Securing GenAI boons with content management best practices
Generative AI (GenAI), though ripe with possibilities, depends on a robust, accurate, and refined content strategy. Putting garbage information in only yields garbage out of AI, as the Deloitte saying goes—and that begins by addressing siloed, scattered content estates preventing GenAI value.
KMWorld’s latest webinar, GenAI Success Begins with Content: 5 Strategies For Accuracy & Precision, gathered experts from Coveo to examine the ways to curate, optimize, and index content to best support GenAI projects within the enterprise.
According to Patricia Petit Liang, product marketing manager, Coveo, GenAI relies heavily on content strategies, especially in the fight against AI-powered misinformation.
“Content cleanliness has never been more important,” said Petit Liang. “We’re seeing a lot of unforeseen results from generative answering in the form of exposing hidden gaps in content management.”
A strong content strategy is the best defense against unreliable outputs, ultimately securing the performance and integrity of GenAI implementations, explained Petit Liang. Yet, simultaneously, “boiling the ocean” is an unsustainable content approach—so organizations must employ an intentional, governed method of content optimization.
Mathieu Lavoie-Sabourin, product manager, Coveo, offered the following content best practices based on its customer success stories:
- Indexing security permissions: Ensures users only get answers they are allowed to see by indexing item-level permissions directly from content repositories.
- Filtering: Index the relevant sections in your content and exclude irrelevant parts.
- Optimization: Make sure that content items are properly titled, contextualized, and have clear text structure to improve LLM efficiency and efficacy.
- Categorization: Normalize key fields across documents and group and link content based on meaningful relationships to enhance retrieval consistency and accuracy.
- Relevance: Only use the most accurate, fresh, trusted, and meaningful content to generate answers; not all searchable content should be used for answers!
Outside of these strategies, Lavoie-Sabourin emphasized the importance of continuous improvement for content management.
“We need to understand what LLMs have been using to answer, and help identify, perhaps, gaps, because it’s hard to know all those gaps in advance—and even documentation may sometimes become outdated,” said Lavoie-Sabourin.
As part of this continuous improvement content approach, David Atallah, product manager, Coveo, introduced Coveo’s Knowledge Hub, a component of the Coveo Relevance Generative Answering solution. The new Knowledge Hub is designed to increase the control and transparency over the content being utilized by GenAI models, enabling users to:
- Assess answers and content quality.
- Trace the source of generated answers.
- Ensure outputs are accurate, compliant with enterprise policies, and aligned with regulatory standards.
Enterprises implementing Coveo’s GenAI solution have already seen the benefits of having a continuous content strategy, including increased self-service success, reduction in annual cost-to-serve, and improved case deflection rate in support use cases.
For the full, in-depth webinar featuring detailed explanations, a Q&A, and more, you can view an archived version of the webinar here.