Every day, we see the power that great information management brings to our customers by connecting users to content and data and super-charging content-rich processes through intelligence and automation. The launch of brings the power of (LLMs) and generative AI (GenAI) to information management users. There’s no doubt that GenAI conversational interfaces have the potential to improve the information retrieval experience for users. At the same time, we must also take care to protect sensitive information and to deliver accurate information to the user. OpenText is widely recognized for bringing trusted and secure information governance through content services to customer processes, helping users get work done quickly and efficiently. Content Aviator is a natural extension of our mission to help organizations master modern work, by delivering relevant, accurate content and keeping users and organizations safe and compliant.
The biggest concerns we hear from customers when we talk about LLMs center on their accuracy, their use in decision-making, and the risks of exposing sensitive content to an AI model. Without careful attention and planning, Generative AI can produce overly general, biased, or outright incorrect responses (sometimes referred to as “hallucinations”). Delivering a valuable and secure experience to users involves a combination of processes and practices that, when done correctly, can dramatically improve user productivity without divulging privileged information.
Delivering a valuable and secure experience to users involves a combination of processes and practices that, when done correctly, can dramatically improve user productivity without divulging privileged information. Publicly available LLMs such as OpenAI’s ChatGPT and Google’s Bard are trained on huge, opaque datasets gathered from the internet. This makes these chatbots proficient at responding to a broad audience with generic language but leads to bias driven by the whims of modern culture and lacks important domain-specific information and proprietary advantages that are critical in the daily work of your users. Ultimately, general-purpose chatbots are of limited utility in the enterprise and are sometimes suspect in their guidance of users. Moreover, they lack the focus on enterprise content or combine public internet and private enterprise content with a lack of transparency if not guided appropriately. So, what can be done about that?
Benefits of the Content Aviator methodology
At OpenText, we tackle these issues head-on with Content Aviator and ensure a powerful GenAI experience delivered through a purpose-built, augmented approach that provides safe and relevant responses:
- Delivering relevant content with precise responses: Good decisions are based on the most relevant content. To achieve this, Content Aviator responds to users by gathering the most relevant documents from a business workspace for analysis through a process called Retrieval-Augmented Generation (RAG). As Business Workspaces represent real-world business entities such Customers, Employees, Products or Purchase Orders, Content Aviator’s responses are intrinsically domain-specific, producing concise, context-aware summaries, insights or answers to questions. All responses cite the source documents in the enterprise repository for reference or further investigation.
- Preventing over-sharing of content to honor security and privacy: A critical element of safely rolling out GenAI is to ensure that responses do not over-share information to which the user has no access or otherwise leak privileged or private information in error. As a first step, content must be effectively protected through proper classification and monitoring. Content Aviator proactively filters for documents available to the user, delivering responses based only on permissible content to prevent data leaks or disclosure of sensitive data.
- Responding with proper context and semantics: Content Aviator is designed to respond in contextually relevant language through “semantic search.” This process connects the user’s prompt with the semantically pertinent documents that relate to the context of their prompt within a business workspace. By going beyond simple keyword matching to semantic analysis, responses are tied to the user’s intent and business context. For example, the term “bandwidth” can relate to network bandwidth or the capacity of a team; through semantic search, the most appropriate responses can be derived from a match of semantic meaning. In addition, OpenText’s deep integration into business applications such as SAP or Salesforce allows Content Aviator to feed the right combination of relevant content and relevant business data to the LLM. This makes Content Aviator unique.
Sound governance principles deliver a foundation for Content Aviator
To get the most out of GenAI investments, organizations must start with robust and secure content services that build on sound information governance principles to deliver GenAI that is both productive and safe.
To get the most out of GenAI investments, organizations must start with robust and secure content services that build on sound information governance principles to deliver GenAI that is both productive and safe. OpenText content services provide critical information governance capabilities upon which we build the Content Aviator experience:
- Business workspace: OpenText content services are best for managing content in context through integrations with line-of-business applications (such as ERP, CRM, or HR systems) and automated business processes. The business workspace encapsulates and organizes this context elegantly and provides a natural nexus of context, automation, and retrieval by establishing repeatable classification, information protection and highly accurate business data. The business workspace forms the perfect domain for feeding GenAI prompt engineering and delivering valuable responses.
- Reduction of content sprawl: OpenText content services are specifically designed to help organizations battle content sprawl. Convenient access to a shared content services library and capturing, classifying and scheduling collaborative content in platforms such as Microsoft Teams are ways OpenText controls content sprawl, leading to better curation of generated responses.
- Security classification: Properly classified content incorporates access permissions and security markings that control who should see the content. The automated application of proper classification, permissions and security markings means that responses are accurately filtered, protecting content from inadvertent disclosure.
- Timely disposition: Incorporating a consistent records policy across content services affords the organization with timely disposition. This assures that expired, irrelevant, and potentially risky content is removed as quickly as policies allow.
What creates great AI? Great information management. Content Aviator builds on a unique business-oriented content structure to infuse GenAI with relevant and safe domain content to connect users quickly with the information they need and are entitled to. Content Aviator will be available for each of the OpenText content services platforms: Extended ECM, Documentum and Core Content. Are you ready to amplify user productivity with GenAI? Get started with Content Aviator today.