DC CAP Enterprise AI Pilot — Capstone
A working Claude Project, at least one documented workflow, and a governance configuration your team inherits.
June 2026
Describe the Claude Project you configured for your team. What organizational knowledge does it contain? What is it designed to help your team do?
Project Name
Who Uses This (1-2 sentences)
Project Instructions (What You Wrote in Claude)
Answer these three questions:
What Problem Does This Solve (2-3 sentences)
Document at least one workflow your team can follow. Step by step: what triggers it, what goes to Claude, what the human reviews, what goes out.
Workflow Name
Modality
Step 1: Trigger
Step 2: What Goes to Claude
Step 3: Human Review
Step 4: What Goes Out
Define the governance guardrails for your project. This is a scored deliverable — your team needs to know what data they can use, what's restricted, and what to do when something goes wrong.
Primary Data Tier for This Project
Select the highest sensitivity tier your project touches.
What's Allowed
What's Restricted
Escalation Path
Rate yourself on each of the 4Ds. Be honest — this is about growth, not performance. Then tell us what didn't work and what should come next.
Which D Was Hardest? Why?
What Didn't Work
What We Should Build Next
One KPI to Track for 90 Days Post-Pilot (View KPI Framework)
Unit: Student Success | Modality: Augmentation | Data Tier: Tier 3 (de-identified aggregates, strategy docs)
Project: Coach Prep Assistant
Project instructions tell Claude to act as a coaching preparation partner familiar with DC CAP's success model, retention benchmarks, and intervention playbook. Uploaded: partner retention summary (Tier 3), intervention decision tree, coaching session template. Governance guardrail: "Never generate content that includes or infers individual student identities. Always reference aggregate data."
Workflow: Weekly Cohort Check-In Prep
Trigger: Monday morning before campus visits. Input: Coach pastes de-identified cohort snapshot (engagement flags, GPA band, last-contact date). Claude produces: prioritized outreach list with suggested talking points. Human review: Coach validates against personal knowledge, adjusts priorities, removes anything that feels off. Output: finalized prep sheet the coach takes to campus.
Governance Config
Allowed: De-identified aggregate data, retention benchmarks, coaching templates, intervention frameworks. Restricted: Individual student names, SSNs, financial aid details, partner-specific MOU terms. Escalation: Data question → Stephanie Gardner (same day). Suspected Tier 1 exposure → Preston Magouirk within 2 hours per incident protocol.
4D Reflection
Delegation: 4 (confident choosing Augmentation for this task). Description: 3 (still refining prompt structure for cohort data). Discernment: 4 (strong at catching over-generalizations). Diligence: 5 (governance is second nature now). Hardest D: Description — getting Claude to produce the right level of specificity without over-prompting.