--- title: "Integration, BPM & AI Glossary | KONDEVS" description: "Plain-English definitions of enterprise integration, BPM and AI terms, from EAI, ESB, SOA and EDI to iPaaS, RAG and MCP, explained by the KONDEVS team." canonical: "https://www.kondevs.com/glossary/" language: "en" type: "website" image: "https://www.kondevs.com/images/og-default.png" --- > For the complete index of this site, see [llms.txt](https://www.kondevs.com/llms.txt); the whole site in one file: [llms-full.txt](https://www.kondevs.com/llms-full.txt). # Glossary Plain-English definitions of the enterprise integration, BPM and AI terms we work with every day. ## Agentic RAG **What is agentic RAG?** Agentic RAG is retrieval-augmented generation in which an AI agent plans the retrieval itself: it chooses sources, refines queries and checks the evidence before answering. Classic RAG retrieves once and generates once. In agentic RAG an agent runs the retrieval loop: it splits a question into sub-questions, decides which knowledge bases, APIs or tools to query, judges whether the evidence is sufficient and relevant, and retrieves again when it is not. This improves answers to multi-step and cross-system questions at the cost of more model calls and latency, so production designs cap the number of steps, log every retrieval and keep answers grounded in cited sources. Related: Retrieval-Augmented Generation (RAG), AI Agent, Multi-Agent System ## AI Agent **What is an AI agent?** An AI agent is a goal-driven system that uses a language model to plan, make decisions and take actions through tools, often with limited autonomy. An AI agent goes beyond answering a single prompt: it pursues a goal by planning steps, calling tools and APIs, retrieving information, and adapting based on results, sometimes coordinating with other specialist agents. Production-grade agents need clear architecture, scoped tool permissions, observability and human-in-the-loop checkpoints. Connected to enterprise systems via APIs, middleware and MCP, agents can automate content pipelines, intelligence hubs and multi-step workflows. Read more: [Cloudflare AI Gateway as an Enterprise Control Plane: Routing, Observability, and the Controls](https://www.kondevs.com/content-hub/cloudflare-ai-gateway-enterprise-control-plane/) Related: Multi-Agent System, Model Context Protocol (MCP), Retrieval-Augmented Generation (RAG), Large Language Model (LLM) ## API-Led Connectivity **What is API-led connectivity?** API-led connectivity is an integration approach that exposes data and capabilities through reusable, layered APIs (system, process and experience). API-led connectivity organises integration into three tiers of reusable APIs: system APIs unlock data from underlying systems, process APIs orchestrate and combine that data into business logic, and experience APIs tailor it for specific channels or apps. Each layer is independently governed and reusable, which speeds delivery and reduces duplicated point-to-point work. It is the dominant model behind modern integration platforms and the composable enterprise. Read more: [What Is Enterprise Application Integration and Why It Matters in 2026](https://www.kondevs.com/content-hub/what-is-eai/) Related: Service-Oriented Architecture (SOA), iPaaS (Integration Platform as a Service), Enterprise Application Integration (EAI) ## AS2 **What is AS2?** AS2 is a secure internet transport protocol for exchanging EDI and other business data with encryption, digital signatures and delivery receipts. Applicability Statement 2 (AS2) carries business documents over HTTP/S with S/MIME encryption and digital signatures, plus signed Message Disposition Notifications that act as non-repudiable delivery receipts. It became a de facto standard for B2B EDI exchange (notably mandated by large retailers) because it provides security and proof of delivery over the public internet without a dedicated value-added network. Related: Electronic Data Interchange (EDI) ## Business Process Management (BPM) **What is Business Process Management (BPM)?** BPM is the discipline of modelling, automating, executing and continuously improving an organisation's business processes. Business Process Management treats work as end-to-end processes that can be designed, measured and optimised. A BPM platform lets teams model a workflow (often in BPMN notation), automate the routing of tasks and data between people and systems, handle exceptions, and monitor performance against SLAs. Done well, BPM turns manual, undocumented handoffs into auditable, repeatable, continuously improved workflows. It pairs naturally with integration, which supplies the data and system connectivity processes depend on. Read more: [Software Dev Postings Rose 15%. The Job That Rebounded Isn't the One That Left.](https://www.kondevs.com/content-hub/software-dev-postings-rose-15-the-job-that-rebounded-isn-t-the-one-that-left/) Related: Process Mining, Camunda, Event-Driven Architecture (EDA) ## Camunda **What is Camunda?