Overview#
An investigator typing a subject's name into a new profile should not have to manually look up their known aliases, enter their last known address, and cross-reference open-source intelligence databases. Those steps take minutes for each record and introduce transcription errors at scale. Smart Fields watches as the investigator types, surfaces suggestions drawn from existing records and public data sources, and waits for explicit approval before committing anything. Every accepted suggestion is logged. Nothing changes without human sign-off.
Smart Fields is an AI-powered inline editing system that reduces data entry time while improving data quality through a multi-tier AI strategy and a human-in-the-loop approval workflow. It is designed for the data volume and accuracy requirements of law enforcement, intelligence organisations, and financial crime teams.
Key Features#
- Real-Time AI Suggestions: Context-aware recommendations appear as users type, with confidence scores and multi-field correlation that anticipates related values based on entity type and existing data.
- Human-in-the-Loop Approval: Every suggestion requires explicit user acceptance through a visual diff showing current versus suggested values. One-click accept or reject, plus batch approval for multiple suggestions at once.
- Multi-Tier AI Routing: Requests route automatically to the appropriate AI model tier based on complexity, balancing speed and accuracy for tasks ranging from simple field completion to complex entity analysis.
- Entity Extraction: Automatically identifies and extracts people, places, organisations, dates, and other entities from unstructured text such as investigation notes and reports.
- OSINT Data Enrichment: Enriches profiles by cross-referencing with public databases and open-source intelligence to discover related entities and fill in missing information.
- Evidence Metadata Extraction: Automatically extracts and suggests field values from evidence file metadata, reducing manual data entry for uploaded materials.
- Permission-Based Access: Role-based access control governs which enrichment features are available to each user, with tenant isolation, usage tracking, and configurable quotas.
- Complete Audit Trail: Records every suggestion, acceptance, and rejection with timestamps and user information for accountability and compliance requirements.
Use Cases#
- Accelerating person and organisation profile creation by suggesting formatted contact details, discovering related entities, geocoding addresses, and pre-filling fields based on available data.
- Streamlining investigation note-taking with automatic entity extraction, tagging, categorisation, timeline event suggestion, and related case linking as investigators type.
- Enriching existing records with open-source intelligence data to fill gaps, discover connections, and improve data quality across the platform without manual research.
- Reducing data entry errors through AI-powered field validation and formatting suggestions that catch inconsistencies and standardise data as it is entered.
Integration#
Connects with open-source intelligence providers, public databases, evidence management systems, and entity resolution tools to provide contextual suggestions drawn from across the platform and external data sources. All suggestion events and user decisions are stored in the PostgreSQL primary data store with organisation-level isolation and a complete audit trail.
Open Standards#
- GraphQL (June 2018 specification): All suggestion requests, enrichment proposal mutations, and HITL decision operations are exposed as a typed GraphQL API, with camelCase field names enforced as the public contract.
- OASIS STIX 2.1 / TAXII 2.1: Intelligence enriched into profile suggestions can be sourced from analyst-configured TAXII 2.1 feed subscriptions, with threat objects represented in the STIX 2.1 format as defined by the OASIS CTI Technical Committee.
- Exif (JEITA CP-3451 / CIPA DC-008): Evidence metadata extraction reads Exif tags from uploaded image files to automatically suggest field values such as GPS coordinates, capture timestamp, and device information without manual transcription.
- ISO 639-1 (Language codes): Detected language codes are attached to extracted text from evidence files, enabling the multi-language translation step that normalises foreign-language content to English before named-entity recognition runs.
- IANA Media Types (RFC 2046): MIME type inspection governs routing of uploaded evidence to the correct analysis tier (image EXIF extraction, PDF OCR, or plain-text entity extraction) and is validated before processing begins.
- ISO 8601 / RFC 3339 (Date and time format): All suggestion audit trail entries, batch creation timestamps, and accept/reject decision timestamps are stored and exchanged as ISO 8601 date-time strings with UTC offset, ensuring unambiguous temporal records for compliance purposes.
- OAuth 2.0 / JSON Web Tokens (RFC 7519): Role-based access control over enrichment features and configurable per-tenant usage quotas are enforced using JWT bearer tokens, with tenant isolation applied at the organisation level throughout the suggestion pipeline.
Last Reviewed: 2026-02-04 Last Updated: 2026-04-14