FAQ for New Users
What can I do with this AI system? The AIR AI system can do almost anything you would do on a public AI system, like ChatGPT, but securely and with access to your organization’s data and systems. It can connect to email, calendars, Teams, and your private data and documents. Users can leverage built-in AI agents for tasks such as meeting transcript summaries, regulatory document analysis, compliance checks, and customer communication. Additionally, the platform supports custom AI agent workflow development, which can be tailored to enhance specific workflows and solve unique business challenges. Note that the AI should not be used for personal purposes, as your chats will be accessible to admins and the compliance team for audit and compliance purposes.
Is it safe to use confidential or personal information with this AI? Yes, using confidential or personal information with the AIR AI system is safe, as the platform is built with robust data security and privacy measures. All data interactions are governed by stringent compliance protocols, including encryption at rest and in motion, role-based access control, and multi-factor authentication. The platform supports tokenization techniques that replace sensitive data with unique identifiers, ensuring privacy protection. Furthermore, AIR does not access or store personally identifiable information (PII) outside the client's secure environment, complying with regulations such as GLBA, GDPR and CCPA.
Why does the public version of AI give a different or longer answer than the AIR system? There are constraints on users in the AIR system versus a public AI system for organizational safety and cybersecurity reasons. The AIR system is designed specifically for secure and compliant use in high-risk environments such as the finance and healthcare sectors. AIR employs well-defined AI agents using Retrieval-Augmented Generation (RAG), which allows for highly accurate and contextually relevant outputs by integrating retrieved data into the generation process. This means that the AIR system is tailored to provide concise and precise responses that align with regulatory standards and enterprise needs. In contrast, public versions of AI, like Copilot, prioritize broad applicability and user engagement, which may result in more verbose responses as they are not bound by the same compliance and precision constraints.
What role does the AIR administrator at my organization have? The administrator in the AIR AI platform plays a crucial role in managing and overseeing the system's operations. Administrators invite new users, create groups, control and monitor user activities, manage permissions, create or add Ai agents, make cybersecurity settings, and ensure that security protocols are adhered to.
How does the AIR AI platform help with regulatory compliance? The AIR AI platform is designed with a comprehensive governance, risk management, compliance, and cybersecurity (AI GRCC) layer which ensures adherence to regulatory requirements. This includes tools for auditing and monitoring, which help firms comply with regulations like GDPR, CCPA, and the NIST AI RMF. The platform facilitates capturing and maintaining records of AI interactions in an immutable database, vital for meeting compliance standards in sectors like finance.
What AI models can I use? Administrators of the AIR platform can integrate a variety of AI models, including popular ones such as Azure Open AI (ChatGPT). The platform supports both proprietary and open-source models. Additionally, clients can deploy these models on their private cloud infrastructure, ensuring a high level of data control and security. It is also important to note that the large AI companies release new models first to consumers on the internet, then later release the tested version for corporate use on the AIR AI system. In addition, the AIR team does its own AI model testing before releasing them to our clients.
How often is the AIR AI system and the AI models updated? The AIR system is typically updated monthly. Updates include enhancements to the AIR software itself as well as access to new AI models once they have been thoroughly tested. This process requires a server administrator on the client side to initiate these updates to maintain control over the deployment and ensure compatibility with the client's existing systems.
How can I upload a document? Documents can be uploaded to the AIR platform through its user-friendly interface, which supports integration with various data sources and applications. Users can connect their systems and data using APIs or direct uploads, ensuring the AI models have access to the necessary information while maintaining security through encryption and governance protocols.
How many documents can I upload? Can I point to a folder instead? You can upload up to 20 documents. Your administrator can make available the ability to upload unlimited documents if you ask. In addition, depending on the system, you can have an AI agent point to a specific folder or document storage area. This feature needs to be configured by your administrator.
How can I generate a graphic with AIR? To generate a graphic with the AIR system, you would typically utilize one of the built-in AI agents for images. You may need to ask your administrator for access.
Does the AI have access to my email or calendar? The AI platform includes built-in agents that can interact with tools like email and calendar applications, offering features such as meeting summaries and personalized email responses. However, access is governed by strict permissions set by the user or administrator, ensuring that the AI only accesses data that it is authorized to handle.
What if I have an idea for an AI agent? If you have an idea for a custom AI agent, AIR provides support to help bring that idea to life. The platform allows for the creation of custom AI agents tailored to specific workflows or business needs. AIR works closely with clients and your administrators, often forming focus groups to develop new agents that enhance operational efficiency and solve domain-specific challenges.
My answer from AI does not look right. What should I do? Start with a thumbs-down to the AI’s response and enter a comment so the team is aware there was an issue. AI large language models (LLMs) are a new technology. They can still make mistakes or “hallucinate” so their results should be checked for accuracy. It is very important to provide the AI with access to the information to answer a question. For example, if you are asking a question about a client, the AI must have access to that data or it cannot give a correct answer. If an AI response seems incorrect, it is recommended to review the input data and the context provided, as these directly impact the AI's output. You can also reach out to the support team for assistance. Ensuring the AI is accessing the correct and intended data sources can often resolve discrepancies.
I really need a longer answer from the AI. How can I get that? To obtain more detailed responses from the AI, consider providing more comprehensive input and context, ensuring the AI has enough information to work from. Additionally, you can ask the administrator to adjust the settings or parameters within the AI model to allow for more expansive answers.
What other built-in AI agents are available? AIR offers a variety of built-in AI agents designed to address different business needs from day one. These include agents for meeting summaries, risk management, compliance, research, marketing, and more. These agents can be customized further to align with specific organizational requirements, providing versatility and immediate utility.
Can software engineers access the AIR AI platform programmatically? Yes, software engineers can access the AIR AI platform programmatically via APIs provided by the system. This allows them to integrate custom solutions and extend the platform's functionality to meet specific business requirements. The platform supports a wide range of integrations with third-party applications, facilitating comprehensive use-case implementations.
What data and software can we integrate with? The platform can integrate with virtually any SQL database, API (programmatic interface), or MCP (model context protocol) connection securely. One of your AIR admins can do this or coordinate with the AIR support team to get this set up. Note that AIR prioritizes privacy and security for these connections, so initial safety and security testing may take some time for a new connection for your company.
How do I learn more about the AIR AI platform? Learning more about the AIR AI platform can be accomplished by accessing the documentation provided to users, which includes detailed guides on platform features, best practices, and troubleshooting tips. Engaging with training sessions offered by AIR or reaching out to support for personalized guidance can also enhance understanding of the platform. You can view the documentation at https://docs.aicrisk.com/ and our private instructional video library for the AIR system at AIR YouTube How To
What if I get an error message? Please reach out to your company admin for explanations or to get permissions or AI agents added. The most typical errors involve uploading more than 20 documents or third-party AI model issues that can be easily resolved. The AIR AI platform automatically blocks toxic language, cybersecurity attacks on Gen AI, etc. For additional support, reach out to [email protected].
Is there more research or white papers I can read? See here for a number of papers including: A Risk Management Framework for Large Language Models
Other sources of information on AI? Read our Substack blog at AI Risk Reward, our monthly AI Newsletter on LinkedIn and listen to our weekly AI podcast here: AI Risk Reward.
Last updated