AUM: A private enterprise AI knowledge base client running locally offline

AUM is an AI intelligence engine designed for enterprises, officially called the "Enterprise Brain". It runs locally on the user's own server or computer, without an internet connection, ensuring absolute privacy and security of business data. Users can let AUM process and analyze all kinds of files, reports and databases within the company through conversations to quickly find information, generate documents or analyze data. This system addresses the risks of data leakage, non-compliance with regulations (e.g. HIPAA, GDPR) and loss of control of intellectual property that can result from using cloud-based AI. AUM supports access to a wide range of mainstream open source AI models, allowing organizations to choose and deploy according to their needs, for example on fully physically isolated servers. It creates proprietary AI brains for different teams (e.g. HR or Legal) to help them with specific operations, such as quickly screening resumes or analyzing contracts.

Function List

• Fully localized deployment: Supports installation and operation on on-premises servers, private clouds, or completely off-network (air-gap) devices, ensuring that data does not leave the organization's premises.
• Enterprise File Search: Ability to quickly retrieve millions of internal and private documents and quickly locate information.
• Multi-document co-processing: It is possible to process a large number of documents at the same time, such as analyzing hundreds of resumes or legal documents at once, and to complete specific tasks as instructed.
• Database dialog analysis: Users can ask questions directly to the database in everyday language and analyze data without writing complex query statements.
• Customized AI Brain: Allows companies to create specialized AI models for different departments (e.g., HR, legal, operations) to make them more knowledgeable about specific business processes.
• Workflow automation: The ability to automatically create documents, draft proposals, send emails, or write content to other applications such as Notion, based on instructions, enhances productivity.
• Support mainstream open source models: The system is compatible with a number of industry-leading open source AI models, including Llama3, Mistral, Mixtral, DeepSeek, etc., which organizations are free to choose from.
• privilege management system: Built-in role-based access control allows fine-grained management of access rights for different users.

Using Help

AUM is an AI tool that runs on a local computer or server, and at its core, it allows organizations to use AI to analyze and process sensitive data internally without sending it to an external cloud.

Deployment

AUM offers a variety of deployment options to suit the security needs of different organizations, and the entire process does not require a connection to the public Internet.

1. On-Premise Servers
   • This is the most common deployment method. The organization needs to prepare a server hardware with sufficient performance.
   • Install the AUM software directly on this server. After the installation is complete, all employees on the company's intranet can use AUM by accessing the address of this server through a browser.
   • All data processing and AI modeling is done on this server, and the data is kept entirely within the company's firewall.
2. Private Cloud / VPC Deployment
   • If an organization is already using a cloud service such as Aliyun, Tencent Cloud, AWS or Azure and has its own virtual private cloud (VPC), it can deploy AUM on the cloud servers inside the VPC.
   • This approach also ensures data isolation because the VPC is a secure environment isolated from the public network and accessible only to authorized personnel of the enterprise.
3. Physical Isolation Deployment (Air-Gapped Infrastructure)
   • It is an option for organizations with very high security requirements, such as government, military or core financial sectors.
   • Install AUM on a stand-alone computer or server that is not connected to any external network at all (including the company intranet).
   • Users can only operate this device by physically touching it, the highest level of security.
4. Multi-Device Local Deployment
   • The upcoming feature allows for distributed deployment of AUM on multiple local devices that work together.

Core Function Operation

Getting started with AUM is done primarily through its conversational interface, just like talking to a chatbot.

1. Questions and searches for documents
No more manually opening folders one by one when you need to find information from a huge amount of company documents.

• workflow:
  • Open the main screen of AUM.
  • Type your question directly into the dialog box. For example, type, "Help me find all the contracts for the last year for Program A," or "Which product had the highest sales in the financial report for the third quarter?"
  • AUM immediately searches for all relevant documents and gives you the answer directly, rather than just listing the documents.

2. Simultaneous processing of multiple documents
Let's say you're HR and you've received over 100 resumes and need to screen them for candidates who meet specific requirements.

