MARA: An AI Assistant for Enterprise Knowledge Discovery

Idea

The idea for MARA came from a challenge we experienced firsthand at HQSoftware. As our team grew, onboarding new employees and helping them navigate the company’s internal knowledge became increasingly demanding. Important information was scattered across different sources, including internal documentation, shared folders, and collaboration platforms. Finding the right answer often meant searching multiple systems or asking colleagues for help.

We realized that this was not just an onboarding challenge, but a broader knowledge-sharing problem. Employees needed a simpler way to access company information, understand internal processes, and find reliable answers without knowing exactly where to look.

This led us to the idea of building an AI-powered assistant that could bring knowledge from different sources into one conversational interface. Instead of manually searching through repositories, employees could simply ask a question and receive an answer based on relevant company documents, with links to the original sources.

We developed MARA to address a challenge we knew firsthand: making company knowledge easier to find, access, and share across the organization.

Challenge

Connecting a chatbot to a language model is not enough to make company knowledge accessible. A general-purpose model knows nothing about internal documents, cannot tell where its answers come from, and has no idea who is allowed to see what. Documents also live in different systems and formats, and many companies do not want to depend on a single AI vendor.

We set the following requirements for the product:

• One conversational interface over all connected knowledge sources
• Answers grounded in company documents, with links to the sources
• Access control: employees get answers only from documents they are allowed to see
• Freedom to choose and switch the LLM provider
• An admin panel that lets the company manage the system without involving developers

Solution

MARA is an AI assistant that enables employees to access organizational knowledge through a single conversational interface. It integrates with multiple data sources, consolidating their content into a shared knowledge base. Employees can ask questions in natural language and receive answers generated by a Large Language Model (LLM) based on information retrieved from connected sources.

The platform also includes an administrative interface for configuring integrations and AI models, managing users and access permissions, and monitoring knowledge base operations. This allows organizations to provide centralized access to internal knowledge while maintaining control over system configuration and information access.

User Workflow

1. Access MARA — The employee logs into MARA and opens the chat interface.
2. Enter a question — The employee types a question in natural language about company processes, internal documentation, or other organizational knowledge.
3. Retrieve relevant information — MARA searches the shared knowledge base for relevant information from connected sources and processes the query using the configured LLM. 
4. Receive an answer with source references — MARA generates a response based on the retrieved information and provides links to the original sources, allowing employees to verify the answer and explore the relevant documents.
5. Continue the conversation — The employee can ask follow-up questions to clarify the answer or explore the topic in greater detail.

Key Capabilities

AI-Powered Knowledge Discovery

MARA enables employees to ask questions in natural language instead of manually searching through multiple document repositories. It retrieves relevant content from the connected knowledge base and generates answers using an LLM.

Answers with Source References

MARA links AI-generated answers to their original documents, helping employees check the underlying information and review the source context.

Role-Based Access Control

Administrators create roles, assign them to employees, and attach roles to folders in MARA’s built-in document library. Employees get answers only from the folders their role can access, so restricted materials such as department-specific documents stay private. Sign-in can be limited to corporate email domains, and users can be invited individually or imported from a CSV file with roles already assigned.

Unified Knowledge Index

Documents from all connected sources are processed and indexed into a shared knowledge base, so a single question can retrieve information from different repositories at once. 

Configurable LLM Providers

Administrators can select an LLM provider and model according to their requirements. Supported providers include DeepSeek, Anthropic, Google Gemini, OpenAI, and OpenRouter.

Document Processing and OCR

MARA processes documents before indexing them for knowledge retrieval. OCR is used to extract text from supported document formats, including PDF and DOCX.

Sensitive Data Anonymization

Administrators can enable sensitive data anonymization when using external LLM APIs to help protect confidential information.

Suggested Questions

Administrators can configure suggested questions displayed when employees open a new chat. Questions can be made available to all users or assigned to specific roles, helping employees find relevant information more easily.

Looking for similar solutions or something unique to your needs?

Whether you have a clear vision or just an idea, our team is ready to help you explore the possibilities and turn your goals into reality.

Sofia Linden

Business Development Manager

Get in touch

Data Sources

MARA connects to several document repositories and indexes them into one shared knowledge base: 

• Google Drive
• Confluence
• Amazon S3
• Built-in document library (local folders) 

Administration

MARA includes a dedicated administrative panel for configuring the platform and managing its knowledge infrastructure.

Administrators can:

• Configure LLM providers and models.
• Connect and synchronize data sources.
• Review document indexing status and errors.
• Create roles and assign them to users and document folders 
• Import users through CSV files and restrict sign-in to specified email domains.
• Configure access permissions for local document folders.
• Set up suggested questions for new chats.
• Publish announcements and system notifications for individual users, specific roles, or all users.
• Monitor background tasks and connector operations.
• Review LLM token consumption by day, user, and model.

Additionally, administrators can enable sensitive data anonymization when using external LLM APIs to help protect confidential information.

These capabilities provide visibility into the knowledge base and the processes that keep the platform running.

Use Cases

MARA can support knowledge discovery across multiple departments:

• Human Resources: Find internal policies, vacation rules, benefits information, and HR procedures.
• Employee Onboarding: Help new employees discover internal processes and documentation.
• IT and Engineering: Retrieve technical documentation, internal guides, and development knowledge.
• Sales: Find sales playbooks, product materials, and internal documentation.
• Legal: Locate internal policies and document templates.
• Finance: Access financial procedures and internal guidelines.
• Production and Operations: Find operating procedures and work instructions.
• Customer Support: Retrieve relevant internal knowledge to support consistent responses.

Team

• Project manager
• ML engineer
• Fullstack developer
• QA engineer

Bring MARA to Your Company

Looking to help employees find answers in company knowledge without searching across multiple systems? MARA brings enterprise documents into a unified AI-powered knowledge experience, with source references, administrative controls, and configurable LLM integrations.

Contact HQSoftware to discuss your knowledge discovery requirements and potential use cases.