How Kargo Built a Conversational BI Chatbot That Generated $238K in New Revenue
Kargo's media strategists were drowning in dashboards - spending hours hunting for client campaign data across complex interfaces. Their Snowflake-powered BI chatbot changed everything, delivering answers 10x faster and directly influencing a $238,000 client upsell. Here's how they did it.
The Dashboard Problem That Cost Kargo Time and Revenue
Kargo's media strategy team faced a common but costly challenge: their dashboards contained too much data. As Brian Herbert, Senior Director of Analytics at Kargo explains, "We had built this massive dashboard that had a bunch of different tiles and basically had any piece of data that they could possibly imagine." But when released, the team found it was "pretty difficult for the media strategist to find the information that they were looking for because of the vast amount of data that we had provided."
The consequences were tangible. Media strategists wasted hours hunting for client campaign history across multiple dashboard views. Without quick access to past performance data, they struggled to build compelling media plans that showcased a client's historical success. The friction in the data access process was directly limiting revenue opportunities.
False belief demolished: More data visibility doesn't always mean better decisions. Kargo learned that curated access to the right data at the right time drives better business outcomes than unfiltered data dumps.
How Snowflake Professional Services Designed the Solution
Having previously worked with Snowflake Professional Services, Kargo's analytics team reached out for help. "Once we realized that we needed some sort of conversational chatbot for BI," Herbert notes, "the first step that professional services took with us was helping us architecture the app."
The solution needed to meet three critical requirements: 1) Natural language interface for non-technical users, 2) Accurate query generation against complex data models, and 3) Full auditability of all data requests. Snowflake's team proposed combining several of their newest technologies into a cohesive solution.
At 1:32 in the video, Herbert explains how Snowflake Professional Services helped them avoid common pitfalls: "What we eventually settled on was building a streamllet application that leveraged Cortex Analyst and large language models to be able to take a natural language question from a user, use Cortex Analyst to generate a SQL query from that question and then throw that into an LLM provided by Cortex."
The Chatbot Architecture: Streamlit + Cortex Analyst + LLM
Kargo's final architecture elegantly combines Snowflake's capabilities:
- Streamlit Interface: Provides the conversational UI that media strategists interact with
- Cortex Analyst: Translates natural language questions into precise SQL queries
- Snowflake LLM: Converts query results into natural language answers
- Audit Log: Stores both the question and generated SQL for compliance
Key insight: The dual output of natural language answers plus the underlying SQL query gives users confidence in the results while maintaining full transparency. As Herbert puts it, "It also allowed us to have an audit log to make sure that they were accurate and providing the right data to customers."
The $238K Revenue Impact: Real-World Results
The chatbot's value became undeniable when it directly influenced a major client upsell. Herbert shares: "There was an instance of a media strategist who used the chatbot to pull past targeting segments for a client that had performed really well. And they then highlighted this in a media plan which run an additional $238,000 for the business."
This wasn't just about speed - it was about discoverability. The conversational interface helped media strategists uncover valuable historical insights that were previously buried in complex dashboards. The chatbot made high-performing past campaigns immediately visible and actionable.
10x Faster Answers and Higher Quality Client Responses
Beyond the direct revenue impact, the chatbot transformed Kargo's BI process:
- Speed: Media strategists get answers "a lot faster" than searching dashboards
- Quality: Client responses have "increased dramatically" in accuracy and relevance
- Adoption: The natural language interface removed technical barriers to data access
- Trust: Visible query audit trails built confidence in the system's outputs
Herbert emphasizes that the solution exceeded expectations: "After using this tool, the media strategists have told us that they're not only getting their answers a lot faster, but the quality of the responses that they're sending to clients has increased dramatically."
Watch the Full Tutorial
See Brian Herbert demonstrate how Kargo's BI chatbot works in the full video (2:15), where he walks through a real media strategist query and shows the dual output of natural language answer plus SQL audit trail.
Key Takeaways
Kargo's BI chatbot success demonstrates how conversational interfaces can transform data access in complex business environments. The solution delivered both immediate ROI and long-term strategic advantages.
