Sarah Nagy’s Post

People have always dreamed of talking directly to their data, but business data lives in large databases accessible only by making manual requests to a small team of technical data specialists. Despite huge spend on data storage, ROI from analysis remains limited. Seek changes that. This week, we’re giving our first group of customers access to Deep Data Reasoning. Out-of-the-box LLMs and reasoning models are great at analyzing small data. But for enterprise use, we need a completely different approach to accommodate billions or trillions of rows, messy real world data, and complex business logic. Deep Data Reasoning combines agents and tool use with a self-building knowledge graph that captures the complexity of enterprise business logic and how it relates to the underlying data. Traditional simple code generation approaches don’t have access to the complex business logic required to provide accurate results. In these types of systems, agents are expected to synthesize large amounts of raw data into an answer, without all of the information they need to reliably avoid hallucinations. Working with a Fortune 100 customer, we found that our approach doubled accuracy compared to this type of simplistic codegen. Deep Data Reasoning is an interface that replaces the antiquated workflow of filling out a JIRA ticket or sending a Slack request to a team of data analysts. Employees within business units such as marketing, product and operations are highly curious and eager to get their hands on as much data as possible. Deep Data Reasoning is like a firehose putting all of this information at their fingertips. Email me at sarah@seek.ai if you want early access or to talk to our eng/research teams!