AI-Powered Job Intake Automation for Recruitment Agency

DNA Recruitment is a national medical recruitment agency placing healthcare professionals into hospitals and medical providers across Australia. The business manages high volumes of incoming job requests from dozens of clients, received across multiple inboxes in no standard format. The operations team was spending significant time manually reading, interpreting, and re-entering job details into their CRM.

Job requests arrived by email from dozens of hospitals and providers, landing across 10 separate inboxes with no enforced format. A single job could arrive as a plain email, a PDF attachment, a Word document, a scanned image, or an Excel spreadsheet, sometimes a combination of all of the above.

Each request required a staff member to:

• Identify which inbox received the job
• Read and interpret the content regardless of format
• Manually enter the job details into Bullhorn CRM
• Maintain the ability to reply from the correct original inbox

At scale, this process was unsustainable. The team needed a reliable way to capture every job automatically without disrupting existing client communication workflows.

Osher Digital built a multi-stage automation pipeline in n8n that handles the full journey from raw email to structured CRM record.

Email consolidation and traceability

All job inboxes forward into a single automation mailbox. Forwarding rules preserve the original sender and recipient details. Microsoft Graph API retrieves a direct web link to each source email, which is stored alongside the extracted job data. Admins can click straight back to the original email and reply as normal. The automation is invisible to clients.

AI-powered data extraction and normalisation

All email content and attachments are routed through a custom AI agent built to extract job facts and return structured data only. The pipeline dynamically handles email body parsing, attachment inspection, and file type detection, including correcting MIME type and extension mismatches that caused silent failures in early testing. The AI agent is constrained by strict instructions: return JSON only, no invented values, no mixing of explanation and data.

Job data lake

Extracted data is written to a master dataset with 30+ structured fields covering role details, location, dates, rates, employment type, grade, specialty, subspecialty, client, hospital, source inbox, and origin email link. This acts as a stable internal record that supports manual review, partial automation, and debugging without needing to reprocess emails.

Bullhorn CRM integration

Bullhorn requires specific internal IDs for most fields, and free-text values are rejected. Osher Digital built a full OAuth2 authentication flow with token handling and refresh logic, and created reference data lookup tables inside n8n that map human-readable values to authoritative Bullhorn API IDs. Lookup rules are hardened: if a confident match cannot be found, the workflow fails explicitly rather than pushing incorrect data silently.

Key technologies used:

• n8n for end-to-end workflow orchestration
• Microsoft Graph API for email traceability
• AI agent (Claude Sonnet) for intelligent data extraction
• Bullhorn CRM for job record creation
• Azure for supporting infrastructure