Purpose
Convert a diagnosed equity gap into a specific, sequenced intervention design that matches the right lever type (process, structural, or behavioral) to the actual root cause — so effort goes toward mechanisms that move outcomes instead of generic training. Solves the common failure where teams reach for an intervention (a workshop, a mentorship program) before diagnosing whether the gap is caused by process design, structural barriers, or behavior.
When to Use
Invoke this skill when the user asks to:
- Design an intervention to close a diagnosed representation, promotion, pay, or attrition gap
- Turn audit or listening-session findings into an actionable program
- Choose between process changes, structural fixes, and behavior-focused interventions for a specific gap
- Fix a specific broken point in the employee lifecycle (sourcing, hiring, onboarding, development, promotion, retention, exit)
- Build a defensible DEI plan tied to a root cause rather than a calendar of activities
- Get a second opinion on whether a planned intervention actually addresses the diagnosed cause
Inputs Required
Before starting, collect:
- The equity gap: The measurable disparity (e.g., "women are 42% of ICs but 18% of managers")
- Root cause type: What the diagnosis points to — process, structural, or behavioral (or "unsure — help me classify")
- Affected population: The group(s) experiencing the gap
- Lifecycle stage: Where in the employee journey the gap appears (sourcing, screening, interview, offer, onboarding, development, promotion, pay, retention, exit)
- What's been tried: Prior interventions and their results, if any
- Constraints: Budget, headcount, leadership appetite, timeline, org size
- Data availability: What's measured today and how often
If root cause type is missing or the user is unsure, run Step 1's classification before designing anything. If constraints are missing, design a default plan and flag decisions that depend on budget or headcount.
Instructions
Step 1: Classify the Root Cause (if not already diagnosed)
A gap has a symptom (the number) and a cause (the mechanism producing it). Interventions must target the cause. Classify into one of three lever families:
- Process cause: The workflow or decision procedure itself produces disparate outcomes — inconsistent interview rubrics, referral-heavy sourcing, unstructured promotion nominations, subjective ratings. Tell by: the disparity concentrates at a decision point and shrinks when the procedure is standardized.
- Structural cause: The system's design, policies, or resource distribution creates barriers — pay bands, benefits, scheduling, credential requirements, geographic access, sponsorship access, "who gets the stretch assignment." Tell by: the disparity exists before any individual decision is made; the rules themselves disadvantage the group.
- Behavioral cause: Individual and interpersonal patterns drive outcomes — bias in ratings, exclusionary team norms, uneven feedback, in-group affinity, microaggressions affecting retention. Tell by: the disparity persists even after process is standardized and structure is neutral.
State the classification explicitly and name the single dominant cause. If two co-exist, rank them and design for the dominant one first — behavioral interventions rarely fix a process problem.
Step 2: Select the Intervention Lever
Match the lever to the classified cause. Prefer the highest-leverage, most durable option available.
Process levers (redesign the decision):
- Structured interview rubrics with pre-committed criteria and scoring
- Blind or de-identified screening at the sourcing/screening stage
- Standardized promotion nomination criteria replacing manager discretion
- Diverse slate requirements at the hiring or promotion stage
- Calibration sessions for performance ratings with challenge questions
Structural levers (redesign the system):
- Pay band audits and correction with ongoing equity monitoring
- Sponsorship program with explicit access rules (not open-application mentorship)
- Revised credential/experience requirements that gate out qualified talent
- Flexible work, benefits, or scheduling policy changes
- Equitable distribution of high-visibility assignments via a tracked rotation
Behavioral levers (change the pattern):
- Manager feedback-equity training tied to a rubric and follow-up audit
- Norm-setting interventions on inclusive team behaviors with accountability
- Bias interrupters embedded in existing meetings (not standalone workshops)
- Peer accountability structures and reporting mechanisms
Rank the top 2–3 candidate interventions by expected impact and feasibility given constraints. Name the primary recommendation.
