Capstone: Your AI Product Case

What This Is About

You’ve worked through nine chapters: Foundations, Strategy, Design, Technical Literacy, Evaluation, Agentic AI, Ethics, Execution, and Leadership.

The Capstone Project connects everything. You choose an AI product or feature — ideally one you actually want to build or improve — and create a complete AI Product Case for it.

No quiz, no multiple choice. A document you could show to a hiring manager, a stakeholder, or your team.

The Assignment

Create an AI Product Case Document (10–15 pages) covering these eight areas.

1. Problem & Opportunity (Chapters 1–2)

Is AI the right solution? Apply the 5 Check Questions from Chapter 1.
Build / Buy / Blend — how do you implement the AI component?
PMF Risk Assessment — does AI threaten or strengthen your existing product-market fit?
Opportunity Sizing with RICE-A (Reach, Impact, Confidence, Effort + AI Complexity)

2. Product Design (Chapter 3)

Which UX pattern does your product use? (Copilot, Agent, Generative, Hybrid)
Trust strategy: How do you build trust? Confidence indicators, explainability, fallbacks.
User onboarding: How do you introduce users to AI features without overselling?

3. Technical Approach (Chapter 4)

Prompting / RAG / Fine-Tuning — which approach and why?
Model Selection with cost/quality/latency tradeoff
Expected cost per query and monthly budget

4. Evaluation Plan (Chapter 5)

Golden Dataset definition: How many examples, what distribution, who labels?
Metrics to track: Precision, Recall, F1, Hallucination Rate, Latency — depending on use case
Red Team Plan: Which attack vectors do you test? Priorities?
Ship/No-Ship Criteria: At what thresholds do you launch?
Bias Check: Which groups could be disadvantaged?

5. Agent Architecture (Chapter 6, if applicable)

Autonomy Level: At which level (L1–L5) does your product operate?
HITL Pattern: Approval Gate, Escalation Trigger, Parallel Review, or Checkpoint Audit?
Tool Strategy: Which tools does the agent need? How is access controlled?

If your product doesn’t use an agentic pattern: skip this section and briefly explain why.

6. Ethics & Governance (Chapter 7)

Responsible AI Reality Check: Apply the 6 steps
Guardrails: What guardrails do you set? Where is the over-blocking risk?
Privacy Tier: What data protection level does your product need?
EU AI Act Classification: Which risk category? What obligations?

7. Execution (Chapter 8)

AI PRD: Write one covering at least the 7 sections from Chapter 8
Lifecycle Phase: Exploration, Evaluation, Production, or Continuous Improvement?
Data Quality Plan: Where does the data come from? How do you ensure quality?
Cross-functional Setup: Who works together? Which roles?

8. Go-to-Market & Leadership (Chapter 9)

Pricing Model: Usage-based, per-seat, freemium, feature-tier?
KPI Dashboard: Three layers — Quality, Business, Operational
Team Structure: Which roles do you need? Centralized, hub-and-spoke, or distributed?

Quality Criteria

Your capstone is good if:

Every area contains a reasoned decision, not just a description
Tradeoffs are explicitly named (e.g., “we accept higher latency for better quality”)
Numbers are included: costs, metric thresholds, timelines
At least one risk is identified that could lead to failure
The document is understandable for a non-technical stakeholder

Tips

There is no “right” solution. There are well-thought-through and poorly-thought-through cases.
Use the Templates as a starting point for individual sections.
If you’re unsure about an area, go back to the corresponding chapter.
Real > hypothetical. The closer your case is to an actual product, the more valuable the result.