PeopleOps AI Onboarding: 30‑Day Enablement Playbook

yaml
playbook_version: 1.4
program: AI Onboarding Sequence
owners:
  people_ops: jane.cho@company.com
  enablement: devon.lee@company.com
  security: priya.raman@company.com
cohorts:
  - role: Support_Agent_L1
    regions: [US, EU]
    rbac_groups:
      okta: ["zendesk_l1", "ai_gateway_support"]
    tools:
      - name: support_copilot
        access: read_kb_generate_reply
        confidence_threshold: 0.72
        retrieval_coverage_min: 0.85
        approval_required: manager
    slo:
      ramp_time_days: {target: 14, warning: 16, breach: 18}
      safe_usage_rate: {target: 0.95, warning: 0.9}
      reopen_rate_pct: {target: 8, warning: 12}
    modules:
      - id: S1
        title: Getting Access + Safety Basics
        duration_minutes: 45
        content: [rbac_overview, residency_map, prompt_logging_walkthrough]
        gate: complete
      - id: S2
        title: Draft-Approve-Post Workflow (Live Lab)
        duration_minutes: 90
        scenario: "Handle 3 real tickets with PII redaction and manager approval"
        success_criteria:
          - all_prompts_logged: true
          - avg_confidence >= 0.72
          - manager_approval_rate >= 0.9
      - id: S3
        title: Quality Coaching + Edge Cases
        duration_minutes: 60
        content: [rejection_reasons, feedback_to_kb, escalation_policy]
    assessments:
      quiz_pass_score: 85
      attempts_allowed: 2
    approvals:
      steps:
        - name: security_signoff
          owner: priya.raman@company.com
          condition: "prompt_logging_enabled and residency=region"
        - name: legal_disclaimer_check
          owner: legal.ops@company.com
          condition: "external_message and approved_template=true"
    telemetry:
      warehouse: snowflake
      tables: [ai_prompt_logs, zendesk_ticket_metrics, adoption_events]
      daily_brief_channel: "#onboarding-ai-daily"

  - role: Sales_SDR
    regions: [US]
    rbac_groups:
      okta: ["salesforce_sdr", "ai_gateway_sales"]
    tools:
      - name: sales_email_copilot
        access: draft_email_generate_summary
        confidence_threshold: 0.7
        approval_required: manager
    slo:
      ramp_time_days: {target: 10, warning: 12, breach: 14}
      safe_usage_rate: {target: 0.95}
      rejection_rate_pct: {target: 5}
    modules:
      - id: A1
        title: Account Research with Knowledge Assistant
        duration_minutes: 60
      - id: A2
        title: Outreach Cadence Lab (Live)
        duration_minutes: 75
        success_criteria:
          - 15 approved drafts
          - avg_confidence >= 0.7
          - no residency_violations
program_policies:
  data_residency:
    EU: process_in_region_true
    US: process_in_region_true
  pii_redaction: enabled
  training_on_client_data: false
  default_deny_external_ai: true
review_cadence:
  daily: adoption_brief
  weekly: manager_coaching_sync
  day_30: scale_go_no_go

{
  "title": "PeopleOps AI Onboarding: 30‑Day Enablement Playbook",
  "published_date": "2025-12-12",
  "author": {
    "name": "David Kim",
    "role": "Enablement Director",
    "entity": "DeepSpeed AI"
  },
  "core_concept": "AI Adoption and Enablement",
  "key_takeaways": [
    "Design role-based AI onboarding paths with hard gates and RBAC from day one to prevent shadow AI.",
    "Measure ramp in business terms: task completion time, ticket quality, and safe usage rates, not training clicks.",
    "Bake governance into training: prompt logging, residency routing, and human-in-the-loop approvals.",
    "Ship in 30 days: audit current tools, stand up access + safety controls, pilot with one cohort, then scale.",
    "Never train on client data; log every prompt and decision for audit-ready evidence."
  ],
  "faq": [
    {
      "question": "How do we prevent shadow AI while we roll this out?",
      "answer": "Default-deny external AI endpoints at the network/identity layer and provision governed access on day one. Then give a clear, fast path with the approved copilot and daily coaching—most shadow use disappears when a better, safer option exists."
    },
    {
      "question": "What’s the minimum stack to start in 30 days?",
      "answer": "Okta/Entra for RBAC, Slack/Teams for the live lab and briefs, your existing systems (Zendesk/Salesforce), a VPC AI gateway with prompt logging and residency routing, and Snowflake/BigQuery for telemetry. We connect to what you already own."
    },
    {
      "question": "How do we measure onboarding success beyond course completion?",
      "answer": "Track time-to-first‑value (e.g., first solo ticket), safe-usage rate (suggestions above threshold with approvals), and quality (reopen/rejection rates). Publish a weekly scorecard by role and cohort."
    }
  ],
  "business_impact_evidence": {
    "organization_profile": "Global SaaS (1,200 FTE) with US/EU ops; Okta, Slack, Zendesk, Salesforce, Snowflake; SOC 2 Type II; HIPAA BAA on select workloads.",
    "before_state": "New hires reached first independent output in 5–6 weeks. No prompt logging or residency routing. Shadow AI usage in week one across two regions.",
    "after_state": "Role-based AI onboarding with RBAC, VPC AI gateway, and approval workflows. First independent output at 3–4 weeks; daily adoption brief and audit-ready logs.",
    "metrics": [
      "Ramp-to-first-solo-ticket cut from 21 to 12 days (43% faster).",
      "Safe-usage rate at 96% with confidence thresholds and manager approvals.",
      "18% reduction in ticket reopen rate during first 30 days.",
      "~350 hours returned in the first quarter across three onboarding cohorts."
    ],
    "governance": "Security and Legal approved because every prompt/response was logged to Snowflake with user/role, RBAC enforced via Okta, data stayed in-region via VPC AI gateway, PII redaction was on by default, and models were never trained on client data."
  },
  "summary": "PeopleOps playbook to cut ramp time with governed AI onboarding: role-based paths, RBAC, prompt logging, and a sub‑30‑day pilot that sticks."
}