How a Digital Advertising Platform Revamped Data Access With a Text-to-SQL AI Agent

How a Digital Advertising Platform Revamped Data Access With StackAI

Overview

A top digital advertising SaaS company partnered with StackAI to make its customer data more accessible across internal teams. By integrating AI-powered text-to-SQL capabilities, the company reduced time spent on manual queries by around 20%, freeing up over 80 hours per month across sales and customer support. Now, with secure, no-code agents accessing Snowflake in natural language, employees generate insights in seconds, not hours.

The Challenge: Data Bottlenecks Slowed Sales and Support

This company provides advanced targeting and personalization tools to enterprise clients. Internally, however, accessing the data to support those services was far from seamless.

Sales and support teams needed regular access to customer data housed in Snowflake, but retrieving insights required writing complex SQL queries. That meant every request had to go through the overburdened data and analytics team, which was spending 20+ hours per week manually pulling audience sizes, purchase trends, and engagement insights.

As the client base grew, so did the volume of requests. But the data bottleneck limited the support team’s effectiveness, slowed sales cycles, and strained resources across the org. The company needed a faster, scalable, and secure way for business teams to get the data they needed—without writing code or waiting on analysts.

The Solution: Natural Language to SQL, Powered by StackAI

To solve this, the firm partnered with StackAI to build a custom AI-powered assistant with text-to-SQL capabilities. Built securely within StackAI’s no-code platform, the assistant allows users to ask questions in plain English (for example, “what’s the total number of users who purchased in the last 30 days?") and get instant, trustworthy results pulled directly from Snowflake.

Here’s how it works: Users type their question into a conversational interface. That prompt is passed to a privately deployed GPT-4o model, which has been trained using the company’s data dictionary, documentation, and historical examples to generate highly accurate SQL queries. The query is then securely executed on Snowflake via a governed, enterprise-grade connection. Metadata and results are sent to a second, privately deployed LLM (Azure OpenAI GPT-4), which generates a natural language answer complete with SQL, explanation, and optional charting. Users can share answers, review past queries, and even reuse templates across teams.

🕒 This AI assistant saves over 80 hours per month across sales and support. 

After being rolled out to even just five active users, the AI assistant saves approximately four hours per week, per user—totaling 80 hours monthly across sales and support. With adoption growing, the team expects similar gains as usage expands. The assistant was built and deployed in just six weeks, and its success has reduced AI development costs by over 80% compared to traditional engineering-heavy builds.

80+ Hours Saved Monthly, and Growing

Today, sales and support teams use the AI assistant daily to generate fast, reliable insights without touching SQL. With just five active users, the company already saves over 80 hours per month, roughly 4 hours per user, per week. As rollout expands to other teams, those gains are expected to scale dramatically.

The assistant went from concept to production in just 6 weeks, saving an estimated 80% of traditional AI development costs.

This company continues to expand its AI footprint with StackAI, making internal data smarter, more accessible, and radically faster to use.