APPX email management for state archives
QEPS turns collected email into searchable archives, working collections, and controlled public access. Automation handles the repetitive first pass. Archivists review the decisions that matter.
Runs where your records belong: inside your secure network, on infrastructure you manage.
What QEPS actually does
QEPS separates the source archive from the collection being processed. That gives archivists room to survey what arrived, select what belongs together, apply policy, and release only what is ready.
Searchable email archives
Build an archive around entities, email accounts, ingested packages, and individual messages. Search and inspect the source before deciding how it should be processed.
Processing collections
Create collections with groups, subgroups, and members, then associate the archives, accounts, packages, or individual messages that belong in each processing group.
Published collections
Publish or unpublish a completed collection without opening the working repository. The public sees searchable, approved records and access copies.
Preservation export
The QEPS roadmap includes export of permanent-retention email, originals and normalized email, separate attachments, and structured metadata for preservation workflows.
Keep the source. Shape the work.
QEPS keeps those ideas separate, then links them. An account or package can be surveyed in the archive and brought into the right collection without losing its source context.
Inside the email archive
An email archive is a searchable repository defined for one entity. It contains the emails sent and received through every email account owned by that entity. The entity is usually a person, but it can also be an organization or another defined subject of archival work.
Keep the entity’s correspondence together while preserving account and source context across the archive.
Ingest the source
To create an archive, the entity’s email must first be ingested. QEPS guides the work from a desktop upload through extraction, validation, parsing, and indexing.
A user-friendly drag-and-drop interface lets staff select files from any storage device accessible from their desktop, review the list, and initiate the upload.
QEPS extracts messages from ZIP, PST, and MBOX files and saves each email in EML format.
Extracted email is virus checked and assigned checksums for fixity purposes.
The system extracts metadata and attachments, and identifies the relationships that form email threads.
Extracted metadata and email content are used to build a searchable index of all email for the entity.
Archival discovery
The archive search is designed for archivists conducting discovery. Results provide a focused path into the correspondence, its attachments, and the relationships between messages.
When an archivist opens an email, QEPS also shows a table of the 50 additional archive emails most like it. Any of those emails can be selected and opened to continue discovery.
Inside the email collection
An email collection brings together related email from one or more archives for processing as a group. When every email is complete, the collection may be published, exported for preservation, or both.
Each member can draw selected email from one or more email archives, keeping the processing objective separate from the source archive.
A collection cannot be published or exported for preservation until all of its email has been processed.
The processing objective
Processing options
QEPS offers multiple processing methods. Each can make a policy assignment and require an archivist’s review when the rule calls for it.
Search for user-defined patterns such as credit-card, Social Security, and driver-license numbers. A match can also be automatically redacted.
Search against a user-defined lexicon—such as a legal lexicon—and act when an email exceeds the specified relevance threshold.
Match designated addresses within an email. The matching address can be automatically redacted when policy requires it.
Search for domain names in email addresses, URLs, and hyperlinks. A matching domain can be automatically redacted.
Search for user-defined hyperlinks and automatically redact the matching hyperlink when required.
Analyze content to assign Record Class. That value can also assign Access Restriction, Record Retention, and labels—and require review.
Analyze content to assign Access Restriction. That value can also assign Record Class, Record Retention, and labels—and require review.
Analyze content to assign Record Retention. That value can also assign Record Class, Access Restriction, and labels—and require review.
Pattern, lexicon, address, domain, and hyperlink processing can assign a record class, access restriction, record retention, or label—and can require manual review. Where supported, the matched value can also be redacted automatically.
Content-based decisions
Record Class, Access Restriction, and Record Retention assignment use the same content-analysis methods. Each targets its associated value and any related action values.
An archivist assigns the target value to a random sample. QEPS uses emails with a given value to find emails with similar content and assign the same value.
A local AI model analyzes the email’s content and applies the program’s rules to assign the appropriate Record Class, Access Restriction, or Record Retention value, plus related actions.
