Excel First Governance

Excel-First Governance

The Excel-first governance model brings Coherence Ops to teams who already live in Excel and SharePoint. Instead of CLI commands, JSON artifacts, and Python scripts, governance happens through named tables, structured prompts, and write-back rows — all inside a shared workbook.

---

BOOT Protocol

Cell BOOT!A1 acts as a system prompt. When the workbook is attached to an LLM, this cell initializes the entire governance context.

Structure:

YOU ARE: Creative Ops Copilot (Coherence Ops: Truth · Reasoning · Memory)
OPENING MOVE: Read tables, then ask "What Would You Like To Do Today?"
MENU:
  1. Build Claims from Deliverables
  2. Refresh Assumptions
  3. Detect Drift
  4. Propose Patch Options
  5. Canon Audit
  6. Exec Summary
  7. Asset Checklist
RULES: Use ONLY workbook data. Cite TableName + Row_IDs. Enforce Canon Guardrails.
OUTPUT: Findings → Recommended Actions → Write-back rows for Excel

Why it works: LLMs parse structured text in A1 as system instructions. Named tables give deterministic cell references. The menu constrains the interaction to governed operations.

Full spec: docs/excel-first/WORKBOOK_BOOT_PROTOCOL.md

---

Table Schemas

Seven named Excel tables map directly to Coherence Ops primitives:

tblTimeline

Milestone tracking with drift and patch status.

Column	Type	Description
Milestone_ID	Text	Unique identifier
Milestone_Name	Text	Short descriptive name
Owner	Text	Responsible person
Due_Date	Date	Target completion date
Status	Text	On Track / At Risk / Blocked / Complete
Drift_Flag	Boolean	TRUE if drift detected
Patch_Ref	Text	Link to tblPatchLog row if patched

tblDeliverables

Asset-level deliverable tracking.

Column	Type	Description
Deliverable_ID	Text	Unique identifier
Deliverable_Name	Text	Asset or output name
Milestone_ID	Text	FK → tblTimeline
Format	Text	File type or medium
Status	Text	Draft / In Review / Approved / Delivered
Assignee	Text	Responsible person
Due_Date	Date	Target delivery date

tblDLR

Decision Ledger Records — formal decision capture.

Column	Type	Description
DLR_ID	Text	Unique identifier
Decision_Title	Text	Short decision name
Decision_Date	Date	When the decision was made
Decision_Owner	Text	Who made the decision
Rationale	Text	Why this choice was made
Evidence_Refs	Text	Supporting data or sources
Status	Text	Active / Superseded / Revoked

tblClaims

Verifiable claims tied to decisions.

Column	Type	Description
Claim_ID	Text	Unique identifier
Claim_Text	Text	The verifiable assertion
DLR_Ref	Text	FK → tblDLR
Confidence	Number	0.0–1.0 confidence score
Evidence	Text	Supporting evidence
Status	Text	Active / Challenged / Refuted
Last_Verified	Date	Most recent verification date

tblAssumptions

Assumptions with confidence scores and time-to-live.

Column	Type	Description
Assumption_ID	Text	Unique identifier
Assumption_Text	Text	The assumption statement
Owner	Text	Who holds this assumption
Confidence	Number	0.0–1.0 confidence score
TTL_Days	Number	Days before review required
Last_Reviewed	Date	Most recent review date
Status	Text	Active / Expired / Replaced

tblPatchLog

Drift → Patch correction log.

Column	Type	Description
Patch_ID	Text	Unique identifier
Drift_Source	Text	What triggered the drift
Drift_Type	Text	time / freshness / bypass / verify / outcome
Severity	Text	low / medium / high / critical
Patch_Action	Text	Corrective action taken
Applied_Date	Date	When the patch was applied
Applied_By	Text	Who applied the patch

tblCanonGuardrails

Blessed rules and constraints — the workbook's constitution.

Column	Type	Description
Canon_ID	Text	Unique identifier
Rule_Text	Text	The guardrail statement
Category	Text	Brand / Legal / Process / Quality
Severity	Text	Must / Should / May
Source	Text	Where the rule comes from
Last_Updated	Date	Most recent update
Status	Text	Active / Deprecated

Full column specs: docs/excel-first/TABLE_SCHEMAS.md

---

Writeback Contract

Every LLM response must include:

1. Findings — what was discovered (citing TableName + Row_IDs)
2. Recommended Actions — what should change
3. Write-back rows — paste-ready rows for Excel tables, specifying target table and columns

This ensures every LLM output is traceable and actionable within the workbook.

---

6-Lens Prompt Model

Six perspectives for multi-dimensional governance:

Lens	Focus	Example Question
PRIME	Constitutional compliance	"Does this decision violate any Canon Guardrails?"
EXEC	Strategic alignment	"How does this impact the campaign timeline?"
OPS	Operational readiness	"Are all deliverables on track for the milestone?"
AI-TECH	Technical feasibility	"Can the asset pipeline handle this volume?"
HUMAN	Team and stakeholder impact	"Who needs to approve this change?"
ICON	Brand and creative integrity	"Does this maintain brand consistency?"

Each lens crosses with three operations:

Operation	What it does
IntelOps	Gather intelligence — scan tables for signals
ReOps	Re-evaluate — challenge assumptions and claims
FranOps	Franchise governance — enforce canon and brand rules

6 lenses × 3 operations = 18 prompt patterns — all available from the BOOT menu.

Full guide: docs/excel-first/multi-dim-prompting-for-teams/README.md

---

Mapping to Coherence Ops

Excel Table	Coherence Ops Primitive	Artifact
tblTimeline	Decision Scaffold	DS
tblDeliverables	Decision Scaffold	DS
tblDLR	Decision Ledger Record	DLR
tblClaims	Atomic Claims	Claim
tblAssumptions	Reasoning Scaffold	RS
tblPatchLog	Patch Packets	Patch
tblCanonGuardrails	Canon	Canon

The same Truth · Reasoning · Memory loop runs in Excel as in the Python runtime — just with a human-readable interface.

---

See Also

• Creative Director Suite — Dataset, workbook, quickstart
• Coherence Ops Mapping — DLR / RS / DS / MG governance artifacts
• Unified Atomic Claims — Claim primitive specification
• Canon — Blessed claim memory and canon entry lifecycle