Knowledge
Key terms from AI, compliance and enterprise technology — explained for CEOs, CIOs and CFOs.
AI systems that autonomously make decisions, execute actions and orchestrate multi-step tasks — as opposed to purely generative models that only produce text.
At Ventum IQ: Our multi-agent platform uses Agentic AI to autonomously handle complex business processes — from tender analysis to contract review.
The EU regulation governing AI systems. Defines risk classes, transparency obligations and requirements for human oversight. In force since 2024, with phased application through 2026.
At Ventum IQ: Our AI strategy consulting incorporates AI Act requirements from the outset — ensuring your AI systems remain compliant.
SAP extension mechanism that integrates custom logic into standard processes without modifying SAP source code. Essential for clean-core-compliant extensions.
At Ventum IQ: Our SAP integrations leverage BAdIs and side-by-side extensions via BTP to preserve the clean core.
Statistical ranking model for keyword-based text search. Complements semantic vector search in hybrid search architectures, as it is better at finding exact technical terms and codes.
At Ventum IQ: Our RAG pipelines combine BM25 with vector search in a hybrid approach for maximum retrieval accuracy.
SAP strategy that minimises custom developments in the S/4HANA core. Goal: maintain upgrade capability, avoid technical debt, realise extensions via BTP side-by-side.
At Ventum IQ: We integrate AI capabilities as side-by-side extensions so your SAP core stays clean.
EU-wide standardised classification system for public procurement. A 9-digit code that categorises products and services in tenders.
At Ventum IQ: Tender IQ automatically analyses CPV codes to identify relevant tenders for your organisation.
A method where a neural network jointly analyses query and document (full bidirectional attention) to assess relevance more precisely than a bi-encoder.
At Ventum IQ: Our search pipelines use cross-encoder reranking as a second stage after the initial vector search.
Contract under Art. 28 GDPR between the data controller and data processor. Governs the handling of personal data when using external services.
At Ventum IQ: We execute a DPA with every client and ensure downstream LLM providers are contractually bound as well.
EU regulation for the operational resilience of the financial sector. Defines requirements for IT risk management, incident reporting and third-party oversight.
At Ventum IQ: Our platform supports DORA-compliant documentation and auditability for financial institutions.
Risk assessment under Art. 35 GDPR, required when processing activities pose a high risk to data subjects. Documents risks and countermeasures.
At Ventum IQ: We support you with DPIAs for AI projects and provide the necessary technical documentation.
A numerical vector representation of text in a high-dimensional space. Semantically similar texts are placed close together, enabling similarity search and intelligent matching.
At Ventum IQ: We generate embeddings from your corporate documents to enable precise semantic search in our RAG architecture.
An AI approach where decisions and results are transparent and verifiable by humans — in contrast to black-box models.
At Ventum IQ: Our systems display sources and decision paths transparently — for C-level traceability.
EU regulation for the protection of personal data. Defines data subject rights (access, erasure, portability) and processor obligations (purpose limitation, data minimisation, technical measures).
At Ventum IQ: GDPR compliance is a design principle — from data processing to our on-premise deployment option.
Safety mechanisms that control AI outputs: pattern matching filters obvious violations, LLM classifiers evaluate nuanced cases, webhook checks validate against external rules.
At Ventum IQ: Multi-layered guardrails are built into every platform solution — configurable per tenant and use case.
Cryptographic method for ensuring the authenticity and integrity of a message. Sender and receiver share a secret key.
At Ventum IQ: HMAC secures communication between our microservices and external interfaces.
International standard for Information Security Management Systems (ISMS). Defines requirements for risk assessment, security controls and continuous improvement.
At Ventum IQ: Our platform and processes align with ISO 27001 — for enterprise-grade information security.
A compact, signed token for the secure transmission of authentication information. Contains claims (user ID, roles, expiry time) in a Base64-encoded JSON format.
