Paul Miles — CLM Glossary Squad Operating Procedures (docx/md + Excel task checklist)
Tag: S-2026-06-12-paul-glossary-operating-procedures
Type: own-writing
Author(s): Paul Miles (Red Strata), drafted by AI assistant on his instruction
Date of source: 2026-06-12
Date ingested: 2026-06-12
Authority weight: own-writing.
Raw file: S-2026-06-12-paul-glossary-operating-procedures.md (markdown master); Word/Excel versions on disk at C:\Users\pgmil\OneDrive\Documents\_ClaudeWorkspace\09 Taxonomy Tool\Glossary Build Kit\guidelines\ (Squad_Operating_Procedures.docx, Squad_Task_Checklist.xlsx + .md copies, and Squad_Operating_Procedures_Approach.md work-breakdown record).
What it claims
An operating-procedure layer (v1.0) supplementing the Squad Operating Guidelines v1.2 and the CLM Glossary process diagram for the glossary squad. It breaks each delivery-flow step into sequenced, role-attributed tasks — 59 tasks across ten stage groups: Sprint-0 set-up (secure source directory, template control, T-05 population/validation, AI tooling, ~50-term calibration), Stage 1 Harvest, Stage 2 Cluster & de-duplicate, Stage 3a Unique terms, Stage 3b Conflicts, Stage 4 SME workshop, Stage 5 Master definition + QA gate, Stage 6 Publish, Stage 7 Govern & change control, and Stage 8 Incremental batches (GR-14 / Business Concept Key flow).
Fourteen tasks are AI-executed (prompts P-01–P-07 with their paired skill files and templates). For each, the document supplies a manual fallback procedure (M-01–M-07) mirroring the paired skill file’s Method section performed by a person — same templates, IDs, controls and review rules, with “Manual” recorded in method/traceability fields — for use when an AI agent is not available or approved. Stages 4 and 6 are human-only by design (no AI processing, no fallback needed). The document restates the binding working rules (skill+guardrails attachment, immutable originals, human-only columns, proportionate review, single-editor file discipline, confidence-governs-routing).
The companion Excel checklist lists the same tasks one row per task (Step, Seq, Task ID, Task, Responsible, AI-assisted, Prompt/Skill/Template refs, Manual fallback ref, Status dropdown, Owner, Date, Notes) for sprint sequencing and status tracking.
Notable quotes
- “AI accelerates, humans decide” and “nothing is invisible” (Section 1 — the two governing principles, carried from the Guidelines).
- “The templates, controls, IDs, statuses and review rules do not change — a person performs the processing” (Section 12, on manual fallbacks).
What’s speculative vs. asserted
- Asserted: the task breakdown, role attributions, AI execution points and manual fallbacks — all derived directly from the kit’s Guidelines, process diagram, prompts P-01–P-07, skill files and template Field Guides; nothing invented.
- Speculative / planning: the note that manual operation “materially reduces weekly term capacity” is a planning consequence of the Guidelines’ sizing assumptions, not a measured figure — the document says to re-baseline with the Squad Lead rather than quoting numbers.
Topics this feeds
- CLM Glossary Acceleration Squad — project page; this is the squad’s task-level operating layer.
Open questions raised
- Whether the squad adopts the Excel checklist as-is or imports it into the existing Jira-import backlog tracker.
- Manual-mode throughput is unquantified; if the AI agent is unavailable for a sustained period the sprint plan needs re-baselining.