How Search Works
An overview of how the Moveworks AI Assistant retrieves and answers from enterprise content, across both indexed and live search.
Moveworks surfaces content to employees through two mechanisms: indexed search, where content is ingested and stored in advance so it can be retrieved instantly at query time, and live search, where results are fetched directly from the source system in real time on behalf of the user.
Both follow the same three-step pattern from the AI Assistant’s perspective. The difference is only where the content lives at query time.
The Enterprise Search Plugin handles every search the assistant performs, whether it’s indexed or live. Each search runs through:
Plugin selection — The Moveworks reasoning engine determines which search plugin to invoke (knowledge search, file search, or a live search plugin) based on the employee’s query. If no plugin is relevant, no results are returned.
Retrieval and ranking — The query is evaluated against the relevant content. For indexed search, this is a combination of semantic and keyword search against pre-indexed content, with query rephrasing applied to expand and anchor the query in context. For live search, the source system’s own API is queried in real time using the signed-in user’s credentials. The top results are ranked by relevance; only results above a predefined score threshold are returned.
Summarization — The reasoning engine produces a summarized response grounded in the top-ranked results, including citations that link back to the source article or file. Not all retrieved results are used in the response — only the ones needed to answer the question.
If no results meet the relevance threshold, the reasoning engine may try a different plugin or offer the employee a handoff to file a ticket.
Both indexed and live search respect source system permissions at query time. Employees only see results for content they have access to in the source system. For indexed file search specifically, Moveworks continuously syncs file permissions alongside content so that access always reflects the current state of the source. For live search, the source system enforces permissions directly using the signed-in user’s credentials.
For more, see Permissions.
Content types, supported sources, ingestion behavior, and FAQ for pre-indexed search across knowledge base articles and files.
How Moveworks performs real-time searches against source systems using the signed-in user’s own credentials.
Reference table showing how frequently Moveworks ingests and indexes each content type, including best-case and worst-case availability windows after a source update.