AI Customer Call Compliance Auditor with Kestra

id: ai-customer-call-compliance-auditor
namespace: company.compliance

description: |
  Audits customer service and outbound sales call transcripts for mandatory regulatory disclosures,
  flags prohibited deceptive claims, computes a compliance risk score using Google Gemini,
  and routes violations to compliance teams via Slack.

triggers:
  - id: scheduled_batch_sweep
    type: io.kestra.plugin.core.trigger.Schedule
    description: Nightly regulatory compliance sweep of recorded customer
      interactions. Shipped disabled by default.
    cron: "0 20 * * 1-5"
    disabled: true

  - id: webhook_transcript_ingest
    type: io.kestra.plugin.core.trigger.Webhook
    description: Authenticated webhook for contact center (Twilio, Genesys, Aircall)
      transcription pipelines.
    key: "{{ secret('WEBHOOK_KEY') }}"

inputs:
  - id: call_id
    type: STRING
    defaults: CALL-2026-9481
    description: Unique contact center call identifier.

  - id: agent_id
    type: STRING
    defaults: AGENT-402
    description: Support or sales representative identifier.

  - id: call_transcript
    type: STRING
    defaults: |
      Representative: Hello, thank you for calling Apex Cloud Solutions, my name is Alex. This call is being recorded for quality and training purposes. How can I help you today?
      Customer: Hi, I am calling to ask about upgrading our subscription to the dedicated enterprise tier.
      Representative: Absolutely! The enterprise plan is $4,500 per month billed annually. We include 24/7 dedicated support and unlimited compute clusters.
      Customer: What is your cancellation and refund policy if our engineering roadmap changes?
      Representative: All enterprise agreements include a 30-day money-back guarantee with zero cancellation penalty. After 30 days, billing renews automatically each year unless canceled 30 days prior.
      Customer: That sounds fair. Let's proceed with the setup.
      Representative: Perfect, I will send the agreement to your email for electronic signature. Thank you for choosing Apex Cloud!
    description: Raw dialogue transcript of the customer interaction.

tasks:
  - id: audit_call_transcript
    type: io.kestra.plugin.ai.agent.AIAgent
    description: Uses Gemini 2.5 Flash to audit transcript against mandatory
      disclosure checklists and prohibited sales claims.
    provider:
      type: io.kestra.plugin.ai.provider.GoogleGemini
      apiKey: "{{ secret('GEMINI_API_KEY') }}"
      modelName: gemini-2.5-flash
    configuration:
      temperature: 0.1
      maxToken: 4096
      responseFormat:
        type: JSON
        jsonSchema:
          type: object
          required:
            - call_id
            - compliance_verdict
            - compliance_score
            - recording_consent_disclosed
            - refund_policy_disclosed
            - recurring_billing_disclosed
            - prohibited_claims_detected
            - auditor_notes
          properties:
            call_id:
              type: string
            compliance_verdict:
              type: string
              enum:
                - PASS
                - WARNING
                - VIOLATION
            compliance_score:
              type: integer
              minimum: 0
              maximum: 100
            recording_consent_disclosed:
              type: boolean
            refund_policy_disclosed:
              type: boolean
            recurring_billing_disclosed:
              type: boolean
            prohibited_claims_detected:
              type: array
              items:
                type: string
            auditor_notes:
              type: string
    prompt: |
      You are an automated regulatory compliance auditor for an enterprise customer contact center.
      Analyze the following customer interaction transcript against regulatory consumer protection standards:

      Call ID: {{ inputs.call_id }}
      Representative ID: {{ inputs.agent_id }}

      Transcript:
      \"\"\"
      {{ inputs.call_transcript }}
      \"\"\"

      Audit Requirements:
      1. Recording Consent: Did the representative clearly notify the customer that the call is recorded?
      2. Refund & Cancellation Terms: Did the representative explain refund windows and cancellation rights if requested?
      3. Recurring Billing: Did the representative clearly state billing terms and renewal mechanisms?
      4. Prohibited Claims: Did the representative make unsubstantiated guarantees (e.g., "guaranteed 1000% ROI", "cannot fail", or hidden mandatory fees)?

