Product descriptions:
Guild is a leading talent development company, partnering with the country’s most innovative employers, including Chipotle, Target, Walgreens, JPMorgan Chase, Hilton, PepsiCo, Tyson, and more, to build the talent needed for today and a resilient workforce for tomorrow.
As AI transformed the workforce, Guild needed to help its employees and internal teams navigate a massive, evolving learning catalog — spanning hundreds of providers and programs — at speed and precision. By building Catalog Atlas on Databricks AI and Agent Bricks, Guild created an enterprise-grade, domain-specific agent that acts as the authoritative source of program information. The solution streamlines daily operations, ensures governed access to accurate information, and empowers teams to make faster, more informed decisions.
Scattered Data, Inconsistent Answers, and Lost Time
Guild’s core mission is to unlock opportunity for America’s workforce through education, skilling, and career mobility. As Guild grew, so did the complexity of managing and accessing information about its extensive catalog of education and skilling programs.
Critical knowledge, such as what programs were available or which providers delivered the best outcomes, was often spread across multiple platforms, from dashboards and spreadsheets to internal wikis and Slack threads.“Before Catalog Atlas, accessing information could take hours or require waiting upwards of 24 hours for a Slack response. Meanwhile, team members were constantly interrupted to answer the same questions,” said Rob Taroncher, Senior Product Planning and Operations Manager. “Now, we can make faster, more informed decisions, helping us better serve our employer partners and learners.”
Standard AI chatbots fell short. With Guild’s unique business rules, provider aliases, and high internal standards on automated AI solutions, relying on generic, ungoverned agents could have introduced errors or surfaced outdated information. Teams needed an AI solution grounded in real, governed data — one that was accurate, auditable, and secure.
Catalog Atlas: Enterprise-Grade Answers, Reliable Performance, Real Value
To break through these barriers, Guild turned to the Databricks Data Intelligence Platform to unify its knowledge, govern access, and build Catalog Atlas — a custom agent that puts trusted, context-aware information at employees’ fingertips.
Guild began its enterprise AI agent journey in Databricks by experimenting with Agent Bricks first. These early prototypes enabled the team to quickly upload data sources, test system instructions, and identify where generic agents struggled. While the beta didn’t yet provide the creative reasoning required for production, it gave Guild invaluable insight into what worked, what didn’t, and how to design a robust long-term solution.
Built on Databricks, Catalog Atlas unified structured and unstructured data into Delta tables, Volumes, and Vector Stores, all governed by Unity Catalog. This architecture enabled enrichment with alias mappings and business rules (e.g., “Cornell” vs. “eCornell”). At the same time, MLflow evaluations created a continuous improvement loop that steadily improved accuracy and reduced manual QA checks by over 90% through the use of automated AI judges.
Databricks also gave Guild the flexibility to adapt as its AI needs evolved. In the Databricks AI Playground, the team conducted head-to-head model comparisons, feeding the same prompt and tool access to multiple models to evaluate query times, token usage, and output quality side by side. Those insights guided them to swap in faster or more cost-efficient models, which, in practice, was as simple as changing a single line of code in their notebook. This agility helped Guild cut query times by more than 80% during development and ensured Catalog Atlas could consistently deliver the best balance of performance and cost.
At the same time, Databricks’ governance model ensured security stayed intact. By bringing the model to the data under Unity Catalog, Guild maintained program information control, masked sensitive columns when needed, and eliminated the risk of outdated or unverified sources being introduced into the agent. Reusable MCP connectors enabled the standardization of trusted tools, such as a custom catalog search tool, allowing them to be leveraged across multiple agents without the need for rebuilding integrations, thereby giving the team confidence in both quality and compliance.
Deployment was seamless: Catalog Atlas was built in notebooks, deployed with AI Model Serving, and delivered through Databricks Apps (via a custom-built chat interface) with SSO authentication enabled by Databricks One, making access secure and straightforward for employees company-wide.
Faster Decisions, Strategic Work, and Scalable AI
The impact of Catalog Atlas is clear and measurable. Automating catalog-related queries saves Guild an estimated 450 hours per year — nearly a quarter of a full-time employee's time — generating $33,750 in annual value that can be redirected toward higher-impact projects. “Catalog Atlas changed our operating tempo. Multi-hour information requests or day-long waits on Slack are now resolved in thirty seconds or less. Teams get the answers they need instantly, so they can focus on supporting learners and employers, and this is just the start,” Rob said.
The qualitative improvements are equally transformative. Teams act with greater agility and confidence, with Catalog Atlas as a governed, single source of truth. The opportunity cost of subject-matter experts being pulled into repetitive questions has plummeted. Employees simply ask and get a reliable answer.
Looking ahead, Guild is considering plans to extend Catalog Atlas to employer-facing admin portals. Such an implementation would empower HR leaders to add new programs to their existing catalog easily. Additionally, integrations with systems like Salesforce are being evaluated for their potential to deliver more personalized recommendations. The architecture behind Catalog Atlas establishes a reusable blueprint for future deployable enterprise-grade AI solutions, characterized by governed data, measurable results, full auditability, model flexibility, and the lowest-cost serving at scale.
Guild’s partnership with Databricks shows how AI can move from experiment to value in the enterprise. With Agent Bricks and Databricks AI, Guild is not only accelerating its mission to unlock opportunity for America’s workforce, but also ensuring AI empowers teams with accurate, actionable information — securely and at scale. “Databricks lets us build with confidence. Catalog Atlas cuts through friction, drives productivity, and gives us a repeatable AI foundation for the future that meets our enterprise needs,” Rob concluded.