Storage & Classification
AWS accounts, category-to-storage routing, and a category → type → subtype hierarchy with OCR and AI-review flags per type.
Employee & HR Experience · Document Management System
Every Benefits Document, Filed, Read, Retained — and Provably Deleted
Categories, OCR extraction, approval routing, retention policies, legal hold, AI search, and an append-only audit trail.
The BR Document Management System is the filing cabinet behind every module: employee uploads, administrator uploads on an employee's behalf, plan documents, and OCR-extracted evidence. Administrators configure storage, a category → type → subtype hierarchy, the fields OCR should extract, allowed file types, and retention rules; then every document is classified on upload, read inline, routed for approval by a business rule when it should be, stamped with an expiration date, searchable by Athena once embedded, and logged on an immutable trail through to its deletion.
OCR at upload
Rule-routed approval
Retention + legal hold
AI-searchable
Reimagining Benefits Document Management
"Think of the DMS as three rooms plus a config room that must be set up first. Config: storage, categories and types, extraction mapping, file types, retention. Operate: the repository, document detail, embeddings, the plan document library. Watch: retention policy, the audit trail, analytics. An admin builds the cabinet and labels the drawers before anyone files a document — and the shredding schedule is written before the first page goes in."
Configuration
Storage, classification, extraction, file types
Upload & OCR
Classify, extract, review, save
Approval & Lifecycle
Rule-routed approval, status, legal hold
Retention & Legal Hold
Jurisdiction, policy, triggers
AI Search, Library & Audit
Embeddings, plan documents, the trail
3
levels of classification — category, type, subtype — with OCR and AI-review flags on the type and subtype, and each category routed to its own storage location
2
different AI pipelines — OCR extracts structured fields at upload time for review before save; embedding vectorizes a document later so Athena can search it
5
jurisdiction levels for retention — federal, state, country, regional, employer-specific — with most-specific-wins matching and a priority tiebreaker
0
deletions under an active legal hold — the job records the attempt as blocked, never silently skipped
From Configuration to Defensible Deletion
Follow one document — a dependent's birth certificate — through the cabinet it lands in, the OCR that reads it, the rule that routes it, the retention clock that starts, the search index it joins, and the trail that records everything.
Build the cabinet before anyone files
Configuration is where the DMS is defined. Storage: AWS accounts and a category-to-storage map so each kind of document lands in the right place. Classification: categories, types, and subtypes, with OCR and AI-review flags on the type and subtype. Extraction mapping: a global catalogue of extractable properties — name, date of birth, identifiers — and a mapping of which properties apply, required or optional, to which type or subtype. File types: the allow-list of extensions and MIME types the upload picker accepts.
What it delivers
- AWS configs with encrypted keys; category → storage routing
- Category → type → subtype hierarchy with OCR-at-upload and AI-review flags
- Property master and property mapping that decide what the OCR panel asks for per type
- File-type allow-list enforced at the picker — an extension not listed is rejected
Why it matters
Every later step — extraction, routing, retention — reads this configuration instead of code.
Adding a property to the master does nothing until it is mapped to a type; adding a type does nothing until its category has a storage location.
Upload once, read at the door
From the repository, an administrator uploads on behalf of any employee; an employee uploads for themselves. The modal asks who the document is for, then category, type, and subtype, then the file — validated against the allow-list. If the type or subtype has OCR on, an extraction panel runs against the mapped properties and shows editable fields before save; a required property that came back empty stops the save until it is filled. Uploads that are abandoned mid-way are recorded as not saved and cleaned up.
What it delivers
- Employee or on-behalf-of upload with the uploader's role and the subject employee on the record
- Cascading category → type → subtype pickers seeded from the page you came from
- Inline OCR of images and PDFs with LLM-assisted extraction against the mapped properties
- Uploaded → Saved lifecycle, with Not saved cleaned up automatically
Why it matters
Metadata is captured while the document is in hand, by the person who knows it.
OCR does not make a document searchable by Athena — that is the separate embedding step, chosen deliberately.
Whether it needs approval is a rule, not a flag
On save, a business rule for document management evaluates the document; if it returns an approval action, the document is routed into BR WorkFlow as a document-approval instance and the approver acts there. Otherwise it finalizes immediately. From Document Detail an administrator can edit metadata, approve or reject a pending document, change status — active, inactive, reopened, archived — place or release a legal hold with a reason and case reference, comment, preview, download, share a deep link, and reveal redacted PII as a logged action.
