SCA's AI Advantage: Strategic Briefing
Author: Ed O'Connell, Director of Digital Strategy & AI Enablement, Springfield Commonwealth Academy
Date: February 2026
Interactive version: adambalm.github.io/sca-explainers/technical-briefing
The Core Metaphor
"A WordPress site without structured data is a pamphlet dropped on the sidewalk — it looks professional today, but when the next economic wind blows, it blows away. What remains is whatever was built on bedrock."
Executive Summary
TL;DR: AI is only as good as the data it can access. Without structured institutional data, AI tools give confident wrong answers. With it, they become genuinely useful. This briefing explains what "focusing on AI" actually requires.
Context
Angelene asked Ed to "focus on AI and credential" while a WordPress team handles the public website. This briefing explains why that's exactly the right division of labor — and what "focusing on AI" actually means.
What Has Been Built
Ed has already created:
- 11 typed document schemas in Sanity CMS (Page, Person, Program, News, Event, Alumni Story, Department, Media Gallery, Boarding, Admissions, Settings)
- Structured content architecture that AI can query with deterministic accuracy
- This briefing itself — generated from structured memory, demonstrating the pattern
What This System Is (Not Naive RAG)
This is Schema-Grounded Agentic Memory — a four-layer architecture:
- Schema-First Architecture: Typed document schemas in Sanity CMS with declared relationships
- Agentic Memory Pattern: Structured knowledge with temporal validity, supersession, and session continuity
- Knowledge-Grounded Generation: AI reasons over declared structure with deterministic GROQ queries, not probabilistic vector similarity
- Human Judgment Layer: Domain expertise for schema design, editorial review, and contextual awareness
Key Distinction from RAG
| Aspect | Traditional RAG | This System |
|---|---|---|
| Retrieval | Vector similarity (probabilistic) | GROQ queries (deterministic) |
| Structure | Inferred from embeddings | Declared in typed schemas |
| Relationships | Implicit in text | Explicit field references |
| Accuracy | "Probably relevant" | "Canonically true" |
Three Capabilities Enabled
- Instant Admissions Response: AI queries program data, drafts response in 5-10 minutes vs 24-48 hours
- On-Demand Board Reports: AI pulls structured data, generates report in 30 minutes vs 2-3 days
- Automatic Student Showcases: Project created → automatically published immediately
What Can Be Built Later
- Vector embeddings / semantic search
- Conversational interfaces (chatbots)
- Knowledge graph visualization
- External AI agent access (MCP, structured data APIs)
Key point: You can't add structure later. You can add everything else later.
Recommendation
Proceed with WordPress for the public site (Shawn's team) while Ed continues building the institutional cognition layer in parallel using Sanity CMS. Both paths are complementary:
- WordPress = the brochure (public-facing website)
- Sanity = the institutional brain (AI-queryable data layer)
Investor Talking Points
- "We're building the institutional brain, not just the public face"
- "Our content structure is AI-queryable from day one"
- "We're not buying AI — we're building the data layer AI needs"