How Kognitos Built an Enterprise Analytics Platform with Letta
What Letta allowed us to do is rapidly iterate on this system. I could just prototype much faster. I was able to see the product working right away.
Kognitos, a leader in AI-native business automation for enterprises, selected Letta to build a sophisticated business analytics tool for a major logistics client. Their story demonstrates how Letta's agents can tackle immensely complex enterprise data challenges, as well as enabling faster iteration cycles than building in-house AI workflows.
The Challenge: Going beyond Text-to-SQL with Agentic AI
Kognitos's core strength lies in its English-as-code automation engine, which provides a dependable, non-hallucinating solution for structured business processes. However, they faced a unique challenge with a new customer—a leading logistics company in the Midwest. The customer needed a way to query a vast and convoluted data landscape spread across multiple systems, including ERPs and a logistics management system containing over 4,000 poorly labeled tables.
The executives weren't just asking for text-to-SQL conversion. They needed a tool that could answer ad-hoc questions and provide personalized experiences for different executive personas (including CEO, CFO, CIO, and others) while maintaining context across conversations. The solution needed to go beyond standard text-to-SQL and provide a “ChatGPT experience”.
"The executives didn't even care that much about the text-to-SQL," explains Vansh Khanna, a Member of Technical Staff at Kognitos. "They cared more about ease of use, and having a plain English semantic layer across their data.”
Finding the Solution: Rapid Prototyping with Letta
The Kognitos team needed a solution that offered flexibility, powerful function calling, and persistent memory. They found their answer in Letta.
With Letta, the Kognitos team was able to implement a proof-of-concept within one day, and quickly evolve it into a production-ready solution. This rapid iteration was made possible by several key Letta features:
- The Agent Development Environment (ADE): "The ADE was really helpful to actually see the agent doing all these things," said Khanna. The ADE provided the visibility needed to debug and understand the agent's reasoning and execution process, dramatically accelerating development.
- Open Source Chat UI: Kognitos used Letta's open-source Next.js chat UI example as a reference, allowing them to quickly build a feature-rich, high-quality prototype with streaming and reasoning visibility—crucial elements for securing early customer buy-in.
- MCP Connectors: Letta's support for Model Context Protocol (MCP) connectors made it simple to integrate with the customer's various internal data sources and services.
The Letta-based agents had an impressive ability to autonomously reason and call tools across many many steps to find the right data. "Previously it would take maybe fifteen back-and-forth messages to get a result out of the chat," Khanna recalls. "But now it's able to deliver results in a single query, with the agent making ten or twenty tool calls autonomously."
The agent could independently receive a high-level user query in natural language, intelligently search through the 4,000+ database tables to identify the most relevant ones, query the tables to get their schemas, decipher cryptic column names and synthesize the data to generate comprehensive reports and charts for the user.
"The agent is given the freedom to fetch data from wherever it needs," Khanna explains. "It figures out the schema and then goes, 'All right, you want to do this? I'm going to make this query, call this query, and create the chart for you.'"
The Result: A Half-Million-Dollar Success Story
The Letta-powered application successfully met every feature and requirement the client needed, while also providing extensive flexibility for future enhancements with minimal additional effort. The logistics client was so impressed with the chat-based analytics tool that they approved a nearly half-million dollar expansion. Beyond this direct expansion, the trust earned and customer focus demonstrated through this implementation led to a broader expansion of the overall contract for Kognitos's core automation products.
Future Plans: Integrating Letta into the Core Product
The success of this project has inspired Kognitos to explore deeper integrations with Letta and agents more generally. They plan to leverage Letta for business analytics, allowing executives across all their customers to query operational data; exception handling, using Letta's memory to help users resolve automation exceptions with context; and a universal chat interface that is a complete reimagining of their product where all functionality becomes accessible through natural conversation.
“The feature set aligned perfectly with our technical requirements and development methodology” said Khanna. It provided the perfect blend of power, flexibility, and ease of prototyping, allowing Kognitos not only to meet but exceed their customer's expectations, transforming a complex data challenge into a major business opportunity.
Try Letta today, or request a demo for your team.