** Camunda is a process orchestration platform that executes BPMN workflows and DMN decisions, coordinating people, systems and services in one auditable process. Camunda runs standards-based BPMN process models and DMN decision tables. It gives developers a code-friendly way to automate end-to-end processes, coordinate microservices and human tasks, and see every running workflow. Camunda 7 (formerly Camunda BPM) was open source; its Community Edition reached end of life in October 2025. Camunda 8, the current platform, is source-available: free for development and testing, licensed for production. It is a common choice for business process automation in integration projects where transparency, exception handling and auditability matter. Read more: [Cloudflare AI Gateway as an Enterprise Control Plane: Routing, Observability, and the Controls](https://www.kondevs.com/content-hub/cloudflare-ai-gateway-enterprise-control-plane/) Related: Business Process Management (BPM), Process Mining ## Electronic Data Interchange (EDI) **What is Electronic Data Interchange (EDI)?** EDI is the structured, standardised electronic exchange of business documents (such as orders and invoices) between trading partners. Electronic Data Interchange replaces paper and email with machine-readable documents in agreed standards (EDIFACT, ANSI X12, etc.), letting partners' systems exchange purchase orders, invoices, shipping notices and more without manual rekeying. EDI is foundational to supply chain, retail, logistics and healthcare. Onboarding partners involves mapping their formats to internal data and enforcing compliance, often over transports like AS2. Read more: [webMethods vs. Open-Source Integration Platforms: A Practical Comparison](https://www.kondevs.com/content-hub/webmethods-vs-opensource/) Related: AS2, Enterprise Application Integration (EAI) ## Enterprise Application Integration (EAI) **What is Enterprise Application Integration (EAI)?** EAI is the practice of connecting an organisation's separate software applications so they share data and processes as one coordinated system. Enterprise Application Integration connects systems such as ERP, CRM, SCM, legacy databases and cloud SaaS that were never designed to work together. Rather than rebuilding applications, EAI adds an integration layer that moves and transforms data between them: through messaging, APIs, events or file transfer. The goal is to eliminate data silos, automate cross-system processes and give the business a single, consistent view of its operations. Read more: [How we made kondevs.com usable by AI agents](https://www.kondevs.com/content-hub/how-we-made-kondevs-usable-by-ai-agents/) Related: Enterprise Service Bus (ESB), Middleware, API-Led Connectivity ## Enterprise Service Bus (ESB) **What is an Enterprise Service Bus (ESB)?** An ESB is a middleware architecture that routes, transforms and mediates messages between applications through a central communication backbone. An Enterprise Service Bus provides a shared backbone for integration: applications connect once to the bus instead of point-to-point to every other system. The ESB handles message routing, protocol and format transformation, and orchestration of services. It reduces the brittle 'spaghetti' of direct connections and centralises monitoring and governance. ESBs are a common implementation of service-oriented architecture, though many organisations now complement or replace them with API gateways and event streaming. Read more: [Cloudflare AI Gateway as an Enterprise Control Plane: Routing, Observability, and the Controls](https://www.kondevs.com/content-hub/cloudflare-ai-gateway-enterprise-control-plane/) Related: Service-Oriented Architecture (SOA), Message-Oriented Middleware (MOM), IBM webMethods ## ETL (Extract, Transform, Load) **What is ETL (Extract, Transform, Load)?** ETL is a data integration process that extracts data from sources, transforms it into a target format, and loads it into a destination system. Extract, Transform, Load moves data from operational systems into a warehouse, lake or another application. Extraction pulls raw data; transformation cleans, validates, reshapes and enriches it; loading writes it to the destination. ETL traditionally runs in scheduled batches, while the related ELT pattern loads first and transforms in the target. Both are core to analytics, reporting and feeding data to machine-learning pipelines. Related: Enterprise Application Integration (EAI), MLOps (Machine Learning Operations) ## Event-Driven Architecture (EDA) **What is Event-Driven Architecture (EDA)?** EDA is a design pattern where systems communicate by producing and reacting to events rather than calling each other directly. In an Event-Driven Architecture, a change of state (an order placed, a payment cleared) is published as an event that any interested system can consume asynchronously. Producers and consumers are decoupled: they need not know about each other or be available at the same time. This enables real-time responsiveness, elastic scaling and resilience. Event brokers and streaming platforms such as Apache Kafka or message queues carry the events between services. Read more: [Software Dev Postings Rose 15%. The Job That Rebounded Isn't the One That Left.](https://www.kondevs.com/content-hub/software-dev-postings-rose-15-the-job-that-rebounded-isn-t-the-one-that-left/) Related: Message-Oriented Middleware (MOM), Enterprise Application Integration (EAI) ## IBM webMethods **What is IBM webMethods?