metis

metis is an enterprise-ready AI platform developed by metaphacts that combines the generative power of large language models (LLMs) with the semantic precision of knowledge graphs. The result is a suite of knowledge-driven AI agents capable of delivering contextual, explainable and trustworthy insights across the organization. metis is designed for enterprises that need to move beyond generic AI and unlock the full value of their internal knowledge assets in a governed, compliant and transparent way.

At the heart of metis lies a semantic model that acts as a trust and context layer, informing and scoping every AI interaction. This ensures that AI-generated results are not only relevant but also aligned with enterprise semantics, traceable in their reasoning, and grounded in the right data sources. The platform is built on top of metaphactory, which is based on open standards, supporting interoperability and reusability across industries and business functions.

Key Value Propositions

• Ease of use: A conversational interface makes AI accessible to non-technical stakeholders, enabling natural language interaction with enterprise knowledge and delivering answers in diverse formats including tables, charts and other visual representations.
• Trust and explainability: Hallucinations are prevented by grounding results in explicit knowledge from the semantic layer. Explainable AI shows how answers were generated, and a built-in quality control framework allows users to validate and correct results.
• Compliance and privacy: Semantic models enforce data governance and scope data access for AI systems. Fine-grained access control mechanisms protect sensitive data, and traceability on how data is processed supports compliance with standards such as GDPR.
• Bias reduction: Enterprise-specific semantics and domain language guide AI responses, minimizing bias and promoting fair, transparent and accountable AI behavior.
• Unified data landscape: metis combines internal enterprise knowledge with public datasets into a single contextual knowledge graph, ensuring AI always draws from the right information sources.
• Extensibility and integration: The platform connects with existing AI frameworks and enterprise tools, supports custom use cases built on in-house knowledge solutions, and adapts to diverse tech stacks and strategic goals.

How metis Works

1. Users interact with AI agents through a conversational interface—no technical expertise or query language knowledge required—to gain insights, build semantic models and more.
2. AI agents interpret user intent, select the appropriate combination of tools, and execute tasks such as query generation, summarization and entity linking.
3. Each tool is orchestrated in real time and guided by the semantic model, which scopes what the AI sees and how it behaves, ensuring traceability and domain relevance.
4. A built-in quality control framework validates and corrects answers, supporting continuous improvement of agents.
5. Pre-defined agents and services provide assistance for tasks such as semantic modeling or search and discovery.
6. An Agent Registry manages both preconfigured and custom agents; new agents can be created by combining specific tools, semantic models and LLM configurations.

Cognitive Capabilities

• Language understanding: Enables intuitive natural language conversations between users and AI agents.
• Intent detection: Allows agents to accurately interpret what the user is trying to achieve.
• Planning and decision-making: Supports complex task execution by selecting the right tools and steps to fulfill user goals.
• Semantic reasoning: Leverages domain-specific semantics from the knowledge graph to ensure contextual accuracy and relevance in responses.

Execution Capabilities

• SPARQL query generation: Converts natural language questions into precise semantic queries.
• Vocabulary mapping: Aligns terminology across systems and data sources for consistent understanding.
• Retrieval-augmented generation (RAG): Enhances answers by grounding them in enterprise data and semantics.
• Entity linking and entity description: Identifies key concepts and connects them to their semantic representations.
• Context-driven response rendering: Generates final answers in clear, human-readable formats.
• Summarization: Condenses complex information into concise summaries.
• Custom tools integration: Supports custom execution logic or services for full extensibility.

Knowledge-Driven AI Agents

• Search and Discovery Agents for querying and navigating structured knowledge, configurable for specific use cases.
• Semantic Modeling Agent to support the creation and maintenance of domain-specific semantic models.
• Agent Registry to manage, configure and orchestrate both prebuilt and custom agents.

Custom agents can be composed by combining a selection of tools and capabilities (such as retrieval, summarization, entity linking and query generation), a semantic model providing contextual grounding, and configuration options including LLM choice, prompt templates and access scope. This composable architecture enables organizations to rapidly build and deploy agents for diverse use cases—from internal knowledge assistants to task-specific copilots—while maintaining contextual precision and explainable logic.

Human-AI Collaboration

• Human-in-the-Loop: Users remain actively involved in critical steps of the AI process—reviewing, validating and refining results. metis provides interfaces and quality control tools that make it easy to give feedback and guide agent behavior, enabling continuous improvement and compliance with enterprise standards.
• Augmented Intelligence: AI agents act as intelligent assistants that enhance human capabilities by automating routine tasks such as query generation and summarization, and surfacing contextually relevant information. Domain experts stay in control while benefiting from scalable, AI-powered support.

Open Standards Foundation

• RDF Data Model: Flexible, open standard for data representation.
• Ontology Language: Formal definition of domain models, extensible at any time.
• Vocabularies: Hierarchical categorization and classification of data.
• Rules and Constraints: Explicit cardinalities and constraints enabling automated reasoning.
• SPARQL Query Language: Flexible querying for graph-shaped information needs.
• Linked Data Platform: Open standards supporting FAIR data principles and data publishing.
• Dataset Descriptions and Data Cataloging Standards: W3C-based metadata making data discoverable, accessible and traceable.
• Web Components and HTML Templates: Rich components for search, visualization, exploration and authoring, with low-code application building via a flexible templating engine.
• Microservice Packaging and REST APIs: Stateless, scale-out design supporting flexible deployment environments, with dynamic publishing of queries as REST APIs.

metis is purpose-built for enterprise environments where trust, governance and explainability are non-negotiable. By bridging enterprise data silos with semantically grounded AI, it enables organizations to move from scattered data and generic AI outputs to a solution that delivers genuine business value across industries including pharma and life sciences, engineering and manufacturing, and beyond.