AI in Feith · Categorization

Categorization, with the reasoning shown.

Record Insight reads each record and categorizes it against your agency's Records Control Schedule and the NARA General Records Schedule. Every decision arrives with a confidence score, a written reasoning, and the audit-trail entry that makes it defensible.

Request a walkthrough All six capabilities ›

5 Label layers, one passSensitivity · disposition · domain · ontology · risk

0 Training on your dataIt reads your records; it never learns from them

0 Public-internet lookupsMatched against your own RCS, inside your boundary

1 Audit entry per decisionConfidence score and written reasoning attached

The problem

You can't manage a record you can't categorize.

Most of an agency's records arrive unstructured and barely described, scattered across OneDrive, SharePoint, network drives, and shared inboxes, with file names and metadata that say little about what's inside. Until each one is matched to a retention schedule, it can't be retained, disposed, or defended.

Rules need structure that isn't there

Rule-based categorization keys off fixed fields, templates, and folder paths. Unstructured records simply don't have them.

Metadata is missing or misleading

A file named “final_v3.docx” in a personal drive tells you nothing about which records schedule should govern it.

Location doesn't equal content

Where a record happens to live says nothing about what it actually is, or how long it must be kept.

To categorize these records, something has to read and understand each one, then find the right schedule to assign it. That's a job for AI, not rules.

What it does

AI that reads records the way a person would.

Built into RMA iQ (Feith's full-lifecycle records platform), Record Insight is the next generation of auto-categorization. It understands a record by its content, metadata, and purpose, assigns it to the right NARA GRS or RCS category, and preserves it through to final disposition.

Reaches the records traditional auto-categorization can't: the unstructured content sitting in drives and shared inboxes with no usable metadata.
Adds to your existing categorizers rather than replacing them: it sits alongside the rules-, metadata-, and location-based ones you already run.
Flags weak schedule descriptions before they cause misfiles: it scores each category description against a structured set of readiness questions and points out what needs clarifying.

Clear schedule descriptions categorize better than vague ones: the system rewards specificity.

Worked example

Unstructured record

Field Assistance Bulletin 2024-1

U.S. Dept. of Labor · pulled from OneDrive with no usable metadata.

Record Insight reads → matches → decides

Reads the full record

Content, metadata & context: the whole record.

Matches your schedule NARA GRS + DoL RCS

Retrieves the matching schedule from your own RCS first, then grounds the decision in it.

Decision

DoL RCS: Interpretive guidance conf 0.94

“Interpretive policy guidance issued to WHD investigators → DoL RCS: Interpretive guidance.”

reasoning attached audit entry written

How it works

Categorized, justified, and dispositioned, grounded in your own schedule.

Record Insight's Categorization Agent retrieves the matching schedule first, then asks the model to justify the fit, so every decision is anchored to your own rules and the schedule it just retrieved.

RMA iQ ingests records

Record Insight

Categorization Agent

Your RCS + NARA General Records Schedule

The model analyzes and justifies the choice

Disposition rules applied to the record

Record Insight pipeline, top to bottom: RMA iQ ingests records; the Categorization Agent queries your Record Control Schedules and the NARA General Records Schedule and gets a schedule match; the model analyzes the record and justifies the choice; then disposition rules are applied to the record.

Why it's different

No rules to maintain

It understands the records instead of relying on rules someone has to configure and keep current.

No training required

Record Insight reads against your retention schedule directly, applying your rules as written.

No hallucinations

Every categorization is grounded in your record control schedule, traceable to the exact clause it matched.

In practice

One pass, five layers of labels.

Record Insight reads everything the record can tell you, then puts each layer to work downstream.

Sensitivity

PII, PHI, CUI-Export, ITAR Cat XVIII

Drives ABAC access and inline encryption.

Disposition

TR-05 five-year, GRS 1.2 destroy-after-audit

Starts retention timers and legal-hold wiring.

Functional domain

Contracts, HR-Personnel, Budget

Powers search facets and workload routing.

Ontology extraction

F-16 avionics parts, pandemic response

Feeds analytics and knowledge graphs.

Risk flags

Attorney-client privilege, FOIA-Exempt (b)(6)

Directs review and redaction workflows.

Explore the rest of AI in Feith

Five more capabilities, each its own deep-dive.

These are the other capabilities in the AI in Feith family — separate tools that each share the same permission model, the same audit trail, and your choice of large language model.

Back to AI in Feith ›

Get started

Manage every record, even the hardest to reach.

Bring a sample of your records and your record control schedule. We'll show you Record Insight running live and walk you through the reasoning and audit trail behind every categorization.