AI Data Analyst

Hacker News

Share card

AI Data Analyst

Hey HN, Michael and Alexey from Narrative BI here. Background We've been working on Natural Language Query tools for quite some time. Back in the pre-LLM era, we launched a natural language to SQL startup called FriendlyData. It was later acquired by Service now where we integrated it into the Now Intelligence platform. Then we started Narrative BI with the thesis that business users like marketers don't always know how to ask data questions, so our goal was to build a product where data insights are generated proactively, before users even ask questions. Evolution The general dynamics of Generative AI going mainstream created an interesting opportunity. We still believe that the future of BI lies in proactive analytics and easy data integrations. However, with increased AI awareness, we've seen a lot of interest from our users in natural language capabilities, so we decided to take it to the next level. Why don't just use ChatGPT? The problem with ChatGPT and other chat-based LLM tools that they hallucinate a lot. ChatGPT answers look plausible so most of users don't bother to verify if the answers match the actual data. However our research shows that ChatGPT hallucination rate is 47% for more or less complex data sets, such as Google Search Console data, for example. For business data analysis accuracy is the key as wrong answers can lead to misinformed decisions. Another limitation is data integration. AI data analyst comes with pre-built data integrations, so you can easily connect your data sources such as GA4, Google Ads, Google Search Console, FB Ads, Hubspot and Salesforce. More data sources coming soon and we're also doing custom integrations for enterprise. We're here to answer your questions and look forward to your feedback!