Document Screening

Overview

This skill adds a document screening stage before extraction or fulfillment workflows in ServiceNow®. It helps teams:

• Validate document completeness and format readiness
• Identify policy-sensitive or high-risk submissions early
• Route documents to automated extraction or human review lanes
• Record auditable screening rationale in work notes

Use this when your intake process receives mixed document quality, high volume, or compliance-sensitive content.

Prerequisites

• Roles: admin, sn_doc_intelligence_admin, or equivalent intake operations role
• Plugins: Document Intelligence app enabled where extraction is required
• Access: Read/write access to sys_attachment, sn_doc_template, sn_doc_intelligence_extraction, and task
• Inputs: Defined screening rules (required fields, prohibited content, confidence thresholds)

Procedure

Step 1: Build the Intake Queue

Collect newly submitted documents and classify by source, type, and priority.

Using MCP:

Tool: SN-Query-Table
Parameters:
  table_name: sys_attachment
  query: sys_created_on>=javascript:gs.hoursAgoStart(24)^table_nameISNOTEMPTY
  fields: sys_id,file_name,content_type,size_bytes,table_name,table_sys_id,sys_created_on,sys_created_by
  limit: 100
  order_by: sys_created_on

Using REST API:

GET /api/now/table/sys_attachment?sysparm_query=sys_created_on>=javascript:gs.hoursAgoStart(24)^table_nameISNOTEMPTY&sysparm_fields=sys_id,file_name,content_type,size_bytes,table_name,table_sys_id,sys_created_on,sys_created_by&sysparm_limit=100

Step 2: Match to Screening Template

Map each document to a known template and required evidence checklist.

Decision points:

• If template exists and is active -> continue to automated screening
• If template missing -> create review task for template owner
• If unsupported format -> route to manual conversion queue

Step 3: Run Screening Checks

Evaluate required attributes and extraction confidence signals.

Using MCP:

Tool: SN-Query-Table
Parameters:
  table_name: sn_doc_intelligence_extraction_result
  query: extraction=[extraction_sys_id]
  fields: field_name,extracted_value,confidence_score,validation_status
  limit: 200

Checks to apply:

• Required field presence
• Minimum confidence threshold (for example 0.85)
• Prohibited content indicators (policy-defined)
• Duplicate submission checks by hash or metadata fingerprint

Step 4: Assign Screening Outcome

Set one of three outcomes:

• Pass: Continue to extraction-to-record workflow
• Conditional Pass: Continue with mandatory human verification
• Fail: Route for remediation and return requester guidance

Using MCP:

Tool: SN-Create-Record
Parameters:
  table_name: task
  fields:
    short_description: "Document screening failed: missing required evidence"
    description: "File [file_name] failed screening checks. Review checklist and resubmit."
    priority: 3
    state: 1

Step 5: Record Audit Notes and Next Action

Document the screening rationale and the exact routing decision.

Tool Usage

Tool	Purpose
SN-Query-Table	Retrieve attachments, templates, and extraction results
SN-Get-Record	Inspect source record context for submitted documents
SN-Create-Record	Create remediation tasks for failed screenings
SN-Update-Record	Update screening status and routing fields
SN-Add-Work-Notes	Persist decision rationale for auditability

Best Practices

• Keep screening criteria versioned and tied to document type
• Treat confidence thresholds as policy controls, not hardcoded assumptions
• Separate business validation failures from technical OCR/extraction failures
• Capture decision evidence to support audit and dispute handling
• Periodically review false positives and false negatives to refine rules

Troubleshooting

Screening Queue Is Empty

Symptom: No documents appear in intake query. Cause: Time-window filter or source table filter is too restrictive. Solution: Broaden query window and verify attachment source integration.

High False Failure Rate

Symptom: Most documents route to fail or conditional pass. Cause: Confidence threshold or required field set is misaligned with real inputs. Solution: Tune thresholds by document type and add exception handling rules.

Related Skills

• document/document-extraction - Extract structured data from documents
• document/smart-documents - Manage document templates and generation
• procurement/invoice-management - Apply screened invoice documents to AP workflows

References

• https://github.com/ServiceNow/ServiceNowDocs/blob/australia/markdown/government-industry/psds-ai-skill-doc-screening.md
• https://github.com/ServiceNow/ServiceNowDocs/blob/australia/markdown/intelligent-experiences/ai-products.md