Ops Workflow Automation with Human Review by Anesah FraserOps Workflow Automation with Human Review by Anesah Fraser

Ops Workflow Automation with Human Review

Anesah Fraser

Completed work

Automation Engineer

Python

Like this project

Completed work

Posted Feb 10, 2026

Developed an ops workflow automation handling emails to CRM/SHEET tasks.

Likes

Views

For independents

For companies

For companies Manage team projects Share hiring workspace Post jobs Discover contractors Sign contracts Approve invoices Global payments & tax compliance

Use Cases

For companies For freelancers For partners For agencies For VCs

Resources

Hire Freelancers

Explore all Top independents Design Engineering Marketing Music & audio Social media Video & animation Product Business consulting & HR Other

Contact

Contact support

Company

Mission Careers Terms & conditions Privacy policy Cookie policy

Ops Workflow Automation with Human Review

Anesah Fraser

Completed work

Automation Engineer

Python

Ops Workflow Automation (Email → Extraction → Sheet/CRM → Summary)

A ops intake automation that ingests inbox-style messages, extracts structured fields, routes low-confidence cases to a human review queue, writes approved items to mock destinations (Sheets/CRM), and records an audit trail for traceability.

Built as a demo package for “automation with guardrails” rather than a toy script.

What this demonstrates

Event-driven intake via /ingest (webhook-style)

Structured extraction (Pydantic models + JSON Schema validation)

Confidence scoring + review reasons (human-in-the-loop routing)

Idempotency protection keyed by message_id (safe retries)

PII redaction in Slack summaries (emails/phone-like strings)

Audit logging for every state change + integration write

Tests covering extraction + workflow behavior

Before / After

Before

Someone reads emails manually

Copies fields into a Sheet/CRM

Posts updates in Slack

Inconsistent structure, no audit trail, high rework

After

Webhook ingests messages

Extracts fields → validates schema → scores confidence

Low-confidence routed to human review with reasons

Approved items written to destinations + Slack summary

Audit log captures who did what, when, and why

Architecture

flowchart LR
  A["Inbox webhook JSON"] --> B["POST /ingest"]
  B --> C["Extractor + validation"]
  C --> D{"Confidence meets threshold?"}

  D -- "No" --> E["Pending review queue"]
  E --> F["POST /items/<item_id>/review"]

  D -- "Yes" --> G["Auto-approved"]

  F --> H["Write to destinations"]
  G --> H

  H --> I["Slack summary (PII redacted)"]

  C --> J["SQLite items + audit log"]
  F --> J
  G --> J
  H --> J
  I --> J

Like this project

Completed work

Posted Feb 10, 2026

Developed an ops workflow automation handling emails to CRM/SHEET tasks.

Likes

Views

Join 50k+ companies and 1M+ independents

Hire creatives Get hired

For independents

For companies

For companies Manage team projects Share hiring workspace Post jobs Discover contractors Sign contracts Approve invoices Global payments & tax compliance

Use Cases

For companies For freelancers For partners For agencies For VCs

Resources

Hire Freelancers

Explore all Top independents Design Engineering Marketing Music & audio Social media Video & animation Product Business consulting & HR Other

Company

Mission Careers Terms & conditions Privacy policy Cookie policy

Contact

Contact support

For independents

For companies

For companies Manage team projects Share hiring workspace Post jobs Discover contractors Sign contracts Approve invoices Global payments & tax compliance

Use Cases

For companies For freelancers For partners For agencies For VCs

Resources

Hire Freelancers

Explore all Top independents Design Engineering Marketing Music & audio Social media Video & animation Product Business consulting & HR Other

Contact

Contact support

Company

Mission Careers Terms & conditions Privacy policy Cookie policy

Ops Workflow Automation (Email → Extraction → Sheet/CRM → Summary)

Built as a demo package for “automation with guardrails” rather than a toy script.

What this demonstrates

Event-driven intake via /ingest (webhook-style)

Structured extraction (Pydantic models + JSON Schema validation)

Confidence scoring + review reasons (human-in-the-loop routing)

Idempotency protection keyed by message_id (safe retries)

PII redaction in Slack summaries (emails/phone-like strings)

Audit logging for every state change + integration write

Tests covering extraction + workflow behavior

Before / After

Before

Someone reads emails manually

Copies fields into a Sheet/CRM

Posts updates in Slack

Inconsistent structure, no audit trail, high rework

After

Webhook ingests messages

Extracts fields → validates schema → scores confidence

Low-confidence routed to human review with reasons

Approved items written to destinations + Slack summary

Audit log captures who did what, when, and why

Architecture

flowchart LR
  A["Inbox webhook JSON"] --> B["POST /ingest"]
  B --> C["Extractor + validation"]
  C --> D{"Confidence meets threshold?"}

  D -- "No" --> E["Pending review queue"]
  E --> F["POST /items/<item_id>/review"]

  D -- "Yes" --> G["Auto-approved"]

  F --> H["Write to destinations"]
  G --> H

  H --> I["Slack summary (PII redacted)"]

  C --> J["SQLite items + audit log"]
  F --> J
  G --> J
  H --> J
  I --> J