      Assign an objective compliance_score (0 to 100):
      - 90-100: PASS (All mandatory disclosures made, zero deceptive claims)
      - 70-89: WARNING (Minor omissions or unclear terminology, no intentional deceit)
      - <70: VIOLATION (Missing recording consent, deceptive pricing, or prohibited guarantees)

  - id: generate_compliance_report
    type: io.kestra.plugin.scripts.python.Script
    description: Parses structured audit metrics and generates markdown compliance
      report artifact in execution storage.
    taskRunner:
      type: io.kestra.plugin.core.runner.Process
    script: |
      import json

      raw_response = """{{ outputs.audit_call_transcript.text }}"""
      audit = json.loads(raw_response)

      verdict = audit.get("compliance_verdict", "UNKNOWN")
      score = audit.get("compliance_score", 0)
      recording = "YES" if audit.get("recording_consent_disclosed") else "NO (FLAGGED)"
      refund = "YES" if audit.get("refund_policy_disclosed") else "NO (FLAGGED)"
      billing = "YES" if audit.get("recurring_billing_disclosed") else "NO (FLAGGED)"
      claims = ", ".join(audit.get("prohibited_claims_detected", [])) or "None detected"
      notes = audit.get("auditor_notes", "")

      report = f"""# CUSTOMER INTERACTION COMPLIANCE AUDIT
      ## Call Details
      - **Call ID**: {{ inputs.call_id }}
      - **Agent ID**: {{ inputs.agent_id }}
      - **Verdict**: {verdict}
      - **Compliance Score**: {score} / 100

      ## Disclosure Checklist
      | Mandatory Requirement | Status |
      |---|---|
      | Recording Consent Disclosed | {recording} |
      | Refund / Cancellation Policy Explained | {refund} |
      | Recurring Billing Terms Stated | {billing} |
      | Prohibited / Deceptive Claims | {claims} |

      ## Auditor Notes
      {notes}
      """

      with open("call-compliance-audit-report.md", "w") as f:
          f.write(report)

      print("Saved call-compliance-audit-report.md")
    outputFiles:
      - call-compliance-audit-report.md

  - id: compliance_gate
    type: io.kestra.plugin.core.flow.If
    description: Evaluates compliance verdict and routes flagged interactions to
      compliance officer Slack channel.
    condition: "{{ json(outputs.audit_call_transcript.text).compliance_verdict ==
      'PASS' }}"
    then:
      - id: log_compliant_audit
        type: io.kestra.plugin.core.log.Log
        description: Records passed audit log with compliance score.
        message: "Call {{ inputs.call_id }} PASSED compliance audit with score {{
          json(outputs.audit_call_transcript.text).compliance_score }}/100.
          Archived."
    else:
      - id: escalate_compliance_violation
        type: io.kestra.plugin.slack.notifications.SlackIncomingWebhook
        description: Alerts compliance officers of flagged disclosure omissions or
          deceptive sales claims.
        url: "{{ secret('SLACK_WEBHOOK_URL') }}"
        payload: |
          {
            "channel": "#compliance-escalations",
            "text": "🚨 *Call Compliance Alert*: Call `{{ inputs.call_id }}` (Agent: `{{ inputs.agent_id }}`) failed regulatory audit.\n*Verdict*: *{{ json(outputs.audit_call_transcript.text).compliance_verdict }}* (Score: {{ json(outputs.audit_call_transcript.text).compliance_score }}/100)\n*Notes*: {{ json(outputs.audit_call_transcript.text).auditor_notes }}\nAudit report generated in Kestra execution storage."
          }

errors:
  - id: alert_on_failure
    type: io.kestra.plugin.slack.notifications.SlackIncomingWebhook
    description: Alerts on-call engineering channel if transcript parsing or agent
      audit fails unexpectedly.
    url: "{{ secret('SLACK_WEBHOOK_URL') }}"
    payload: |
      {
        "channel": "#compliance-ops",
        "text": "⚠️ *Pipeline Error*: Call compliance auditor flow `{{ flow.id }}` failed on execution `{{ execution.id }}`."
      }

Customer service and outbound sales contact centers operate under strict regulatory scrutiny (such as FTC, CFPB, and FCC consumer protection mandates). Representatives are legally obligated to disclose call recording consent, accurately represent refund and recurring billing policies, and avoid deceptive or unsubstantiated sales claims.

Auditing thousands of call recordings manually is impossible: quality assurance (QA) teams typically sample less than 2% of recorded calls. Deceptive statements or omitted disclosure mandates go unnoticed until customer complaints trigger costly regulatory fines and reputational damage.

This blueprint automates 100% call compliance auditing. Contact center transcription pipelines dispatch completed call transcripts to Kestra via webhook. An AI compliance agent powered by Google Gemini analyzes the dialogue against mandatory disclosure checklists, computes an objective compliance risk score (0 to 100), and outputs structured findings. Clean interactions are logged and archived, while non-compliant calls automatically generate a downloadable audit report and dispatch an urgent escalation card to Slack for supervisory review.

