Policy answers hid across SharePoint and Confluence; permission?aware search curated one canonical HR source

Overview

A university’s HR team fielded repeated questions about policies and benefits because information was buried across SharePoint sites, Confluence spaces, and the Learning Management System (LMS). Different versions of the same policy circulated, search results varied by platform, and employees weren’t sure which source was current. Intelligex launched a permissions?aware enterprise search that indexed content across systems, normalized metadata, and surfaced a curated, canonical HR knowledge base. Results respected each system’s access controls, answers pointed to the latest approved documents, and HR received fewer tickets—while SharePoint, Confluence, the LMS, and existing publishing practices stayed in place.

Client Profile

• Industry: Higher education (university with academic, administrative, and auxiliary units)
• Company size (range): Multi?college campus with central HR and distributed department administrators
• Stage: HR content spread across SharePoint team sites, Confluence spaces, and LMS course pages; duplicate PDFs and outdated posts; search limited to individual platforms
• Department owner: Human Resources & People Ops (HR Communications and HRIS)
• Other stakeholders: Central IT, Information Security/Privacy, Accessibility, Legal/Policy Office, Benefits Administration, Labor Relations, Faculty Affairs, Student Employment, Employee Communications

The Challenge

Employees needed fast answers about leave, benefits, remote work, and pay practices. Instead, they navigated multiple portals with uneven search. The same policy might exist as a PDF on SharePoint, a wiki page on Confluence, and a training artifact in the LMS. Titles didn’t match, effective dates were missing, and search returned stale results as often as current ones. HR responded to repetitive questions, pasted links into tickets, and manually flagged outdated files.

Permissions and policy complexity compounded confusion. Some content was restricted by bargaining unit or campus, other content was open. Search on one platform returned items a person couldn’t actually access in another. Meanwhile, policy updates rolled out through email and meetings but took time to appear consistently across sites. Employees guessed which source to trust, and HR spent time clarifying rather than enabling self?service.

Accessibility and consistency were uneven. PDFs lacked clear headings, course pages embedded files rather than linking to a canonical policy, and wiki labels varied by team. Even when employees found a relevant page, the answer might be out of date or not applicable to their role.

Why It Was Happening

Content and search were fragmented. SharePoint and Confluence offered their own indexing and permissions, the LMS held HR training and references, and there was no cross?platform search that respected access controls while unifying metadata. Teams published what they needed in the moment; taxonomy and version control lived in documents rather than in a shared model.

Governance and signals were missing. There was no single catalog of canonical policies with effective dates, owners, and related FAQs. Search relevance favored recently edited pages over approved policy sources. Without a curated layer and an index that honored each system’s permissions, employees hit multiple results and guessed, and HR became the arbiter after the fact.

The Solution

Intelligex implemented a permissions?aware enterprise search that indexed SharePoint, Confluence, and LMS content and surfaced a curated HR knowledge base as the canonical entry. The search respected source system access controls, normalized titles and metadata, and boosted results that matched approved, current policies. Each canonical article linked to the underlying source of truth and related forms, with role and campus applicability noted. The architecture used Microsoft Graph search for SharePoint/OneDrive content (Microsoft Graph Search), Confluence APIs for wiki content (Confluence REST API), SharePoint content APIs (SharePoint REST API), and applied accessible content practices aligned to WCAG.

• Integrations: Connectors to SharePoint and OneDrive collections; Confluence spaces; LMS reference modules; identity provider for security trimming and role context; HR service desk for escalation when no answer matched.
• Metadata and taxonomy: Canonical policy catalog with titles, summaries, effective/expiration dates, owners, bargaining unit and campus tags; synonyms and acronyms; redirects from deprecated files.
• Permissions?aware indexing: Security trimming mapped to source ACLs; role and campus filters; results only for content a user can access; fallbacks and explanations when restricted documents exist.
• Curated knowledge base: HR?owned articles that summarize policy, link to the current source, and embed related forms and FAQs; version history and change notes; review cadence.
• Relevance and validations: Boosting for canonical and current content; demotion of stale or superseded items; checks for missing effective dates or owners; broken link detection.
• Publishing workflow: Draft, review, and publish steps for KB articles; maker?checker for policy changes; notifications to content owners when related source files change.
• Dashboards and reporting: Top queries, zero?result searches, click?through to canonical answers, stale content alerts, and escalation patterns to the HR service desk.
• Accessibility and format: Guidance and checks for headings, alt text, and link clarity; preference for web pages over image?only PDFs; “last reviewed” banners on canonical entries.
• Security and privacy: Role?based access to authoring; minimal personally identifiable information in results; encryption in transit and at rest; immutable logs of changes and indexing actions.

