Diagnostic · Step 3 of 3
Document Analysis
Alongside the diagnostic and the AI conversation, Sextant reviews a lightweight pack of existing company documents so findings rest on operating evidence, not self-report alone.
Document pack
A lightweight review of six document types grounds the diagnostic in operating evidence rather than opinion alone. Most companies already have these on hand; nothing is created for the review.
Representative Sextant output — fictional company and modeled assumptions.
Document Pack
Requested materials
A short list of documents that already exist inside the company — no new artifacts to produce, just what's on hand.
Requests fulfilled
5 of 6
Organizational chart & reporting linesOrg & Governance
Establishes who owns AI decisions today and where reporting lines create or close gaps between leadership intent and frontline execution.
Current org chartCommittee / working-group rostersRACI for technology decisions
Complete▸
Strategy & board materialsStrategy
Shows how AI investment is framed against the company's broader strategic priorities and what leadership has already committed to publicly or internally.
Most recent strategy deckBoard minutes referencing AI or automationAnnual operating plan
Complete▸
AI, data & governance policiesGovernance
Benchmarks written policy against the NIST AI RMF and surfaces where governance exists on paper but not in practice.
AI acceptable-use policyData classification & retention policyModel / vendor risk review process
Complete▸
Core process SOPsOperations
Identifies which high-volume workflows are documented well enough to redesign around AI-assisted execution versus those that live only in tribal knowledge.
Top 5-10 SOPs by volume or costProcess maps or swimlane diagrams
Received · in review▸
Systems & data inventoryTechnology
Maps the current systems and data landscape to identify integration points, silos, and the unification work required before agent deployment.
System / application inventoryData warehouse or lake architecture diagramIntegration & API inventory
Complete▸
Prior AI pilot post-mortemsValue & Measurement
Reveals what has already been tried, what was measured (or not), and where past pilots stalled — the fastest way to avoid repeating a known failure mode.
Pilot summary decksPost-deployment measurement reportsVendor pilot contracts or SOWs
Awaiting upload▸
Pipeline
What happens after upload
Each document moves through a fixed pipeline before anything from it can inform a report.
01
Secure intake
Documents are received through an access-logged secure intake channel with a per-file audit trail.
02
Structured extraction
Each document is parsed into structured text with page-level locators so every downstream claim can be traced back to its source.
03
Evidence candidates
Excerpt-grounded claims are proposed against the diagnostic question set and queued for human reviewer sign-off.
04
Cross-document synthesis
Approved evidence is rolled up across documents into the thematic findings the diagnostic team carries into Map.
Confidentiality
Documents are processed under a zero-data-retention agreement with the model provider — only derived text and page locators are used for analysis, never the raw files themselves. Every access is logged, and a human reviewer approves each document before any excerpt from it becomes citable.
Human Review Gate
Every excerpt is reviewed before it counts
The model proposes excerpt-grounded evidence candidates. An Areté Intelligence reviewer approves or rejects each one before it can appear in any report — nothing the model proposes reaches a report unreviewed.
148
Proposed
121
Approved
19
Rejected
Cross-Document Synthesis
What the documents say, together
Approved excerpts are synthesized into a small set of grounded findings, each traceable back to its source document.
Governance posture
Written AI governance exists but coverage is uneven — strong on data classification, thin on model and vendor risk review.
  • Acceptable-use policy was last revised within the current fiscal year and references NIST AI RMF categories.AI Use Policy v2 · p. 4
  • No standing process was found for reviewing third-party AI vendor risk prior to procurement.Data Classification Standard · p. 7
  • Data retention rules are documented but do not yet address model training or fine-tuning data.AI Use Policy v2 · p. 9
Organizational shape
Decision rights for AI initiatives sit with a steering committee, but day-to-day ownership is diffuse across functions.
  • An AI steering committee meets quarterly and reports through the CFO, not a dedicated AI or data leader.AI Steering Committee Roster · p. 1
  • No function was found with explicit budget authority for cross-functional AI initiatives.Org Chart 2026 · p. 2
  • Reporting lines route technology decisions through operations rather than a central technology function.Org Chart 2026 · p. 3
Strategic priorities
Leadership has named AI as a strategic priority publicly, but the operating plan does not yet translate that into funded initiatives.
  • The FY26 strategy deck lists AI adoption as a top-three priority for the year.FY26 Strategy Offsite Deck · p. 6
  • The annual operating plan does not allocate a discrete budget line for AI initiatives.FY26 Strategy Offsite Deck · p. 31
  • Board minutes reference AI twice in the last four quarters, both times as a standing agenda item without a resolution.Q2 Board Minutes Excerpt · p. 2
Notable gaps
The clearest gap is measurement: initiatives are tracked for delivery, not for the value they were meant to create.
  • Two of the five reviewed process SOPs have no defined success metric beyond completion.Core SOP Bundle · SOP 3
  • No prior AI pilot in the reviewed set has a documented post-deployment measurement report.Systems Inventory 2026 · tab 2
  • Data architecture diagram shows three unresolved integration points flagged 'pending' for over a year.Data Architecture Diagram · p. 1
Document evidence supports the narrative and citation layer of the Cohort Report and Command Center. It does not by itself change the numeric readiness scores — those come from the diagnostic and interviews.
See the Cohort Report →