What it delivers
- Business-rule gate on save → BR WorkFlow document approval when required
- Document Detail with guardrail banners for legal hold and workflow locks
- Legal hold with reason and case reference; deletion attempts logged as blocked while it is active
- PII redaction toggle with a logged reveal; AI insights panel with quality, lifecycle, and suggestions
Why it matters
The same upload path finalizes instantly or routes for approval, depending on your rule — not a developer's constant.
Approve and reject for routed documents happen in WorkFlow; there is no button that bypasses it.
Retention is a policy with a clock and a witness
Retention types name the jurisdiction and authority — federal, state, country, regional, employer-specific. Retention policies set the period in years, whether the document auto-deletes, and a priority, scoped to a category, type, or subtype. On save, the most specific matching policy wins — subtype over type over category, priority breaking ties — and the expiration date is stamped on the document. Retention trigger configurations decide what happens as expiry approaches: notify through a Communication Hub template, flag for review, or archive.
What it delivers
- Retention types by jurisdiction and authority (IRS, DOL, state, employer)
- Policies with period, auto-delete, priority, status, and notes per classification
- Triggers at N days before expiry — NOTIFY, FLAG, or ARCHIVE
- Expiring-soon counts, coverage KPIs, and BR Insights for missing, conflicting, or outlier policies
Why it matters
The shredding schedule is written down, applied consistently, and visible before it acts.
Expiration is stamped at save; what happens at expiry depends on the trigger configuration and the auto-delete flag — and never on a document under hold.
Searchable, distributed, and accounted for
Embeddings make a saved document searchable: an administrator performs an embedding on a document, a URL, or a web page, the text is chunked and vectorized, and Athena can find it semantically from any module. The Plan Document Library distributes an uploaded SPD or SBC to a targeted audience for signature or acknowledgment, sends the notification through Communication Hub, and tracks each employee's response — viewed, accepted with e-signature, declined, pending. The Audit Trail lists every Document Log and deletion-audit row, with integrity verification and IntelliSuggest warnings such as one actor performing most deletions.
What it delivers
- Embeddings with chunk analytics, per-document delete, and a programmatic path for other modules
- Plan Document Library items with signature or acknowledgment type, audience targeting, deadline, and per-employee responses
- Append-only Document Log: upload, save, type change, legal hold, read, archive, delete — plus deletion outcomes
- Analytics: upload trends, category, status, type, retention compliance, large-file impact — each with drill-down and export
Why it matters
The cabinet can be searched, its plan documents can be signed, and its history can be proven.
Embedding does not auto-sync: re-run or delete it when a source document is reclassified or removed.
Core Capabilities of the Document Management System
Sixteen screens under one sidebar — configuration, operation, and watching — with every action written to a trail that auditors, carriers, and courts can read.
Upload with Inline OCR
Employee or on-behalf-of uploads validated against the file-type allow-list, with image and PDF extraction against mapped properties, editable before save and enforced for required fields.
Rule-Routed Approval
A business rule on save decides whether a document finalizes or routes into BR WorkFlow as a document approval — approvers act in WorkFlow, and the document is locked meanwhile.
Retention & Legal Hold
Jurisdiction-scoped policies with most-specific-wins matching stamp an expiration on save; triggers notify, flag, or archive before expiry; legal hold blocks every deletion path with a logged outcome.
AI Search & Plan Document Library
Opt-in embeddings that make documents searchable by Athena across modules, and a library that distributes SPDs and SBCs for e-signature or acknowledgment with per-employee tracking.
Audit Trail & Analytics
Append-only Document Log and deletion audit with integrity verification and IntelliSuggest warnings; analytics by trend, category, status, type, retention, and file size with drill-down and export.
Athena inside Document Management
Find It, Explain It, Prove It
Once a document is embedded, Athena can find it semantically from any module — an SPD clause, a policy letter, a carrier invoice. Inside the DMS, Athena explains the insight strips: why a policy is an outlier, which documents will expire together, why a deletion was blocked, and what a document's AI quality assessment means. For employees, it answers "what does this plan document say about…" from the library they were asked to sign.
- Semantic search — "the SPD section on dependent coverage age" answered from embedded chunks, with the source document
- Retention explanations — which policy matched a document and why — subtype over type over category, priority as tiebreaker
- Deletion diagnostics — blocked, skipped, and deleted outcomes explained per run, with the hold or trigger that caused them
- Quality and lifecycle — the AI insights panel on a document — extraction quality, status history, suggested next step
- Library follow-up — who has not yet signed an SPD, by population, before the deadline
Why were two documents not deleted in Monday's job?
Athena
Both are carrier invoices under an active legal hold placed by Finance on 09/10 for audit query FIN-2026-0093. The retention job reached them, recorded BLOCKED_BY_HOLD in the deletion audit, and moved on. They will be re-evaluated on the next run after the hold is released; nothing was skipped silently.