At Ventum IQ: JWT-based authentication secures access to all platform APIs and user interfaces.
A large language model such as GPT-4, Claude or Mistral. Trained on extensive text data, capable of understanding, generating and contextualising natural language.
At Ventum IQ: We deploy various LLMs — via API or on-premise with vLLM — depending on cost, latency and data privacy requirements.
A system of multiple specialised AI agents that cooperate in a division-of-labour fashion. Each agent has its own capabilities, tools and responsibilities.
At Ventum IQ: Our product platform is built on MAAIS architecture — specialised agents for analysis, research, generation and quality control.
An open protocol that connects AI models with external data sources and tools. A single connection exposes multiple tools that are automatically discovered.
At Ventum IQ: MCP enables our agents to access SAP, databases and other enterprise systems through a standardised interface.
Mutual certificate authentication: not only the server but also the client must present a valid certificate. Stronger security than standard HTTPS.
At Ventum IQ: mTLS secures service-to-service communication within our platform architecture.
An architectural principle where a single platform instance serves multiple isolated tenants (organisations). Each tenant has its own data, configuration and access rights.
At Ventum IQ: Our platform is fully multi-tenant — with strict data isolation between tenants.
EU directive on cybersecurity for critical infrastructures. Defines reporting and due-diligence obligations for organisations across 18 sectors.
At Ventum IQ: Our security architecture supports NIS2 requirements — from incident reporting to technical safeguards.
A single-use random value that prevents replay attacks. The receiver stores used nonces and rejects duplicates.
At Ventum IQ: Nonces protect our API communication against replay attacks and ensure the integrity of every request.
An architecture pattern where an LLM retrieves relevant document chunks from a knowledge base before generating an answer. Reduces hallucinations and enables up-to-date, source-based responses.
At Ventum IQ: RAG is the core of our products — from ChatFlow to Tender IQ to Contract IQ.
An access control model that assigns permissions to roles rather than individual users. Typical roles: Admin, Manager, Member.
At Ventum IQ: Fine-grained RBAC governs access to documents, agents and configurations — individually configurable per tenant.
A method for combining multiple search result rankings. Weights top ranks more heavily than lower ones, regardless of the absolute scores from individual search methods.
At Ventum IQ: RRF merges BM25 and vector search results into an optimal ranking within our hybrid search pipelines.
Protocol for unidirectional real-time notifications from server to browser. Simpler and more scalable than WebSockets for the typical case: server informing the client of changes.
At Ventum IQ: SSE delivers real-time streaming of AI responses and status updates to the user interface.
A process management model where a project passes through defined phases (stages) and quality gates. Each gate verifies the completeness of required information.
At Ventum IQ: Our AI implementation projects follow a stage-gate model — from discovery through pilot to production.
Measures under Art. 32 GDPR for the protection of personal data: encryption, access control, pseudonymisation, backup concepts.
At Ventum IQ: Our TOMs include AES-256 encryption, RBAC, audit logging and regular security reviews.
A specialised database for high-dimensional vectors (embeddings). Enables fast similarity search across millions of documents. Examples: Qdrant, Pinecone, pgvector.
At Ventum IQ: We use vector databases as a core component of our RAG architecture — for fast and precise document retrieval.
Open-source framework for high-performance LLM inference. Uses PagedAttention and continuous batching for maximum GPU utilisation in on-premise deployments.
At Ventum IQ: vLLM is part of our AI Sovereignty solution — for cost-effective and privacy-compliant on-premise AI.
International standard for accessible web content. Level AA covers: keyboard navigation, screen reader support, colour contrast, focus management.
At Ventum IQ: Our user interfaces are developed to WCAG 2.1 Level AA — for accessible AI applications.
A security architecture where every access — internal and external — is verified. No implicit trust, not even within the network.
At Ventum IQ: Our platform follows the zero-trust principle — every request is authenticated, authorised and logged.
Whether you want a demo of our products or a strategic consultation — we are here for you.