How it works

audit_call_transcript (ai.agent.AIAgent): Evaluates dialogue text using gemini-2.5-flash against mandatory consumer disclosure mandates, returning a validated JSON schema payload (compliance_verdict, compliance_score, disclosure flags, and auditor notes).
generate_compliance_report (scripts.python.Script): Formats the audit breakdown and disclosure checklist into an executive call-compliance-audit-report.md artifact stored on the Kestra execution.
compliance_gate (core.flow.If):
- Compliant (PASS): Logs confirmation and records compliance pass metrics.
- Non-Compliant (WARNING / VIOLATION): Dispatches an interactive alert card to #compliance-escalations in Slack detailing agent ID, verdict score, and omitted disclosure items.
alert_on_failure (errors block): Catches downstream API errors and notifies the compliance operations channel.
Triggers: Nightly batch audit schedule (disabled: true by default) plus an authenticated Webhook trigger for telephony platforms (Twilio, Genesys, Aircall).

What you get

100% automated call QA coverage eliminating manual sampling blind spots.
Standardized audit scoring against consumer protection disclosure checklists.
Downloadable markdown audit reports stored natively in Kestra execution storage.

Who it's for

Compliance Officers, Legal Counsel, and Quality Assurance managers.
Contact Center Operations and Customer Success leadership.
Telephony and Voice Platform Engineers (Twilio, Talkdesk, Genesys).

Why orchestrate this with Kestra

Building AI evaluation gates inside bespoke telephony webhooks creates opaque failure states and lacks persistent audit lineage. Kestra provides declarative DAG orchestration, built-in artifact storage, centralized secret management, and native Slack escalations for human oversight.

Prerequisites

Telephony or transcription provider exporting call transcripts.
Google Gemini API key with access to gemini-2.5-flash.
Slack incoming webhook endpoint.

Secrets

GEMINI_API_KEY: API key for Google Gemini model inference.
SLACK_WEBHOOK_URL: Slack Incoming Webhook URL for compliance escalations.
WEBHOOK_KEY: Authentication secret for event-driven webhook invocation.

Quick start

Set GEMINI_API_KEY, SLACK_WEBHOOK_URL, and WEBHOOK_KEY in your Kestra namespace secrets.
Click Execute in the Kestra UI to run the blueprint against the default compliant transcript.
Modify call_transcript in the inputs to simulate an omitted recording disclosure or misleading refund claim, then observe the Slack alert.

How to extend

Add an automated speech-to-text task (such as Whisper or AWS Transcribe) upstream to process raw MP3/WAV audio recordings directly from S3 or Google Cloud Storage.
Connect io.kestra.plugin.core.http.Request to automatically tag non-compliant calls in CRM systems (Salesforce, HubSpot) for mandatory agent coaching sessions.

AI Customer Call Compliance and Regulatory Disclosure Auditor

How it works

What you get

Who it's for

Why orchestrate this with Kestra

Prerequisites

Secrets

Quick start

How to extend

Links

Tasks

Orchestrate with Kestra

Share this Blueprint

Create a Blueprint

New to Kestra?

AI Customer Call Compliance and Regulatory Disclosure Auditor

How it works

What you get

Who it's for

Why orchestrate this with Kestra

Prerequisites

Secrets

Quick start

How to extend

Links

Tasks

Orchestrate with Kestra

Share this Blueprint

Create a Blueprint

More AI Blueprints

Summarize Text with an AI Agent, Generate Multilingual Summaries in Short, Medium, or Long Format

Convert Natural Language Prompts to Structured Todoist Tasks with Deepseek AI

Generate and Send a Personalized AI-Powered Daily Digest Email Every Morning

Generate a book chapter summary using Perplexity AI and email it.

Generate AI Blog Posts in Your Brand Voice and Publish to WordPress as Draft

Trigger the Right Automation with an AI Agent Based on User Intent

More Business Blueprints

Weekly monday.com Portfolio Digest from One Raw GraphQL Query

Create a Dated Sprint Retro Board from a Miro Template on a Schedule

Audit New Miro Boards with a Trigger and Slack Notifications

Stale Miro Board Cleanup with a Confirmation Gate

React to New monday.com Items with Acknowledgement and Slack Summary

Monthly Miro Board Inventory Digest to Slack

New to Kestra?