Implementation

• Discovery: Inventory of HR content across SharePoint, Confluence, and the LMS; audit of duplicate and stale pages; review of permissions models by site; catalog of high?volume HR inquiries; accessibility and privacy requirements.
• Design: Canonical policy schema (titles, dates, owners, applicability); search relevance strategy and security trimming; metadata mapping and synonym lists; publishing workflow and review cadence; dashboards and zero?result escalation.
• Build: Configured connectors to SharePoint, Confluence, and the LMS; implemented identity?based security trimming; built the curated KB with authoring and review; added validations for stale or orphaned content; instrumented analytics and logging.
• Testing/QA: Ran in shadow mode with internal HR users; validated permissions by role and campus; tested relevance and synonym handling on the top queries; exercised stale content and broken link detection; tuned boosting and demotion rules.
• Rollout: Launched to HR and pilot departments first; expanded to all employees after tuning; retained existing site search as a fallback; published “how to search” tips and quick links from HR portals.
• Training/hand?off: Trained HR content owners on the KB workflow and taxonomy; briefed IT on connectors and identity mapping; provided communications templates for campus rollout; updated SOPs for policy updates and deprecations.
• Human?in?the?loop review: Scheduled content councils to review top queries, zero?result trends, and policy updates; recorded decisions with rationale and effective dates; updated canonical entries, synonyms, and redirects accordingly.

Results

Employees found current policies faster and landed on a single canonical answer that matched the official source. Search respected permissions across systems and highlighted content applicable to their role and campus. The curated knowledge base reduced duplicate documents and pointed to forms and training in context, so employees finished tasks without opening a ticket.

HR shifted from answering repeat questions to governing content. Dashboards showed where searches failed, which policies needed updates, and which departments needed reminders to retire old files. The university kept SharePoint, Confluence, and the LMS as publishing tools; the change was a search and curation layer that made policy answers consistent, accessible, and auditable.

What Changed for the Team

• Before: Policy details were scattered across sites and files. After: A canonical HR knowledge base linked to the current source with clear applicability.
• Before: Search results varied by platform and permissions. After: Enterprise search respected access across systems with filters for role and campus.
• Before: Duplicate PDFs and stale pages caused confusion. After: Redirects and demotions hid superseded content, and owners received stale alerts.
• Before: HR answered the same questions repeatedly. After: Employees self?served common answers, and zero?result queries informed updates.
• Before: Accessibility and formats were inconsistent. After: Content followed accessible patterns with “last reviewed” banners and clear links.
• Before: Policy updates took time to propagate. After: Publishing workflows and notifications kept canonical entries aligned with source changes.

Key Takeaways

• Index where people store content; connect SharePoint, Confluence, and LMS rather than moving files.
• Make security first?class; apply permission trimming so results match what a user can actually access.
• Create a canonical layer; curate HR knowledge with owners, dates, and applicability to end duplicate answers.
• Tune relevance with governance; boost current policy, demote stale content, and close gaps from zero?result analytics.
• Embed accessibility and metadata; structured titles, headings, and tags drive better results and inclusion.
• Integrate, don’t replace; keep existing sites—add search, taxonomy, and workflows to make them usable.

FAQ

What tools did this integrate with? The search indexed SharePoint and OneDrive content through Microsoft Graph (Microsoft Graph Search) and SharePoint APIs (SharePoint REST API), Confluence spaces via the Confluence REST API, and LMS reference modules through existing connectors. Identity and single sign?on provided security trimming, and the HR service desk handled escalations when no answer matched.

How did you handle quality control and governance? A canonical policy catalog defined owners, effective/expiration dates, and applicability. Publishing used a draft?review?publish workflow with maker?checker for policy changes. Stale and duplicate content was detected and routed to owners for deprecation. Relevance rules boosted canonical entries and demoted superseded items, and decisions were logged with rationale and effective dates.

How did you roll this out without disruption? The search ran in shadow mode with HR first, validating permissions and relevance on top queries. A pilot group tested across departments while legacy search remained available. After tuning, the enterprise search link replaced older bookmarks, and existing sites continued to host content without migration.

How were permissions and privacy preserved? Results were security?trimmed using the source systems’ access control lists. Users saw only content they could access in SharePoint, Confluence, or the LMS. Sensitive terms in snippets were minimized, and access to authoring and analytics followed role?based permissions. All queries and changes were logged.

How did you keep content current? Canonical entries displayed “last reviewed” dates and owners. When source files changed or reached review dates, owners received alerts. Zero?result and high?bounce queries created tasks to add or update articles, and redirects retired superseded PDFs or pages.

What about accessibility and multiple formats? The team adopted accessible content practices aligned to WCAG. Canonical entries preferred web pages over image?only PDFs, enforced clear headings and link text, and included alt text for images when needed. Where PDFs were required, accessible versions replaced older files.

Can search results be filtered by campus, role, or bargaining unit? Yes. Metadata captured applicability by campus, employment class, and bargaining unit, and filters appeared automatically for users. Default relevance boosted entries most likely to apply to a user’s profile while still allowing manual refinement.