Databricks Agent Bricks: AI-Powered Sales Offer Targeting
For sales teams, a potent offer is only as effective as its precise delivery. At technology companies like Databricks, providing free credit offers aims to onboard new customers or accelerate existing adoption. Yet, a deceptively simple question often bogs down sales representatives: which customer accounts are eligible for these offers, and which should be prioritized for outreach?
What appears straightforward can quickly become an opaque, time-consuming endeavor, often requiring multiple teams. Sales personnel frequently find themselves sifting through extensive documentation, scouring internal communication channels, and manually investigating account specifics with operations teams. This cumbersome process creates unnecessary back-and-forth, slows down sales momentum, and ultimately delays the delivery of valuable offers to customers. Even when eligibility is clear, discerning the highest-impact prospects for prioritization remains a challenge.
To address this internal bottleneck, Databricks leveraged its own “Agent Bricks” platform, designed for building sophisticated AI agents atop enterprise data. The result was a multi-agent system engineered to deliver clear, actionable guidance directly to its sales force. Remarkably, this comprehensive tool was developed in less than two days by an intern in Business Strategy and Operations, underscoring the platform’s speed and simplicity. The system empowers sales representatives to rapidly identify qualifying customer accounts, understand the precise reasons for any ineligibility, and rank eligible accounts to focus on the highest-impact prospects first.
The core of this solution lies in Agent Bricks’ Multi-Agent Supervisor, which orchestrates the activities of three specialized AI agents. Much like an air-traffic controller, the Supervisor intelligently delegates specific parts of a query to the appropriate agent, then seamlessly stitches their individual responses into a single, cohesive answer.
The first component is an Offer Details Agent, powered by a Knowledge Assistant. This agent is trained on unstructured internal offer documentation, including PDFs and slide decks, enabling it to deeply understand the intricate rules, eligibility requirements, and the entire outreach and delivery process for various offers. Crucially, the Knowledge Assistant processes these documents in their native format, eliminating the need for any pre-processing or embedding work.
Next is the Offer Eligibility Agent, built using an AI/BI Genie. This agent analyzes structured customer account data, securely governed within Unity Catalog, to determine not only which customers qualify for specific offers but, equally important, why others do not. It can pinpoint the exact eligibility criteria an account fails to meet and even suggest follow-up steps should a sales representative wish to troubleshoot further. To facilitate this analysis, the underlying data tables are meticulously structured with columns relevant to each eligibility criterion.
Finally, an Account Prioritization Agent, also powered by an AI/BI Genie, examines structured go-to-market data. This agent ranks eligible accounts by considering factors such as usage data, growth signals, and the overall relevance of the offer to the customer. This provides sales teams with a clear, prioritized list, directing them to the most promising contacts first.
This innovative multi-agent system streamlines a previously cumbersome sales process, transforming manual guesswork into data-driven precision. The ease with which such a sophisticated AI solution could be built directly on existing customer data and offer documentation, without requiring extensive research into complex agent architectures or engagement with specialized technical teams, highlights a significant leap forward in enterprise AI application.