• workflow:
  • Upload or assign all resume files to the AUM workspace.
  • Give instructions in a dialog box, e.g., "From these resumes, identify all the candidates who have more than 5 years of experience in Java development and who graduated from a 985 university."
  • AUM reads all resumes and then compiles a list of qualified candidates.
  • Similarly, legal teams can use it to compare differences in the terms of multiple contracts at the same time.

3. Dialogue with databases
If you are a marketing manager and you want to know the recent sales data but you don't know how to write SQL query statements.

• workflow:
  • First, a technician plugs the company's sales database into AUM.
  • After that, you can ask questions in the dialog box in the same way you would in everyday speech. For example, "Which channel's promotion brought in the most new users last month?" or "Help me generate a chart showing the sales trend for each product over the last six months."
  • AUM translates your words into query commands that the machine understands, then finds the data from the database and presents it to you in an easy-to-understand way.

4. Creating customized "AI brains"
Workflows and expertise vary from department to department.AUM allows for the creation of exclusive AI for specific teams.

• workflow:
  • The administrator goes to the backend settings and selects "Create New Brain".
  • Upload information specific to the department, e.g. for the "HR Brain", you can upload the company's Employee Handbook, Recruitment Process, Performance Standards and other documents.
  • Once the setup is complete, HR employees can ask the dedicated brain, "What's the onboarding process for new hires?" or "Help me draft a job description for a Java engineer". This AI will answer based on the internal information you provide, ensuring the answer is accurate and compliant.

application scenario

1. law firm
   When handling cases, lawyers need to read and analyze a large number of contracts, evidence and legal documents. With AUM, they can quickly retrieve key clauses from thousands of documents, discover contractual loopholes or conduct due diligence, and the entire process is done in-house, completely eliminating the risk of leaking sensitive client information.
2. medical research organization
   Patient data and clinical trial reports handled by hospitals and research institutions are strictly protected by regulations such as HIPAA.AUM can be deployed on hospitals' internal servers to help researchers analyze massive amounts of anonymized medical record data, retrieve the latest medical literature and clinical knowledge, and accelerate the research process without violating regulations.
3. financial institution
   Investment banks and research departments need to analyze large volumes of market reports, company earnings reports and compliance documents to make decisions. Using AUM, analysts can securely parse these internal documents for investment research and risk assessment without having to worry about exposing confidential trading information to any third-party cloud service.
4. Government and defense sector
   AUM can be deployed in an "air-gap" network that is completely physically isolated from the outside world for intelligence retrieval, policy research, and internal knowledge management to ensure the security of state secrets, which is a top priority for government departments that deal with classified intelligence and confidential policies.
5. Internal knowledge base of a large enterprise
   When a large company's internal processes, technical documentation, and board reports are scattered across systems, it can be difficult for employees to find the information.AUM can integrate this private data to build a unified internal knowledge center of intelligence. Employees can quickly access operational data, inquire about company policies, or get technical support through conversations.

QA

1. What is the fundamental difference between AUM and cloud-based AI tools like ChatGPT?
   The most fundamental difference is data privacy and control. Cloud-based AIs such as ChatGPT process your data, which carries data leakage and compliance risks for organizations that contain sensitive information. AUM, on the other hand, is deployed entirely within the organization's own hardware facilities, and the data never leaves the organization from start to finish, ensuring data sovereignty and security for 100%.
2. What are the hardware requirements for running AI models on AUM?
   AUM supports high-performance systems, allowing organizations to unlock more advanced and powerful versions of AI models for faster response times if their hardware is powerful enough (e.g., has free compute resources). It allows organizations to leverage their hardware resources to run the latest AI technologies.
3. What AI models does AUM support?
   AUM supports a wide range of mainstream open source large language models such as Mistral, DeepSeek, Llama3, Mixtral, etc. This gives organizations the flexibility to choose the most appropriate model for their needs and hardware conditions. This gives organizations the flexibility to choose the most suitable model according to their needs and hardware conditions.
4. Does AUM offer a free version?
   Yes, the official website mentions that there is a "perpetually free" version available that does not restrict access to the AI.