The article then introduces PromptQL's agentic approach to data retrieval, which mimics human behavior. This approach involves gathering relevant data first, such as emails from the previous week, and then applying the appropriate LLM to classify if any follow-ups are needed. This method is touted as being more accurate and efficient, as it mirrors the way humans would naturally approach data retrieval.
Key takeaways:
- Today's AI assistants in closed domains often miss data that doesn't fit their search criteria, causing them to falter when users go off script.
- These AI assistants might not respond usefully to queries like 'Find all receipts from October' because 'October' was not vectorized.
- PromptQL's agentic query planning aims to retrieve data like a human, first gathering relevant emails from last week, then applying the right LLM to classify if there are follow-ups required.
- PromptQL's approach is designed to be more accurate and flexible, as shown in the agentic data access benchmark.