AI Knowledge Access with MBOT

Controlled access to organisational knowledge through a conversational assistant

MBOT provides a practical approach to helping employees find answers in organisational documents while retaining control over content, access and hosting. This illustrative case study examines its application in a document-intensive organisation; it does not represent a verified client deployment.

The Business Challenge

Policies, operating procedures and software guides are often spread across documents and departmental repositories. Staff must locate the right file, identify the current version and interpret it before acting. Routine questions then reach subject experts and support teams, adding avoidable effort and delaying access to information.

For enterprise and government environments, the requirement extends beyond convenient search. Access must reflect user responsibilities, sensitive material must remain controlled, and administrators need a clear record of which documents are available to the assistant.

The MBOT Approach

MBOT brings approved organisational content into a conversational interface. Users type or speak a question; the application checks their role, retrieves relevant passages from accessible documents and uses a configured language model to compose a response. This retrieval-based approach is designed to ground answers in the organisation’s own knowledge.

Expected Operational Value

The intended benefits are less manual document searching, more consistent access to current guidance and reduced dependence on specialists for routine queries. These benefits should be validated through a pilot using representative users, approved content and agreed success measures.

Architecture and Deployment

The documented design combines a browser interface, role-aware retrieval and organisation-managed knowledge stores. Deployment choices should reflect information sensitivity and the intended user group.

ComponentFunction
Angular web applicationProvides chat and administration screens for documents, users and access rights.
Python application layerProcesses uploaded files, creates embeddings, coordinates retrieval and exposes REST APIs.
SQL databaseStores users, roles, document versions, active status and the audit trail.
Chroma vector databaseRetrieves relevant passages from the available knowledge base.
Configured language modelComposes answers from retrieved content; supports local or cloud hosting options.
Hosting and information control

MBOT’s documented design stores documents, embeddings and chat history on premises. A locally hosted language model supports offline operation. If a cloud model is selected, the information passed to that model and its handling must be reviewed before deployment. On-premises storage alone does not establish that all processing remains within the organisation.

Relevant enterprise and government applications

Enterprise teams can apply the approach to HR policies, standard operating procedures, product documentation and ERP user guidance. Government departments can evaluate it for internal circulars, administrative procedures, staff training and departmental manuals. Government applications are suggested use cases, not documented deployments.

An initial employee-facing scope provides a focused way to assess usefulness and access controls. Citizen-facing services, transaction processing and access to live operational records would require separately agreed content, integrations and validation.

Proposed COMnet engagement

A proposed engagement would begin with use-case discovery and content classification, followed by hosting design, access configuration and a controlled pilot. Pilot findings would inform integration, user training, support arrangements and phased expansion. The delivery scope and responsibilities should be agreed for each engagement.

A focused next step

Speak to COMnet Solutions about an MBOT demonstration using a selected set of approved documents. Agree the user group, hosting model and pilot success criteria before wider adoption.