Below are some key points to consider when working with content created by Generative AI, which will help you optimize its use and ensure that the responses from your virtual assistant across different support channels are useful and accurate.
- Structure the content: The content should be logical and well-organized to ensure that the AI can process it adequately. This means using clear titles and comprehensive descriptions.
- Differentiate similar content: To maintain topic coherence in your personalized knowledge base, avoid duplicating content. If similar topics are present in your personalized content, make sure that the title and the content are distinct enough for the AI to manage them correctly.
- Use of Gen AI and multibots: If using Gen AI to rewrite content, it is recommended to do so in a separate bot, especially if you have the multibot option. This will help prevent the main bot from being affected by any errors that occur while optimizing the content.
- Avoid sensitive information in GEN AI: Since AI can make mistakes, it is advisable to carefully train any information that may represent sensitive or critical data in the personalized engine with the appropriate review and approval process.
- Tables, graphs, and images: Although AI can read these formats, it is recommended to review the responses generated to ensure it has correctly interpreted the information. This ensures that the content is clear, accurate, and actionable, avoiding possible conceptual errors in automated responses. If necessary, convert the information into descriptive text for greater reliability.
- Character correction and use of OCR: When training with PDF documents, OCR can alter characters. To avoid this, it is recommended to use fonts like Times New Roman that are more reliable in optical character recognition (OCR) processes.
- Use of links: Links should have a secure navigation format (https), be public, and visible for information access.
- Additional instructions: Include simple instructions in the content, such as "talk to an advisor," if the user needs further clarification or human support. This guides the user toward appropriate assistance when the AI cannot resolve the inquiry.
- Testing before production: Test the generated responses in a controlled environment before activating them in production to identify any potential errors or necessary adjustments.
- Constant review: Frequently review the content generated by the AI to validate that it remains aligned with your company's policies and does not deviate from key messages.
- Clarity and precision: Verify that the generated responses are clear and specific, avoiding ambiguities or misconceptions.
- Training with educational content: To train the AI on academic topics, it is advisable to use a separate bot for each subject to avoid confusion. Students often do not provide enough context in their inquiries, so dividing topics will improve the bot's accuracy.