Data LLM: Get insights from your data

Every organization has tons of data but it’s always a challenge to find actionable insights from it. Much like finding a needle in the haystack, you have to look through tons of reports to discern concrete patterns from which you can make decisions. Unless one has a deep understanding of the various data sources and SQL or has a data analyst at hand, getting actionable insights is tough. This bottleneck prevents a lot of business users from making better decisions using data. 

While democratizing access to these insights has been a distant dream for a long time, the evolution of generative AI has provided a promising pathway to make this dream a reality. The current state of the art Large Language Models can do this heavy lifting for you as they are capable of understanding the question asked by the user, fetching the right tables required to answer the question, generating the query to extract the results from the data warehouse and summarizing them just as a data analyst would do. However implementing this solution is not a trivial task as the whole database cannot be given as a context to a Large language model. So, a technique similar to Retrieval Augmented Generation (RAG) but for querying data warehouses is needed and Abacus.AI’s Data LLM Agent provides this as a packaged solution. All you have to do is ask the Data LLM Agent a question in natural language and it does the rest, giving you the results just as a personal data analyst or a team of analysts would. With this, any team member regardless of technical expertise, be it a marketer or a product manager can instantly pull customer data reports such as “Number of signups by the day of the week in the US”, “No of customers who churned at a regional level” etc.

Setup and Usage

Before using the Data LLM Agent, you need to set it up which involves creating an AI Agents project, providing access to your data warehouse and within a few clicks Abacus.AI Data LLM Agent connects to your data warehouse and extracts metadata about the tables and stores in a document retriever. Once set up, the Data LLM agent is ready to be used and can respond to queries from your users. When the Data LLM agent receives a query, it performs a sequence of steps

• First, the Data LLM Agent queries the document retriever and gets the relevant tables and their metadata needed to answer the question
• Second, the agent takes the help of an LLM – be it OpenAI, Claude, Palm, Llama2, NSQL, etc – to form the SQL query for your data warehouse to answer the question
• Third, the agent queries your data warehouse using the connector and fetches the relevant data
• Finally, the agents uses an LLM to summarize the results to answer the question asked by the user and responds back

An illustration for the setup and working of the Data LLM AI Agent is shown below

Interacting with the Data LLM Agent

To interact with the Data LLM Agent, Abacus.AI provides a bunch of options

First, you can interact using the Dashboard provided within the product. An example of this is shown below

Second, Abacus.AI offers Bots for business communication platforms like Teams and Slack. These allow you to engage with the Data LLM Agent directly from the messaging application and generate custom reports.

Third, Abacus.AI offers a consumer-facing application similar to ChatGPT that can be exposed to both employees within the company as well as customers and partners. Imagine how effective it could be to provide a chatbot for your customers. They could easily access their data, avoiding complicated interfaces or waiting times to speak with customer support. For example, they could request, “Pull up my purchases from the last week where I spent over $50.”

Finally, Abacus.AI also provides an API endpoint that you can use to power any app. The endpoint can also be used to link multiple agents so they can interact with each other.

Security

While democratizing access to data across your organization as well as to your customers adds a ton of value, it possesses security risks. To address the security risks and to prevent unauthorized access to data, Abacus.AI provides RBAC and the required security settings to ensure that users can only access the data that they have permissions for. Abacus.AI can easily integrate with multiple IAM services such as Okta, Azure AD and Google Workspace to provide seamless access management and ensure secure, controlled access to your data.

If you are interested in learning more or want us to create a Data LLM Agent on your data, contact us at contact@abacus.ai for a free POC and/or a consultation