QEPS evaluates the actions from its processing sources and determines the appropriate values for the email.
Completion and human control
An email is complete only when it has a record class, access restriction, and record retention—and every mandated review has been performed.
Archivists may review any email, assign those values directly, and redact its content. A manual action always takes precedence over a system action.
QEPS can automatically create attachment access copies based on media type. Archivists can review and redact those copies as needed.
Processing with rules, context, and review
Teams define the vocabularies and actions that matter to their program. QEPS applies them consistently across a collection and calls attention to conflicts, exceptions, and uncertain results.
What your team defines
What QEPS examines
What comes out
Human intelligence + local analysis
Pattern matching, lexicon analysis, and semantic comparison help find sensitive content, related messages, and likely processing actions without reading every message from a blank page.
Reviewers validate results, settle conflicting assignments, change individual records, control redactions, and approve the collection before publication.
QEPS can run its semantic model inside the deployment. Message text does not need to be sent to a public AI provider for meaning-based search.
Local AI for email collections
QEPS runs a local large language model on the QEPS server. It reads each email against the rules your team defines, then returns a decision a reviewer can inspect, validate, or change.
Record classRecord
Access restrictionPublic
Record retentionPermanent
High confidence. The result cites the message content and the user-defined rules that support each assignment.
Flag low-confidence, conflicting, or likely-redaction results for an archivist.
Message text, attachments, and processing context stay within the QEPS deployment instead of being sent to an online AI service.
A local model avoids per-message or per-token charges as collections grow, letting teams plan around infrastructure they manage.
Local processing avoids repeated transfers, provider queues, and request limits when QEPS is reviewing large volumes of email.
QEPS identifies messages that need manual redaction review or result validation. With well-defined rules and confidence thresholds, AI can reduce manual review and classification to as few as 1% of a collection.
Installation and deployment
QEPS is not a hosted solution. It is installed on a Linux server inside your secure network, keeping email processing, local-model inference, search, and archive data in your environment for security and privacy.
Memory64 GB RAM minimum
CPU16+ high-performance cores
GPUNVIDIA RTX 509032 GB VRAM minimum
Working storage2 TB+ high-speed storage
Archive and collection storage varies with the number of email accounts, the volume collected, and the processing work retained. A reasonable initial deployment starts at 16 TB; long-term requirements can approach 128 TB or more.
These estimates do not include storage required for system backups.Optional APPX service
Public access
QEPS creates a separate searchable index for approved collections. Staff can publish, unpublish, or withhold a collection while the original archive and internal processing data remain under administrative control.
Search by keyword, natural-language prompt, subject, date, sender, or recipient.
Narrow with labels, entity tags, dates, addresses, and domains.
Read the approved email as PDF and open the permitted attachment copy.
Email in its archival context
A QEPS collection can carry AXAEM references for its bibliographic record, agency, entity, and contact. That connects processed email to the broader records-management and archival description around it.
Explore AXAEM ↗QEPS resources
These CoSA ShopTalk recordings explore the challenges of state-government email and the practical role of automation and AI. Additional QEPS reference materials will be collected here as they become available.
State archives face immense volumes of email and strict records-management duties. This session examines how AI can identify official records and non-records, flag sensitive information such as PII, and apply restrictions, retention schedules, and descriptive metadata—improving consistency while keeping long-term access in view.
Watch the recording ↗An introduction to practical ways AI can support state archives. The session considers where automated analysis can assist staff, the policy and review controls that keep people accountable, and opportunities to make archival work more consistent and scalable.
Watch the recording ↗State-government email creates a large, complex workload before public access or preservation. This session presents a hybrid approach: use computer algorithms for repetitive work and direct human review to the exceptions and decisions that need judgment.
Watch the recording ↗See the real workflow
A useful QEPS demonstration starts with the kind of email your team receives and the decisions you need to make. Ask about a guided demo or a trial run with representative material.
