The AI Knowledge Package: How Commander Prepares SharePoint for AI Consumption

Every organization deploying Copilot or any AI against SharePoint is asking the same question: "Are we ready?" Most have no structured way to answer it.

Commander answers it with the AI Knowledge Package — a structured, machine-readable export of your SharePoint environment that both a SharePoint administrator and an LLM can consume. Here's what's actually in it.

The Package Structure

Every Commander AI Knowledge Package has three layers:

Manifest — package identity, scope (tier, object ID, object title), provider inventory, and version metadata. Tells you exactly what was evaluated and when.

Knowledge modules — raw harvested data from SharePoint. For each provider module (site features, permissions, lists, content types, fields, etc.), Commander extracts the live state of the environment: what's active, what's unknown, what's configured, and what's changed.

Assessment modules — scored evaluation of each knowledge module. Every assessment produces: a 0–100 score, a rating (Healthy / Review Recommended / Needs Attention), a risk level, enumerated strengths, enumerated findings with severity (critical/high/medium/low), and actionable recommendations.

Three Tiers of Coverage

Commander evaluates SharePoint at three tiers — each reveals what the tiers above cannot:

Tier	What Commander Evaluates
Site	Site features (104 evaluated in one test environment — 27 active, 77 unknown status), general configuration (storage, administrators, quotas), permissions (role assignments, groups, role definitions), recycle bin state, and web enumeration
Web	Lists and libraries (type, item count, permissions, activity), web-level features, general web properties, and web-scoped permissions
List	Content types (visibility, mutability, form mappings), fields (type, required, indexed), forms (inherited vs. overridden), list items, and list-scoped permissions

The Web tier flags that 19 of 19 lists need detailed analysis. The List tier reveals why — and scores each list individually.

How the Health/AI Readiness Report Works

When Commander assembles the package into a Health / AI Readiness Report, each module contributes:

1. A score — Site Features: 80/100. Site General: 90/100. Web Lists: 80/100. List Content Types (individual list): 100/100. These are not averages — each module scores independently based on its own rule engine.

2. Strengths — what passed validation. "Site Feature inventory was normalized — 104 features reconciled." "All 6 form mappings resolve." "Storage healthy — 15% of quota in use."

3. Findings — what needs attention, ranked by severity. "81 feature records only partially reconciled against reference definitions" (medium). "19 of 19 lists require detailed List-tier assessment" (high). "Only one Site Collection administrator — continuity dependency" (medium).

4. Recommendations — specific, actionable next steps tied to each finding. "Confirm a second accountable Site Collection administrator and periodically validate both accounts."

   
   

Why This Matters for AI Readiness

Before you point Copilot or any LLM at SharePoint, you need to know what it will see. Commander's package gives you that answer in a format both your security team and an AI can parse:

• Permissions sprawl — surfaced at every tier, not just the site level

• Unknown features — 77 of 104 features had uncertain deployment status in one test environment. Those features create governance blind spots that AI inherits

• Structural gaps — orphaned lists, single-administrator dependencies, stale content

• Configuration integrity — content types, form mappings, field configurations

The package is the prerequisite. The Health Report is the answer. Together they replace "I think we're ready" with "here's exactly what AI will see."