AI Brand Reports for CPG Decision-Makers

Case Study: Brandtuit

Brandtuit partnered with Melt Studio to develop an AI-powered brand management platform for CPG brand managers—bringing brand data into one place, analyzing it with AI, and generating ready-to-use, client-presentable reports in a secure, multi-tenant environment.

Projc

THE CHALLENGE

Unifying Global + Private Data for Faster Insights

Brand teams needed actionable insights from many sources. The platform required global industry knowledge plus tenant-specific private data—with strict tenant isolation—so users could ask questions, uncover patterns, and generate reports efficiently.

INITIAL STATE

OUR SOLUTION

We designed a multi-agent system with a two-tier knowledge layer (Global + Tenant-Specific) to ingest, analyze, and synthesize information in real time—producing non-obvious recommendations and professional-quality reports through iterative sprints.

Requirements

Two-tier knowledge scope.

Defined the Contextual Intelligence Layer: a Global Knowledge Layer (frameworks, definitions) and a Tenant-Specific layer for proprietary content—plus retrieval rules that preserve isolation.

Application

Agent + retrieval building

Specified backend services via API, integrating LangChain for retrieval orchestration and agent workflows, LangGraph for complex reasoning flows (as needed), and LangFuse for observability.

Prototype

Report pipeline iteration

Prototyped knowledge-store needs: vector storage and metadata for traceability and access control, with vector DB options noted (Pinecone/Weaviate), plus embedding APIs and search endpoints.

DELIVER STATE

CUSTOMER FEEDBACK

Report Quality Improved via V10 Evaluation Pipeline

Sprint reviews documented report-quality gains in “Report V10,” driven by a new evaluation + score aggregation system. The team hardened the pipeline with a model-specific “prime” step and improved preparation for client-ready delivery.

Trusted by Startups Across the U.S.

We have delivered products for startups from Anchorage to New York.

Jonathan

Truvle

January 2, 2026

Product is now in a testable state. Overall sentiment was positive. The team noted a large jump in usability and that the product feels useful and ready for thorough testing.

January 2, 2026

Avani Kanubaddi

Brandtuit

December 30, 2025

Thanks, everyone, for the hard work that went into this first beta client. We have now officially released our first Zena report to a client. We will meet with him on Monday to review his candid feedback on what is working well and what needs improvement.

This was a significant step, so we should take a moment to commemorate and celebrate it. We will have two more beta tests in the new year, so the work continues. As we turn the calendar, I wanted to take a moment to appreciate this team. I wish everyone a fantastic start to 2026, and I hope we can continue to build on this promising start.

December 30, 2025