** webMethods is an enterprise integration platform providing an ESB, API management, B2B/EDI, managed file transfer and process automation. IBM webMethods (Software AG webMethods until IBM acquired it on 1 July 2024) is a mature, comprehensive integration suite used by large enterprises for application integration (ESB), API management, B2B and EDI partner integration, managed file transfer and BPM. Its strengths are breadth of capability on a single platform, proven scale and vendor support; the trade-offs are licensing cost and a specialist skill set. KONDEVS has delivered webMethods integration and BPM projects across European finance, rail, insurance and logistics. Read more: [Cloudflare AI Gateway as an Enterprise Control Plane: Routing, Observability, and the Controls](https://www.kondevs.com/content-hub/cloudflare-ai-gateway-enterprise-control-plane/) Related: Enterprise Service Bus (ESB), Electronic Data Interchange (EDI) ## iPaaS (Integration Platform as a Service) **What is iPaaS (Integration Platform as a Service)?** iPaaS is a cloud-hosted platform that provides tools to build, deploy and manage integrations between applications and data without managing infrastructure. Integration Platform as a Service delivers integration capability (connectors, data mapping, workflow, API management and monitoring) as a managed cloud service. Teams build integrations in the platform rather than installing and operating middleware themselves. iPaaS suits hybrid landscapes that mix SaaS and on-premises systems, and accelerates delivery with prebuilt connectors, though enterprises weigh it against self-managed platforms for high-volume, mission-critical flows. Related: API-Led Connectivity, Middleware ## Large Language Model (LLM) **What is a large language model (LLM)?** An LLM is an AI model trained on vast text data to understand and generate natural language, powering chat, summarisation and reasoning tasks. Large Language Models are neural networks trained on extensive text corpora to predict and generate language. They underpin conversational assistants, document processing, summarisation, classification and code generation. In the enterprise, LLMs are most valuable when integrated with reliable data and tools (through retrieval-augmented generation, function calling and protocols like MCP) so their output is grounded, governed and connected to real systems rather than used in isolation. Read more: [LLM visibility: how AI assistants find, read and cite a company](https://www.kondevs.com/content-hub/llm-visibility-in-ai-answers/) Related: Retrieval-Augmented Generation (RAG), AI Agent, MLOps (Machine Learning Operations) ## Message-Oriented Middleware (MOM) **What is Message-Oriented Middleware (MOM)?** MOM is infrastructure that lets applications communicate by sending and receiving messages through queues or topics, asynchronously and reliably. Message-Oriented Middleware decouples applications in time and space: a sender places a message on a queue or topic and continues working, while the receiver consumes it when ready. The middleware guarantees delivery, ordering and buffering even if one side is temporarily unavailable. This underpins both classic integration and event-driven systems. Common implementations include JMS providers, IBM MQ, RabbitMQ and Apache Kafka. Read more: [What Is Enterprise Application Integration and Why It Matters in 2026](https://www.kondevs.com/content-hub/what-is-eai/) Related: Event-Driven Architecture (EDA), Enterprise Service Bus (ESB) ## Middleware **What is middleware?** Middleware is software that sits between applications and operating systems to enable communication, data management and integration across them. Middleware is the connective software layer that lets distributed applications and systems work together: handling messaging, data transformation, transactions, authentication and orchestration so individual applications don't have to. Integration middleware (ESBs, message brokers, integration platforms) is the category most relevant to enterprise integration, providing the plumbing that moves and reshapes data reliably between heterogeneous systems. Read more: [How we made kondevs.com usable by AI agents](https://www.kondevs.com/content-hub/how-we-made-kondevs-usable-by-ai-agents/) Related: Enterprise Service Bus (ESB), Message-Oriented Middleware (MOM), iPaaS (Integration Platform as a Service) ## MLOps (Machine Learning Operations) **What is MLOps?