Sales Enablement Engine: 30-day call summaries and follow-ups

```yaml
service: sales_enablement_trust_layer
owners:
  - name: Priya Shah
    role: Director, RevOps
    email: priya.shah@company.com
  - name: Alex Romero
    role: Security Architect
    email: alex.romero@company.com
regions:
  - id: us-east-1
    residency: US
    model_endpoint: azure-openai-pep://use1-gpt4o
  - id: eu-west-1
    residency: EU
    model_endpoint: azure-openai-pep://euw1-gpt4o
controls:
  rbac:
    policy_source: salesforce_roles
    permitted_roles:
      - AE
      - SDR
      - SalesManager
      - SalesOps
  prompt_logging:
    enabled: true
    retention_days: 365
    sink: snowflake://revops_ai_logs
  redaction:
    pii_types: ["email", "phone", "credit_card", "ssn"]
    strategy: mask
    mask_token: "[REDACTED]"
  residency_routing:
    rules:
      - match: account.region == "EMEA"
        route_to_region: eu-west-1
      - match: default
        route_to_region: us-east-1
quality:
  summary_rubric:
    required_sections: ["problem", "stakeholders", "timeline", "objections", "next_steps"]
  confidence_thresholds:
    auto_approve: 0.88
    require_human_review: 0.70
  groundedness_check:
    method: semantic_alignment
    min_score: 0.82
workflows:
  call_summary:
    sources: ["gong", "zoom", "teams"]
    destinations:
      salesforce:
        objects:
          - type: Task
            fields:
              Subject: "Call Summary"
              Description: ${summary.text}
              Custom_Objections__c: ${summary.objections}
              Next_Steps__c: ${summary.next_steps}
      slack:
        channel_template: "#mgr-brief-${team}"
        message: "${account}: ${summary.key_points} | conf=${metrics.confidence}"
  follow_up_email:
    composer: outreach
    sequence: "DISCOVERY_FU_V2"
    draft_policy:
      if_confidence_below: 0.88
      require: AE_APPROVAL
    send_sla_minutes: 10
approvals:
  - name: LowConfidenceSummary
    trigger: metrics.confidence < 0.70
    steps:
      - role: AE
        action: REVIEW_EDIT
      - role: SalesManager
        action: APPROVE
  - name: PIIEscalation
    trigger: redaction.detected == true
    steps:
      - role: SalesOps
        action: VERIFY_REDACTION
      - role: LegalOps
        action: ACKNOWLEDGE
observability:
  metrics:
    - name: follow_up_latency_minutes_p95
      target: 10
    - name: auto_approve_rate
      target: 0.70
    - name: rubric_coverage_rate
      target: 0.90
  alerts:
    - condition: follow_up_latency_minutes_p95 > 10
      route: pagerduty://revops-ai
    - condition: rubric_coverage_rate < 0.85
      route: slack://#revops-ai-alerts
notes:
  - "Never train models on customer data."
  - "All logs immutable; exportable for audit."
  - "Runbooks: go/sales-ai-runbooks"
```

{
  "title": "Sales Enablement Engine: 30-day call summaries and follow-ups",
  "published_date": "2025-12-03",
  "author": {
    "name": "Lisa Patel",
    "role": "Industry Solutions Lead",
    "entity": "DeepSpeed AI"
  },
  "core_concept": "Industry Transformations and Case Studies",
  "key_takeaways": [
    "Stand up governed call summaries and actioned follow-ups in 30 days using your existing Gong/Zoom/Teams recordings and Salesforce/Outreach stack.",
    "Return 40% of rep admin time to selling and lift Stage 2→3 conversion with precise next-step sequencing and objection handling.",
    "Meet Legal’s bar on audit trails, prompt logging, RBAC, data residency, and never training on your data—no shadow AI.",
    "Instrument quality with confidence scores, objection taxonomies, and human-in-the-loop approvals for low-confidence cases.",
    "Scale beyond the pilot with enablement playbooks, telemetry, and an executive roll-up showing adoption and revenue impact."
  ],
  "faq": [
    {
      "question": "How do you prevent risky or off-brand follow-up emails?",
      "answer": "We enforce a confidence threshold and groundedness checks. Below 0.88 confidence, emails remain drafts for AE review; below 0.70, manager approval is required. The trust layer masks PII and maps tone to your brand guidelines."
    },
    {
      "question": "Will this flood reps with more tasks?",
      "answer": "No. Tasks are created in Salesforce only if the rubric detects clear next steps or risks. We optimize to reduce clicks—auto-fill fields, attach summarized context, and suppress duplicates."
    },
    {
      "question": "What about EMEA data and GDPR?",
      "answer": "All EMEA accounts route to EU regions. Prompts and outputs are logged with data residency controls and encryption. We provide DPIA-ready documentation and never train models on your data."
    },
    {
      "question": "Which models do you use?",
      "answer": "We prefer your private endpoints (Azure OpenAI or AWS Bedrock). We benchmark quality on your transcripts during Week 2 and can swap models without changing the workflow."
    }
  ],
  "business_impact_evidence": {
    "organization_profile": "120-rep global SaaS sales org, Azure VNet, selling to security leaders across NA and EMEA.",
    "before_state": "Manual note-taking and inconsistent objection capture; 33-minute median to draft follow-ups; Legal blocked earlier AI attempts over residency and logging gaps.",
    "after_state": "Governed summaries auto-pushed to Salesforce; Outreach drafts auto-prepared with AE approval under threshold; daily Slack rollups for managers.",
    "metrics": [
      "40% rep admin time returned to selling",
      "Stage 2→3 conversion lifted from 37% to 41.5% in pilot cohorts",
      "Median follow-up latency cut from 33 minutes to 7 minutes",
      "72% of follow-ups sent without manual edits at steady state"
    ],
    "governance": "Legal/Security approved due to prompt logging, RBAC aligned to Salesforce roles, regional routing to EU endpoints, KMS encryption, and a commitment to never train models on client data."
  },
  "summary": "RevOps can launch governed call summaries and auto-follow-ups in 30 days to cut admin time and lift conversion—integrated with Salesforce, Gong, and Outreach."
}