A county parks department operates a network of community centers offering recreational, educational, and cultural programming each season. Each activity proposal must pass through a structured review process before it is published — two departments evaluating submissions independently, each recording their decision before the overall approval status can advance. That process had been attempted twice before this solution, and neither attempt had resolved the core problem.
The original process lived in spreadsheets with no workflow capability and no mechanism for capturing reviewer decisions in a structured way. A move to SharePoint introduced shared access but the structural limitations compounded at scale — no enforcement of which fields were visible at which review stage, no mechanism to block a reviewer from advancing the overall status while a department review was still pending, and AI-generated content suggestions introduced as a workflow step but with no presentation layer to surface them effectively. After two prior attempts, the process still had no reliable enforcement, no audit trail, and no structured review workflow.
The replacement was built around a clear principle: each architectural layer owns what it does best. The canvas app owns form entry, client-side state, and submission validation. Classic Dataverse workflows own model-driven form field visibility and review transition enforcement — server-side, synchronous, and not bypassable. Cloud flows own automated state transitions, AI integration, and notifications. Dataverse owns the data model and its integrity.
Several architectural decisions shaped the implementation. A Business Process Flow (BPF) — the standard Power Platform visual review stage indicator — was evaluated and deliberately rejected. The multi-department review structure, where Community Programs and Recreation update their fields independently across different sessions, would create save conflicts if BPF stage advancement required all departmental fields to be updated simultaneously. Instead, three separate classic workflows enforce completion rules independently, producing the same enforcement result without the BPF's concurrent-update limitations. Formula columns were introduced to solve a Dataverse delegation problem that would otherwise silently truncate search results as submission volume grows. Staff-led activities are given a submission-time shortcut — the Community Programs review is automatically completed at submission for activities staff are personally leading, eliminating a review step that was never required for that category.
The activity submission canvas app handles four distinct scheduling types — single activities, premier events, activity series generated across date ranges, and ongoing series defined by selected months. Staff can pre-fill a new submission from any previously approved activity, reducing repetitive entry across recurring seasonal programming. Validation on submit checks seven required fields and surfaces specific missing-field messages without red borders appearing on initial form load.
On submission, a cloud flow immediately calls Azure OpenAI Service on Azure Government Cloud with the activity name and description. The model returns a grammar-corrected description and three to five alternative title suggestions, written back to the submission record before any reviewer opens it. Reviewers see the original submission and AI suggestions side by side. The AI enhances — the reviewer decides. No AI output is published without human review.
The model-driven reviewer application provides regional views filtered by five geographic service areas. Classic workflows control field visibility at each review stage and block the overall status from advancing while either department's review remains pending, surfacing a named error message identifying which review is outstanding. When both reviews are complete, a cloud flow sends an automated notification to the community center's site email — resolved from a formula column — and advances the record. Reviewers can @mention colleagues in Timeline notes, triggering instant in-app notifications without requiring a separate communication channel.
This engagement pattern applies directly to any government agency, parks department, school district, or nonprofit that manages recurring program or activity submissions requiring multi-party review before public delivery — and particularly to organizations that have already tried to solve this with spreadsheets or SharePoint and found those solutions insufficient at scale. The platform architecture is repeatable for grant application review, curriculum proposal workflows, event permit processing, community program registration management, or any process where submissions from distributed staff must flow through independent departmental approvals before reaching a final disposition. The specific design decisions documented here — the Business Process Flow rejection in favor of independent classic workflow enforcement, the formula column delegation fix, and the AI enhancement layer that generates suggestions without making decisions — are directly transferable to similar multi-department review architectures regardless of the program domain.
Schedule a 30-minute conversation about your operational challenges.
ore.brigham@brigville.com