** MLOps applies DevOps-style practices to machine learning: versioning, automated training, deployment, monitoring and retraining of models in production. MLOps is the set of practices and tooling that take machine-learning models from notebooks to reliable production systems. It covers data and model versioning, reproducible training pipelines, automated deployment, and ongoing monitoring for performance and data drift so models can be retrained before they degrade. MLOps is what keeps AI solutions accurate, maintainable and trustworthy once they are integrated into business processes. Read more: [Software Dev Postings Rose 15%. The Job That Rebounded Isn't the One That Left.](https://www.kondevs.com/content-hub/software-dev-postings-rose-15-the-job-that-rebounded-isn-t-the-one-that-left/) Related: ETL (Extract, Transform, Load), Large Language Model (LLM) ## Model Context Protocol (MCP) **What is the Model Context Protocol (MCP)?** MCP is an open standard that lets AI models connect to external tools and data sources through a consistent, auditable interface. The Model Context Protocol standardises how AI agents and applications communicate with external tools, data and services. Instead of bespoke glue code for each integration, systems expose capabilities through MCP servers that any MCP-aware model can use over a common interface. For the enterprise this means agents can reach CRM, ERP and analytics through standardised, governed, auditable connections: bringing integration discipline to AI agent development. Read more: [LLM visibility: how AI assistants find, read and cite a company](https://www.kondevs.com/content-hub/llm-visibility-in-ai-answers/) Related: AI Agent, Retrieval-Augmented Generation (RAG) ## Multi-Agent System **What is a multi-agent system?** A multi-agent system is a group of specialised AI agents that divide a task between them and coordinate through messages, shared state or an orchestrator. Instead of one general agent doing everything, a multi-agent system gives each agent a role, such as planner, researcher, writer or reviewer, with its own instructions and tools. An orchestrator or a defined protocol routes work between them and merges the results. The pattern suits complex workflows and lets each agent hold only the permissions it needs, but it adds coordination overhead and new failure modes, so enterprise deployments need clear hand-offs, observability across agents and human approval for consequential actions. Related: AI Agent, Agentic RAG, Model Context Protocol (MCP) ## Process Mining **What is process mining?** Process mining analyses event-log data from systems to discover how processes actually run, revealing bottlenecks and deviations from the intended design. Process mining reconstructs real process behaviour from the event logs that enterprise systems already produce. Algorithms reveal the actual paths cases take (including variants, rework loops and bottlenecks) versus the designed 'happy path'. This evidence base lets organisations target improvements, predict SLA breaches and validate that automation delivers the intended outcomes. It is a natural complement to BPM and to AI-enhanced process intelligence. Read more: [How AI Is Transforming Business Process Management](https://www.kondevs.com/content-hub/ai-transforming-bpm/) Related: Business Process Management (BPM), MLOps (Machine Learning Operations) ## Retrieval-Augmented Generation (RAG) **What is Retrieval-Augmented Generation (RAG)?** RAG is a technique that grounds a language model's answers in retrieved, authoritative documents rather than relying solely on its training data. Retrieval-Augmented Generation improves the accuracy and currency of large language models by retrieving relevant content (from a knowledge base, document store or database) and supplying it to the model as context when generating an answer. This reduces hallucination, lets responses cite up-to-date and proprietary sources, and keeps sensitive data in systems you control. Agentic RAG extends this with planning, multi-source routing and iterative refinement. Read more: [LLM visibility: how AI assistants find, read and cite a company](https://www.kondevs.com/content-hub/llm-visibility-in-ai-answers/) Related: Agentic RAG, Large Language Model (LLM), AI Agent, Model Context Protocol (MCP) ## Service-Oriented Architecture (SOA) **What is Service-Oriented Architecture (SOA)?** SOA is an architectural style that structures software as reusable, loosely coupled services accessed over a network through defined interfaces. Service-Oriented Architecture decomposes capability into discrete services (each with a contract describing what it does) that other systems call without knowing the internal implementation. Loose coupling means services can evolve independently, and reuse reduces duplication across the enterprise. SOA is often realised with an enterprise service bus for mediation. Modern API-led and microservice approaches share its core principle of well-defined, reusable, independently deployable services. Read more: [What Is Enterprise Application Integration and Why It Matters in 2026](https://www.kondevs.com/content-hub/what-is-eai/) Related: Enterprise Service Bus (ESB), API-Led Connectivity