Before today, when looking for an analysis of revenue, or supplier performance, or the impact of tariffs, one first would need to reach out to a business or financial analyst. The request then would go through several databases, pulling in the appropriate data and performing the requisite reviews. Should all go well, this process could take nearly five business days and cost $18,000, according to Dashboards are Dead. Should it not go well — a hiccup like not having access to the appropriate data or not having set up the right dashboard — the process can take much, much longer.
Today, that all changes. Today we introduce hila: Deep Analysis. This mode fundamentally shifts how an enterprise can perform significant analysis. Take the video below for example — from a high-level question about sales revenue, hila reasons through the data and breaks down the best questions, as well as understands the questions that will answer the initial question. In the example of sales revenue, the initial user prompt was: How has sales revenue changed between 2022 and 2024? Analyze the main drivers of this change.
The subsequent questions based on that initial complex and ambiguous prompt were:
Each of these answers contributes to an overall picture of the business. In the end, these answers are summarized and provide a holistic view of the entire business. It goes through all of the appropriate data to pull out the crucial information for a business review.
But how does it work, and how can we reason through something like this with confidence in our accuracy, speed and cost? While each component deserves particular attention, the core to our ability to accurately answer questions against new data is in the knowledge model that we create against the data itself.
By working directly from the data and applying domain-specific knowledge, we can build a model specific to a company and its data. This knowledge model is built quickly, in the case of NetSuite, it’s done within a few minutes.
The result is a comprehensive model that understands what can be asked against the data to provide answers. This knowledge model then becomes the backbone for agents to reason through ambiguous questions — through the knowledge model, the agents understand how to break down the question in a way that provides correct and robust responses. This then ensures that the responses are 100 percent accurate.
These agents perform several crucial functions. They detect the use intent, perform intelligent retrieval, disambiguate and sanitize the query, they create code from natural language, they perform confidence and sanity checks, they ask follow ups, and they create the appropriate tables, charts and summaries. They do all of this within seconds. The additional agents in Deep Analysis use the knowledge model to reason through complex and high level questions — they can then break down the query into questions that the data can answer. In this way, the system, reasons on only the data, and responds accurately, with all of the speed and cost efficiency built into hila.
Moreover, by using hila, the costs and latency is minimal — we can provide this analysis in a couple of minutes for a tiny fraction of what these dashboards would cost.
And we can do something that a dashboard can’t — you can ask follow up questions or change a response.
Step into the future of analytics with us: Contact us