Questions and glossary
Glossary
22 terms that appear consistently throughout the product and this knowledge base.
Stable term names used consistently across the knowledge base, the product and Slimaco AI materials.
| Term | Definition |
|---|---|
| Agent | A specialized task executor created from a plain-language description; runs on demand or on a schedule, within permissions and approval rules. |
| Scheduled action | Running an agent on a schedule, e.g. every morning or every Monday. |
| Knowledge mining | Picking up agreements, decisions and changes in how work is done from day-to-day communication and sending them for review. |
| Assistant | A configurable AI role with a persona, knowledge, tools and visibility. |
| Slimaco assistant | The main conversation and work interface between the user and the platform. Formerly called Copilot. |
| Skill | A repeatable instruction or runbook used in a controlled task execution. |
| Company knowledge (Knowledge Core) | The layer of sources, documents, versions, approval, indexing and access that the assistant and agents work on. |
| RAG | Retrieving relevant knowledge fragments before generating an answer. |
| Evidence | Proof drawn from a knowledge source that an answer or conclusion is based on. |
| Citation | User-visible information about the source supporting a fragment of an answer. |
| HITL | Human in the loop; mandatory human involvement in an approval or decision. |
| Review gate | A point where a process is paused for human evaluation. |
| Write guard | A policy requiring consent before an action that writes data or triggers an external effect. |
| MCP (Model Context Protocol) | An open standard for connecting AI assistants to external data sources and tools. Via MCP, ChatGPT, Claude, Gemini and Microsoft Copilot will be able to use Slimaco knowledge (coming soon). |
| Hard data | Figures and statuses from department systems, such as the number of a customer's open tickets; they update in Slimaco continuously, without review. |
| Workflow | A versioned sequence of work with steps, tools, decisions and review gates. |
| Artifact | A structured work output, e.g. a draft, report, recommendation or analysis card. |
| Entitlement | Whether a feature is available under a client's plan or configuration. |
| BYO AI | A mode using a client's own agreed model access, or a separate model billing arrangement. |
| Golden Set | A representative set of questions, answers and criteria used to evaluate quality. |
| No-answer | A controlled response indicating there isn't enough evidence to answer. |
| Shadow AI | Uncontrolled use of AI tools outside the organization's rules and visibility. |
Last updated: 2026-10-14 · Content owner: Slimaco AI team