Beyond the Silo: Building Athena AI, A Cross-Brand Shopping Agent

The MACH AI Exchange Hackathon wasn't just another coding sprint – it was a real-world test of the MACH philosophy in the age of AI. Our B2C team took on a challenge that plagues countless multi-brand enterprises (the classic "House of Brands" problem): how do you prevent various brands from feeling like fragmented, separate experiences?

The customer doesn't care if Brand A's products live on Elastic Path and Brand B's products are in some other vendor. They just want the best solution. Our mission was to eliminate those technical silos and deliver unified, intelligent product discovery using a truly composable architecture.

The result? Athena AI, an AI-powered Cross-Brand Shopping Agent designed to be the single point of truth for a multi-brand portfolio.

The Athena AI Architecture: Orchestration is King

Our solution was an exercise in smart orchestration. Athena AI functions as an intelligent layer that sits above the existing brand systems, transforming complex customer queries into curated, cross-brand recommendations, all without forcing disruptive redirects.

To achieve this, we integrated a powerful, multi-vendor stack:

• Intelligence & Interface: We leveraged Bloomreach Clarity to handle the conversational front end, interpreting natural language queries and intent.
• The Orchestration Backbone: Uniform (with Scout AI) played a crucial role. Uniform organized and assembled product content from disparate systems, while Scout AI enriched that data, filling in gaps and adding contextual tags necessary for the conversational agent to make accurate, informed recommendations.
• The Data Providers: The backend demonstrated connectivity to various commerce and product systems, including Elastic Path and others, proving the solution's true vendor agnosticism.

The Devil in the Details: Integration and Evaluation

Data Connection & API-Ready Product Data: This part of the hackathon confirmed two vital truths. First, integration remains a key process. Even with the best tooling, we had to ensure Uniform was correctly wired to the Elastic Path APIs. But second, and perhaps more importantly, we witnessed the power of API-native Product Experience Manager data in action. Elastic Path's product information, complete with rich attributes, hierarchies, and relationships, was already structured for programmatic consumption. This meant Scout AI could immediately enrich and contextualize products without extensive ETL work. In an agentic world, "data readiness" isn't just about having data; it's about having data that's inherently consumable by digital agents.

Metrics Architecture: One of the most insightful parts of the project was stepping back to define how to judge the success of such a product. I led the effort to establish the system's KPIs and we approached this from a top-down perspective, defining goals across three critical domains.

1. Business Impact: Measuring key outcomes like Conversion Rate Uplift from cross-brand recommendations and Average Order Value (AOV) increase.
2. User Experience (UX): Tracking things like Response Time, Answer Accuracy, and User Satisfaction (e.g., did the user click on the recommendation?).
3. Operational Excellence: Evaluating the efficiency of the AI layer itself, such as the cost-per-query or the efficacy of data normalization.

Thinking through these metrics at a component level patterns – essentially an AI-driven ETL flow with a client UI on the end – forced us to consider performance not just in human terms, but in system-to-system terms. I really appreciated the chance to dig into the metrics and think about the different, and changing, dynamics at play in an agentic world, blending classic metrics with new perspectives to underpin business value.

Key Takeaways for an Agent-Centric World

The speed and quality of our outcome highlight two major takeaways:

1. The Power of Composable Collaboration: It was remarkable that a team comprising people from different companies and meeting only a couple of times a week could successfully integrate and build something genuinely novel. This speaks to the motivation, skill, and sheer power of collaborating on a composable stack. The API-first architecture reduces friction required to unify these disparate components and I was especially motivated seeing how a common goal and AI support helped us glue disparate systems together so quickly.
2. The Inevitable Agent-Centric Shift: The early standard for AI implementation is often the chatbot, front-and-center. But our technical hurdles, including the need to meticulously define interface points and interaction contracts between systems, points to another shift. We are rapidly moving toward an agent-centric world where digital systems (such as AI agents) do most of the communication. The architecture of a system must now be designed not just for human developers or end-users, but for cooperating digital agents consuming and utilizing data. This, in itself, is an interesting process and a good exercise in forcing builders to rethink system usage, as our human audiences now include digital consumers, as well.

We didn't just build a solution; we gained a key insight into how systems must evolve to thrive in an autonomous, AI-driven future.

This was a fantastic exercise in distributed innovation, and the Athena AI project is a testament to what a MACH Alliance hackathon can achieve. For Elastic Path, it validated our belief that API-native commerce isn't just a technical choice; it's a strategic enabler for the agentic commerce future we're all building together. Here's to further collaboration and building!