Enterprise Knowledge Discovery & AI Assistance

Useful enterprise AI starts with finding the right information and establishing who may use it. Documents may be distributed across content-management systems, file shares and other repositories, with different owners, formats and permissions.

Understand the product roles

OpenText Knowledge Discovery, associated with the IDOL technology lineage, supports enterprise search and analysis across diverse content sources. Connector coverage and document processing should be checked against the actual repositories and files in scope. [6]

OpenText Content Aviator provides generative AI assistance within supported OpenText content-management environments, including conversational access and summarisation. “Aviator” describes a family of offerings; Content Aviator, Knowledge Discovery and other Aviator products should not be treated as one interchangeable package. [7]

Potential applications in shipping

  • Technical knowledge: locate relevant manuals, service bulletins and previous maintenance reports for a particular equipment model.
  • Fleet procedures: find the approved procedure and distinguish it from superseded versions.
  • Procurement support: retrieve specifications and prior correspondence associated with an order or component.
  • Commercial knowledge: find relevant clauses and supporting correspondence, subject to the user’s permissions and professional review.

These are proposed use cases, not completed PT customer references. A technical assistant should cite the source and version, identify missing evidence and leave operational decisions with authorised personnel.

Permissions must be designed and tested

Knowledge Discovery provides mechanisms for source-aware document security, but connectors, identities, group membership and query security require configuration. [8] Test both permitted access and denied access, including what happens after a document is deleted or a user loses permission.

Keeping original files in place does not mean no information is copied or transmitted. Assess indexes, extracted text, prompts, generated responses, logs and backups. Hosting and model endpoints must be evaluated for the selected product configuration.

What a pilot should prove

Choose representative repositories, users and questions. Test exact part numbers as well as natural-language questions; include scanned files, tables, obsolete documents and conflicting versions. Measure retrieval relevance, source accuracy, permission enforcement, update delays and response time.

Start with a knowledge-readiness assessment.
Identify the information sources, access rules and decisions the assistant is intended to support. The result should be a defined pilot and a realistic operating model.