Experience layer
Azure AI Foundry / Copilot
Copilot-based tutoring experience designed for privacy-conscious academic environments.
Flagship use case
Piper: Feedback to institutional action.
A real applied workflow where preservice teacher feedback becomes structured insight, stakeholder communication, and human-approved product improvements — coordinated through Meridian.
Piper is a concept-first AI tutoring experience designed for mathematics learning. It emphasizes questions, representations, and structured thinking rather than direct answers — producing meaningful feedback signals that can be analyzed and acted on through Meridian.
How the workflow moves
From interaction to approved change.
Seven predictable stages. Reasoning, delivery, and execution happen in distinct systems; a human decision-maker reviews the work before anything ships.
- 01
Piper interaction
Student engages the tutoring experience
- 02
Feedback intake
Structured responses collected
- 03
Analysis & patterns
Themes and signals surfaced
- 04
Meridian delivery
Approved work routed to real destinations
- 05
Stakeholder deliverables
Summaries, decks, messages
- 06
Human approval
Decision-maker reviews
- 07
Product updates
Approved changes executed
System layers
Each layer does one job.
The architecture keeps reasoning, feedback capture, analysis, orchestration, execution, and governance as distinct responsibilities — not collapsed into a single monolithic model.
Feedback layer
Structured feedback collection
AI-assisted Google Forms (or equivalent) capture preservice teacher responses in organized shape.
Analysis layer
Google Gemini / Vertex AI
Pattern detection, summarization, and insight generation across the collected feedback.
Delivery layer
Meridian
Ridian's delivery and routing layer — policy-aware routing that connects approved Ridian OS outputs to real destinations across reasoning providers, collaboration tools, and enterprise systems of record.
Execution layer
OpenAI / Codex-style APIs
Used to generate artifacts and implement approved updates once a human has signed off.
Governance layer
Human-in-the-loop approval
No automatic deployment. Decision-makers review outputs before anything ships.
Deliverables
What the system produces.
Artifacts are generated from real feedback, reviewed by a human, and delivered into the tools institutions already use.
- KPI summaries Headline numbers for leadership
- Charts & insights Visualized patterns from feedback
- Slide decks Board- and department-ready
- Stakeholder emails Drafts for faculty and administrators
- Documentation Written records of findings and decisions
- UI/UX recommendations Proposed product improvements
Human approval
Institutional oversight by design.
Every output — summary, deck, email, documentation, or recommended change — is reviewed by a human decision-maker. Proposed updates are approved or rejected before execution. Nothing deploys automatically. The platform exists to support institutional judgment, not to replace it.
Piper is the tutoring experience referenced throughout this workflow — surfaced here as a flagship example of Ridian applied to a real institutional setting.