KobiMaster AdsEngine
AI Agent System That Automates Amazon Sponsored Products Ad Management
Intelligent, automated, and controlled ad management with an LLM + tool-calling architecture. A decision engine powered by artificial intelligence that manages Amazon advertising operations end to end through discovery, optimization, health checks, budget management, protection, and reporting layers.
Automation Engine
AI Ads Engine
An automation engine that connects Amazon Ads to a daily optimization and protection cycle.
Discovery
Recommends campaigns by product/category
Optimization
Improves performance with bid/keyword recommendations
HealthCheck
Recommends pause/start actions for Buy Box, stock, and listing risks
Budget
Shifts budget to better-performing areas
Protection
Applies warn/slow down/stop actions when limits are exceeded
Reporting
Generates a report on what happened / why / next move
Runs Daily
Discovery
Scans products without campaigns and reference market signals to identify products with advertising potential.
Runs Daily
Optimization
Analyzes current campaign and keyword performance, optimizing bids according to ACoS targets.
Hourly Operation
Protection
Monitors critical guardrail rules such as ACoS, daily spend, BuyBox, and minimum stock in real time; prevents ad losses from growing with rapid intervention in risky situations.
Runs Daily
HealthCheck
Checks the listing quality, BuyBox, stock, and Prime status of active campaign products.
Weekly Runs
BudgetManagement
Segments campaigns by performance class and shifts budget toward campaigns that are performing well.
Weekly / Monthly
Reporting
Combines performance data and AI decision history to generate reports in natural language; provides management and operations teams with a clear, explainable, and decision-supporting advertising performance report.
Current Status and Next Steps
✅ Completed
  • 6 AI roles and tool definitions completed
  • Pipeline architecture set up as PromptStep[]
  • ApprovalQueue structure prepared
  • EventHistory structure prepared
  • OpenAI LLM integration is working
🔧 In Progress
  • Reading tools' real DB and Amazon API connections are not fully in place yet
  • Amazon Advertising API execute backends have not been completed
  • Claude and other LLM provider integrations have not been added
  • ApprovalQueue pending dedup process has not been completed
  • HealthCheck tool should be updated according to the new data model

Next Goal: Connect the execute tools to the Amazon Advertising API and conduct the first live test.

Strategic Note: In the first live test, the focus should not be on running broad automated optimization. The priority should be measuring controlled decision quality through the ApprovalQueue in low-risk campaigns.