosing fixed rules, model-assisted steps, or bounded agents based on ambiguity, action space, reversibility, and consequence.
Read the field noteThe LLM Evaluation Scorecard: Test the Business Workflow, Not the Demo Prompt
A practical evaluation design for representative cases, groundedness, task success, safety, latency, cost, review burden, and release decisions.
Read the field notePrompt Injection Controls for RAG and AI Agents: Design for Compromise
A practical security playbook for separating instructions from data, constraining tools, validating outputs, protecting secrets, and containing failures.
Read the field noteAI Knowledge Base Architecture: Preserve the Source of Truth
A practical architecture for source systems, ingestion, provenance, permissions, retrieval, citations, updates, deletion, and human ownership.
Read the field noteHow to Pilot AI in a Document Workflow Without Automating the Wrong Decision
A bounded pilot blueprint for extraction, classification, comparison, drafting, review, exception handling, and measurable operational value.
Read the field noteThe AI Production Handoff Checklist: Leave Capability, Not Dependency
A practical handoff checklist for ownership, runbooks, evaluation, monitoring, access, suppliers, cost, change, incidents, training, and retirement.
Read the field noteThe AI Incident Response Runbook: Detect, Contain, Decide, Learn
A practical runbook for harmful outputs, data exposure, tool misuse, drift, cost spikes, service failure, and unreliable knowledge.
Read the field noteAI for Manufacturing Quality: Choose the First Use Case by Evidence and Consequence
A selection playbook for inspection, document review, non-conformance triage, root-cause support, and knowledge retrieval in quality operations.
Read the field note