Step 3: Design the Primary Intervention
For the chosen intervention, specify:
- Mechanism of action: The exact causal chain — how this changes the number, in one sentence
- What changes concretely: The specific procedure, policy, or behavior that will be different, described precisely enough to implement
- Scope: Which teams, roles, or decisions it applies to, and any phased rollout
- Prerequisites: Data, tooling, leadership sign-off, or policy changes required first
Step 4: Sequence and Assign
- Implementation sequence: 3–5 phases from pilot to full rollout, with a defensible order (pilot narrow → validate → scale)
- Ownership map: For each phase, name the accountable role (RACI-style — who decides, who does the work, who is consulted)
- Timeline: Realistic milestones, distinguishing quick wins (0–90 days) from durable change (6–18 months)
Step 5: Define Success Metrics
- Leading indicators: Behavioral or process signals visible within 30–90 days (e.g., rubric adoption rate, slate diversity, calibration challenge frequency)
- Lagging indicators: The actual gap metric and the cadence to measure it — set a specific target and baseline
- Guardrail metric: One number that must NOT get worse (e.g., overall time-to-hire, quality-of-hire, non-target-group experience scores)
Step 6: Unintended-Consequence Check
Name 2–4 predictable side effects and how to prevent them:
- Backlash or perceived unfairness among non-target groups
- Tokenism, stigma, or "diversity hire" perception
- Gaming the metric (e.g., diverse slates filled without genuine consideration)
- Process burden causing quiet abandonment by managers For each, state the specific mitigation built into the design.
Output Format
- Structure: Six labeled sections in order, matching the instruction steps
- Length: 900–1,300 words
- Format: Structured document — prose for reasoning, bulleted specifics, and a compact ownership/timeline table
- Sections: Root Cause Classification → Lever Selection (ranked) → Primary Intervention Design → Sequence & Ownership → Success Metrics → Unintended-Consequence Check
- Tone: Organizational effectiveness advisor — evidence-oriented, direct, willing to say a popular intervention won't work if it doesn't match the cause
Quality Checklist
Before delivering output, verify:
- [ ] The intervention's lever type matches the classified root cause (no behavioral fix for a process gap)
- [ ] The mechanism of action states a specific causal chain, not a hope
- [ ] Both leading and lagging metrics are named, with a baseline and target for the lagging one
- [ ] At least one guardrail metric is specified
- [ ] Ownership is assigned to named roles per phase — not "the DEI team" for everything
- [ ] 2–4 unintended consequences are named, each with a concrete mitigation
Notes
- The most common design error is defaulting to training (a behavioral lever) for what is actually a process or structural problem — training doesn't fix an unstructured interview or an inequitable pay band.
- Structural and process levers are usually more durable than behavioral ones because they change outcomes even when individual attention lapses; recommend them first when the cause supports it.
- Diverse-slate and mentorship interventions are frequently misapplied: slates without structured evaluation just add a checkbox, and open-application mentorship rarely reaches those without existing access — sponsorship with explicit access rules does.
- If the gap spans multiple lifecycle stages, design for the earliest stage first; downstream gaps often shrink once the upstream mechanism is fixed.
How to use this skill
This file is a Claude Code skill in Anthropic's standard format — markdown with YAML frontmatter at the top.
Installing in Claude Code:
~/.claude/skills/<folder-name>/SKILL.md— auto-loads in every session anywhere on your machine (recommended for skills you'll use across multiple projects)<your-project>/.claude/skills/<folder-name>/SKILL.md— auto-loads only when working in that project (recommended for project-specific skills)
Restart Claude Code or open a new session to register.
Using in Claude.ai, ChatGPT, Gemini, or other AI tools:
- Drag this file into a new chat, or copy and paste its contents
- Send your task or request as your next message — the AI will apply the skill where it fits
For Claude Code skill details: https://docs.anthropic.com/en/docs/claude-code/skills