Blocked by hold
2 · FIN-2026-0093
Skipped · FLAG open
6 · awaiting review
Show the hold record
Show the 6 flagged
Verify the audit rows
Which SPD signers are still pending?
Athena
284 of 1,102 in the 401(k) division have not responded to the SPD 2027 signature request; 41 have viewed it without signing. The deadline is 10/31 and the notification has already been sent once, so a reminder would go through Communication Hub as a new send for a person to approve.
Ask Athena about the repository…
What the Document Management System Changes
Configuration, not code
Storage, classification, extraction, file types, and retention are administrator settings.
Metadata captured at the door
OCR reads the document in the upload modal, and required fields are enforced before save.
Approval when a rule says so
The same upload path finalizes or routes into WorkFlow depending on your business rule.
Retention that is written down
Jurisdiction-scoped policies, triggers before expiry, and a legal hold that blocks every deletion path.
Documents Athena can find
Opt-in embeddings make plan documents and evidence searchable across modules.
A trail nobody can edit
Every action and every deletion outcome is a log row, with integrity verification and audit warnings.
Where It Shows Up in the BR Experience
The DMS is an administrator application with an employee-facing edge. Administrators configure and operate it in the console — on the platform's one data model, running on SAP BTP and AWS, with document objects in AWS storage. Employees upload and view their own documents in BR Profile and the mobile app, and respond to plan document library items from their own portal. Every other module — dependent verification, billing, workflow, communication — stores its attachments here.
- Admin: settings — AWS config, storage config with category mapping, category / type / subtype, extraction mapping, file types, plan document library
- Admin: DMS — documents repository with BR Insights strip, document detail workspace, embeddings, retention config, retention policy by jurisdiction tab
- Admin: administration — audit trail with verification and IntelliSuggest, analytics with drill-down and Excel export, dashboard with KPIs and Essential Insights
- Employee: BR Profile and the mobile app — upload for self, view and download own documents, respond to plan document requests with e-signature — scoped strictly to their own records
- Other modules: attachments — tickets, cases, notes, verification evidence, invoices, and communication templates all reference documents stored here
Bring a folder and a retention schedule
We'll configure a category tree and its storage, upload a document with OCR on, route one through WorkFlow, stamp a retention date, place and release a legal hold, embed an SPD for Athena, and open the audit trail that recorded all of it.
OCR + required fields
Legal hold wins
Append-only trail
Encryption at rest & in transit
Document Management FAQs
Every document the platform touches — employee uploads, administrator uploads on an employee's behalf, plan documents such as SPDs and SBCs, verification evidence, invoices, and attachments from tickets, cases, notes, and communications — as encrypted objects in AWS storage with metadata, classification, status, retention, and an audit trail on the platform.
If a document type or subtype has OCR enabled, the upload modal extracts the properties mapped to that type from the image or PDF and shows them for review before save. Required properties must be present. OCR captures structured fields; making a document searchable by Athena is a separate embedding step.
Only when your business rule says so. On save, the document-management rule is evaluated; if it returns an approval action, the document is routed into BR WorkFlow as a document approval and approvers act there. Otherwise the document finalizes immediately.
Retention policies are scoped to a category, type, or subtype and to a jurisdiction level — federal, state, country, regional, or employer-specific. On save, the most specific matching policy wins, priority breaks ties, and an expiration date is stamped. Trigger configurations notify, flag, or archive a set number of days before expiry; auto-delete runs only where the policy allows it.
No. An active legal hold blocks automated deletion, trigger actions, and manual deletion alike; the attempt is recorded in the deletion audit as blocked by hold, never silently skipped. Releasing the hold requires a reason and is logged.
Upload the SPD or SBC through Documents, then create a Plan Document Library item of type signature or acknowledgment, target the audience, and set a deadline. Notification goes out once through Communication Hub, and each employee's response — viewed, accepted with e-signature, declined, pending — is tracked per person.
Only their own documents — uploads for themselves and documents uploaded on their behalf — and the plan document requests addressed to them. Scoping is enforced in the backend by uploader and on-behalf-of fields, not just by page access.
No. Categories, types and subtypes, the file-type allow-list, property masters and mappings, retention types, policies and triggers, and the business rule that routes a document into WorkFlow approval are configuration inside Benefits Reimagined. Benefits and HR administrators maintain them without custom development; the storage itself runs on AWS under the platform on SAP BTP and AWS, with the platform data model holding the metadata, log and embeddings.
A Filing Cabinet That Can Testify
Configured before the first page, read at the door, routed by rule, retained by policy, searchable by Athena, and logged to the last deletion.