OpenAI has introduced “Company Knowledge,” a feature for ChatGPT aimed at enhancing productivity for business, enterprise, and educational users. This tool leverages GPT-5 to allow searches across platforms like Slack, SharePoint, Google Drive, and GitHub, aiming to create an internal search engine for company data. While the feature offers source references and can answer vague inquiries, it must be manually activated and does not support web searches or image generation.
Despite its potential, the effectiveness of large language models (LLMs) like ChatGPT in handling diverse data is questionable. Issues such as “AI workslop” could result in inaccurate citations and decreased productivity, costing companies significantly. Current LLMs perform best with well-defined tasks or exploratory searches. Context engineering is crucial for maintaining answer quality; poorly structured information can degrade performance. Companies need to emphasize the limitations of these systems, ensuring human oversight to avoid reliance on potentially flawed outputs. OpenAI’s announcement overlooks these risks, underscoring the necessity of careful usage.
Source link