In summary: 1) Conversational BI unlocks value buried in complex dashboards, 2) The right architecture (Streamlit + Cortex Analyst + LLM) balances usability with accuracy, and 3) Transparent audit trails build trust in AI-generated insights. Most importantly - this isn't theoretical. Kargo's chatbot directly influenced a $238K upsell by making historical performance data instantly actionable.
Frequently Asked Questions
Common questions about conversational BI chatbots
Kargo's media strategy team struggled to find client campaign history data in their complex dashboards. They needed a centralized way to quickly access past campaign performance data to inform new media plans.
The existing dashboard solution contained too much information, making it difficult to find specific client data. Media strategists were wasting hours hunting for the right metrics across multiple dashboard views.
- Dashboards had become overwhelming with excessive data points
- Finding historical client campaign data was time-consuming
- Lack of quick access to past performance limited upsell opportunities
One media strategist used the chatbot to identify high-performing past targeting segments for a client. By highlighting these in a new media plan, they secured an additional $238,000 in business.
The chatbot made previously hard-to-find performance data instantly accessible. This allowed strategists to quickly surface and leverage historical successes when pitching clients on expanded campaigns.
- $238,000 directly attributed to chatbot-enabled upsell
- Identified high-performing historical segments instantly
- Transformed buried data into actionable sales insights
The chatbot is built on Snowflake as the core data platform, using Streamlit for the application interface. It leverages Snowflake Cortex Analyst to translate natural language questions into SQL queries, and Cortex LLMs to generate natural language answers.
This combination provides both answers and an audit trail of the queries used. The full stack runs within Snowflake's ecosystem, eliminating the need for external API calls or additional infrastructure.
- Streamlit for the conversational UI
- Cortex Analyst for SQL query generation
- Cortex LLM for natural language responses
Media strategists reported getting answers 10x faster than searching dashboards. The quality of client responses improved dramatically because they could quickly access accurate historical data.
The solution also created an audit trail of all queries for compliance and accuracy verification. This combination of speed, quality, and transparency transformed Kargo's BI process.
- 10x faster answer retrieval
- Higher quality client-facing reports
- Full auditability of all data requests
The system uses Snowflake Cortex Analyst to automatically generate SQL queries from natural language questions. These queries are executed against Kargo's Snowflake data warehouse, with results fed into Cortex LLMs to produce natural language answers.
Users see both the answer and the query used to generate it. This transparency helps verify accuracy while maintaining the simplicity of natural language interaction.
- Cortex Analyst converts questions to precise SQL
- Executes against Snowflake's optimized data warehouse
- LLM converts results to natural language with query audit
The chatbot accesses Kargo's entire Snowflake data warehouse, including campaign performance data, targeting segments, creative optimization metrics, and historical client campaign results.
It provides a unified interface to all analytical workloads previously scattered across multiple dashboards. The natural language interface makes this broad data accessible without requiring technical expertise.
- Campaign performance metrics
- Historical client results
- Targeting segment effectiveness
- Creative optimization data
Snowflake Professional Services helped architect the application, recommending the Streamlit + Cortex Analyst + LLM approach. They provided expertise in implementing conversational BI solutions that could handle Kargo's complex data environment.
The services team ensured the solution balanced usability with accuracy and auditability. Their guidance helped Kargo avoid common pitfalls in implementing natural language interfaces for business intelligence.
- Architected the optimal solution stack
- Provided implementation best practices
- Ensured compliance with audit requirements
GrowwStacks helps businesses implement AI-powered BI solutions like Kargo's chatbot. Whether you need a conversational interface to your Snowflake data, automated reporting systems, or custom AI data products, our team can design and deploy solutions tailored to your specific needs.
We specialize in building practical AI solutions that deliver measurable business impact, not just technical demos. Our implementations focus on real-world usability, accuracy, and ROI.
- Custom conversational BI interfaces
- Snowflake Cortex implementations
- End-to-end AI solution development
- Free consultation to assess your needs
Ready to Transform Your BI With Conversational AI?
Every day your team spends hunting for data in dashboards is revenue left on the table. GrowwStacks can build you a Snowflake-powered BI chatbot in as little as 4 weeks - just like Kargo's $238K-generating solution.