Field notes
from running workflows.
What breaks in self-built AI workflows, how we stabilise and run them, and what we measure. No marketing fluff, no sponsored content.
All dispatches.
Routing Rules vs. AI Screening in Intake: Decision Guide (Without Tool Chaos)
When a rule set is enough and when AI in intake really makes sense - with decision matrix, risks, and minimal starter setup.
Topic Rotation for Law Firms: 12 Topic Families (System Instead of Idea List)
A topic system for law firms: 12 topic families that are reusable - including rotation plan and format matrix.
AI Content Labeling & Guardrails (AI Act): Pragmatic for Law Firms
Pragmatic guardrails for law firms: when labeling makes sense, what approval/QA looks like, and which no-go claims to avoid.
Avoid Tool Collection: 9 Rules for Workflow Systems in Law Firms
Nine rules that protect you from tool chaos: Ownership, status model, releases, data checks - so automation runs as a system.
Owner & Change Process: Governance for Automation (Law Firm)
A minimal governance setup: Owner, backup, change process, and release rules - so workflows don't fall apart in daily operations.
Documentation as a System: Templates That Delivery Teams Actually Use
Documentation that gets used: three templates (runbook, process, handoff) and a simple routine - instead of notes in chat.
Conflict Check/New Retainer: Minimal Process That Doesn't Block Intake
A minimal process for conflict check in intake: when to check, what data is enough, and how to build the flow so nothing blocks.
Owner & Role Model for Automation in Law Firms (RACI + Escalation)
A lean role model (owner, backup, escalation) so workflows in law firms don't fail due to "nobody is responsible" - including RACI to copy.
KPI Set for Automation in Law Firms: 5 Metrics + Measurement Setup
5 KPIs that actually drive decisions (instead of KPI sprawl) - including measurement plan, target value logic, and "stop" rules for law firm workflows.
Data Quality in Intake: Validation, Dedup, Edge Cases (Law Firm)
Pragmatic rules for intake data: Validation at the source, duplicate logic, and edge cases that otherwise destroy routing/follow-ups - including rule set to copy.