Overview#
An investigation team has accumulated thousands of pages of case documents, analyst reports, and evidence transcripts. A new analyst joins mid-investigation and needs to get up to speed fast. Instead of reading everything from scratch, she opens the AI Partner and asks: "What are the key financial connections between the three main suspects?" Within seconds, the AI answers from the actual case files, not from generic knowledge, citing specific documents and page references.
The AI Partner Platform makes that possible. It provides conversational AI infrastructure that lets organisations build, deploy, and monitor AI-powered chat experiences grounded in their own private knowledge bases. Every interaction draws on ingested documents through semantic search, so answers reflect real operational content rather than hallucinated generalities.
Key Features#
- Partner Development Kit (CDK): High-level conversation management APIs for building AI-powered chat experiences with multi-turn sessions, persistent context, and automatic message history management.
- Bring Your Own Knowledge (BYOK): Ingest documents from multiple sources (documents, webpages, databases, APIs, and custom sources) to ground AI responses in your organisation's private knowledge base using semantic search.
- Interactive Playground: Test and iterate on AI partner configurations, prompts, and knowledge bases in a sandbox environment before deploying to production.
- Conversation Management: Full session lifecycle management including active, paused, archived, and expired states, with token usage tracking and conversation statistics.
- Semantic Search: Vector-based document retrieval enables AI partners to find and reference the most relevant knowledge when answering questions.
- Document Processing: Automatic document chunking with sliding window overlap ensures comprehensive coverage of ingested content.
- Content Deduplication: Intelligent deduplication prevents redundant knowledge base entries, keeping your AI partner's knowledge clean and efficient.
- Usage Analytics: Track token consumption, conversation volumes, response quality, and session metrics for operational visibility.
- Programmable API Access: Full API support for managing AI partners, conversations, knowledge bases, and analytics programmatically.
Use Cases#
Law enforcement investigation teams deploy an AI partner loaded with case evidence and analytical reports, letting analysts ask natural language questions about case details without manually searching thousands of pages.
Intelligence agencies build internal knowledge assistants grounded in classified operational procedures, enabling cleared personnel to quickly locate protocols and precedents without exposing sensitive material to external services.
Financial crime units ingest regulatory guidance, typology reports, and internal policies into an AI partner that helps investigators verify whether specific behaviours match known money-laundering patterns.
Government onboarding programmes create AI-powered training assistants that help new staff learn complex operational workflows, referencing actual policy documents rather than generic training materials.
- Internal Knowledge Assistant: Deploy an AI-powered assistant that answers questions using your organisation's internal documentation, policies, and procedures.
- Investigation Support: Provide investigators with an AI partner that can search across case files, evidence summaries, and analytical reports to surface relevant information quickly.
- Customer-Facing Chatbot: Build customer-facing conversational experiences grounded in your product documentation and support knowledge base.
- Training and Onboarding: Create AI-powered training assistants that help new team members learn organisational processes and best practices.
Integration#
The AI Partner Platform integrates with other Argus modules to enhance conversational AI capabilities:
- Knowledge Management: Ingest and index documents from across the platform to build comprehensive AI knowledge bases.
- Case Management: AI partners can reference case data and investigation context to provide relevant assistance.
- AI Partners (Usage Tracking): Token consumption and costs are tracked through the AI Partners usage management module.
- Audit Trail: All AI partner interactions and knowledge base changes are logged for compliance and accountability.
Open Standards#
- GraphQL (June 2018 Specification): the entire API surface for managing AI partners, conversations, knowledge bases, and analytics is exposed as GraphQL queries and mutations, with a strongly typed schema defining all inputs and return types.
- JSON / JSONB (RFC 8259): all structured metadata, conversation context, AI partner configuration, and analytics payloads are stored and exchanged as JSON; PostgreSQL JSONB columns are used for flexible, indexed storage of partner and message metadata.
- RFC 4122 (UUID v4): every entity, including conversations, messages, knowledge sources, and document chunks, is assigned a randomly generated UUID v4 as its canonical identifier.
- ISO 8601: all timestamps for conversation creation, message history, knowledge indexing, and usage analytics are stored and returned as UTC-normalised ISO 8601 datetime values.
- OAuth 2.0 Bearer Token (RFC 6750): outbound requests to the Cloudflare Workers AI inference and embedding APIs are authenticated using a Bearer token in the HTTP authorisation header, conforming to the OAuth 2.0 Bearer Token Usage specification.
- Cosine Similarity (IEEE vector geometry): semantic retrieval of knowledge chunks uses cosine similarity via the pgvector PostgreSQL extension, computing angular distance between 768-dimensional dense embedding vectors to rank document chunks by relevance to a query.
Last Reviewed: 2026-02-05 Last Updated: 2026-04-14