🌟 Inspiration
Traditional marketing agencies offer human-like collaboration but often miss the mark on precision. On the other hand, ad tech tools are powerful but require too much input, lacking the human feel. We wanted to bridge that gap — build an AI agent that feels like you're talking to a real expert, but with the precision of data-backed decisions and automation.
⚙️ What it Does
AdBuddy.ai is your AI-powered advertising consultant that:
- 🗣️ Chats like a real consultant to understand your business
- 📊 Analyzes competitors and trends using Qloo’s taste intelligence
- 🎯 Helps plan accurate campaigns: targeting, strategy, and creatives
- 💡 Suggests ad ideas and audience insights—all in one smooth workflow
No input overload. Just smart conversations with results.
🛠️ How We Built It
We built AdBuddy.ai using:
- 🧩 Frontend: React.js
- 🚀 Backend: FastAPI
- 🧠 Agent Framework: Langchain-based LLM agentic workflows
- 🗄️ Database: MongoDB
- 📡 External Intelligence: Qloo Taste Insights API for behavior-based audience and brand data
🧪 Major Workflows (placeholder for now):
- 🧠 Campaign Generation Workflow
Generates a complete ad campaign from natural conversation input, enhanced with Qloo’s taste-driven data.
</> Source Code → AdBuddy.ai/backend/app/services/campaign.py
graph TD
subgraph Campaign Generation Workflow
A[Start] --> B[Initial Planning Node];
B -- Uses LLM with conversation transcript --> B1(Generate Campaign Title & Qloo Query);
B1 --> C[Fetch Qloo Data Node];
C -- Invokes Qloo Workflow with Qloo Query --> C1(Get Optimized Qloo API Parameters);
C1 -- Parameters --> C2[Call Qloo Insights API];
C2 -- Audience Data & Insights --> D[Generate Enhanced Campaign Node];
D -- Uses LLM with all collected data --> D1(Create Full Campaign Plan, Creatives, KPIs, etc.);
D1 --> E[End];
end
- 🔍 Qloo Parameter Generation Workflow
Transforms a high-level query into a precise Qloo Taste Insights API call by resolving tags, audiences, and locations.
</> Source Code → AdBuddy.ai/backend/app/services/qloo.py
graph TD
subgraph Qloo Parameter Generation Workflow
A[Start] --> B[Planner Node];
B -- Uses LLM --> B1(Generate Initial Parameters & Identify Resolvers);
B1 --> C[Process Resolvers Node];
C --> C1{Has Tag Resolvers?};
C1 -- Yes --> C2[Resolve Tags];
C2 -- 1. Call Qloo Tags API <br> 2. Use LLM to select best tags --> C;
C1 -- No --> C3{Has Audience Resolvers?};
C3 -- Yes --> C4[Resolve Audiences];
C4 -- 1. Call Qloo Audiences API <br> 2. Use LLM to select best audiences --> C;
C3 -- No --> C5{Has Location Resolvers?};
C5 -- Yes --> C6[Resolve Locations];
C6 -- Call Here Maps Geocoding API --> C;
C5 -- No --> D[Return Final Parameters];
D --> E[End];
end
🚧 Challenges We Ran Into
- Qloo's insights API is extremely rich, with dozens of parameters.
- Building a flexible agent that can intelligently fill and balance these parameters to get the most accurate and relevant results was a major technical challenge.
- Ensuring the user experience stayed human-like, even while dealing with complex API and agent logic.
🏆 Accomplishments That We're Proud Of
- 🔋 A fully functional AI agent that’s not just a demo — it's ready to solve real business challenges.
- 🤖 Feels like talking to a human consultant, but smarter and scalable.
- 🧠 Reusable agentic flow for multiple tasks — from competitor analysis to campaign building.
📚 What We Learned
- Designing agentic workflows is less about coding and more about creating decision-making logic.
- A single agent with the right tools can be multi-functional, adapting from competitor analysis to campaign creation.
- UX matters — no matter how smart your backend is, users want clarity and ease.
🚀 What’s Next for AdBuddy.ai
- 🔄 One-click action: Turn planned campaigns into live TikTok/Meta ads
- 🕵️♂️ Product discovery tools: Help users explore and test new product ideas
- 🧪 Market testing surveys: Get feedback on campaign performance and effectiveness
- 💡 Continuously refining the agent via user feedback loops
Log in or sign up for Devpost to join the conversation.