| AI Agent |
AG |
An autonomous AI entity capable of goal-directed reasoning, planning, and action. Unlike a simple chatbot or scripted automation, an agent perceives its environment, evaluates options, and selects actions to achieve a specified objective, with or without continuous human direction. It is the primary actor that owns the business goal and drives the solution forward. |
| **AI Orchestrator (Coordinator) |
AC |
The coordination logic, workflow control, and multi-agent management layer that sequences and routes AI operations. The orchestrator does not reason about the domain problem itself; it governs how and when other elements are engaged, breaking down goals into tasks, managing dependencies, handling failures, and ensuring that outputs from one step correctly feed into the next. This is where deterministic process logic is enforced. |
| **Context Management (Memory State) |
CN |
The conversational and memory state, prompt engineering, and interaction relevancy layer. Context Management ensures the model is always operating with the right frame of reference, preserving prior exchanges, user preferences, and session history, and filtering what is surfaced to the model at each step. Without it, even a powerful model operates as if every interaction is its first. |
| AI Model (Reasoning) |
ML |
The models, inference engines, and reasoning frameworks that form the cognitive core of the solution. This is the probabilistic layer where language understanding, inference, and generation occur. Model selection, including size, fine-tuning, and whether a proprietary or open model is appropriate, is an architectural decision that must be made in concert with the other TOCKAM elements, not in isolation. |
| Knowledge Service (Access and Config) |
KA |
The semantic retrieval, RAG, embedding, and knowledge management capabilities that connect the model to an organization’s own data and domain expertise. Knowledge Services reduce model hallucination and generalization by grounding responses in curated, organization-specific content (procurement policies, technical documentation, regulatory guidelines) rather than relying solely on what the model learned during training. |
| Autonomous Tools |
TL |
External functions, services, plugins, and third-party capabilities that extend AI into the operational environment. Tools are the integration boundary between the AI solution and the real world, invoking APIs, querying databases, triggering workflows, and producing artifacts. They convert AI reasoning into concrete, executable outcomes and are typically the most deterministic component in the TOCKAM stack. |