CLICKFORCE is one of the leaders in digital advertising services in Taiwan, specializing in data-driven advertising and conversion (D4A – Data for Advertising & Action). CLICKFORCE’s mission is to provide industry-leading, trend-driven and innovative marketing solutions that help brands, agencies and media partners make smarter advertising decisions.
However, as the advertising industry rapidly evolves, traditional analytical methods and common AI outputs are no longer sufficient to provide actionable insights. To stay competitive, CLICKFORCE turned to AWS to build Lumos, a next-generation AI-driven marketing analytics solution powered by Amazon Bedrock, Amazon SageMaker AI, Amazon OpenSearch, and AWS Glue.
In this post, we show how CLICKFORCE built Lumos using AWS services to transform advertising industry analytics from weeks of manual work to an automated one-hour process.
Digital advertising challenges
Before adopting Amazon Bedrock, CLICKFORCE faced several obstacles in building actionable intelligence for digital advertising. Large-scale language models (LLMs) tend to produce general recommendations rather than actionable industry-specific intelligence. Without understanding the advertising environment, these models lacked the industry context needed to align their recommendations with real industry realities.
Another major challenge was the lack of a unified internal dataset, which reduced the reliability of the output and increased the risk of hallucinations and inaccurate insights. At the same time, marketing teams didn’t have a standardized architecture or workflow and relied on disconnected tools and techniques like vibe coding, making processes difficult to maintain and scale.
Creating a comprehensive industry analysis report was also a lengthy process, typically taking two to six weeks. The timeline resulted from multiple labor-intensive steps: 1-3 days to define objectives and set up the research plan, 1-4 weeks to collect and validate data from various sources, 1-2 weeks to perform statistical analysis and create graphs, 1-2 weeks to extract strategic insights, and finally 3-7 days to draft and finalize the report. Each stage often required back-and-forth coordination between teams, which further extended the schedule. As a result, marketing strategies were frequently delayed and based on intuition rather than insights backed by timely data.
Solution overview
CLICKFORCE was built to address these challenges. Lumosis an integrated AI-powered industry analysis service powered by AWS services.
The solution is designed around Amazon Bedrock Agents for contextual inference and Amazon SageMaker AI to fine-tune the accuracy of Text-to-SQL. CLICKFORCE chose Amazon Bedrock because it provides managed access to the underlying models without the need to build or maintain infrastructure, while also providing agents that can coordinate multi-step tasks and integrate with enterprise data sources through a knowledge base. This allowed the team to establish insights based on real, verifiable data, minimize hallucinations, and rapidly experiment with different models, while reducing operational overhead and speeding time to market.
The first step was to build an integrated AI agent using Amazon Bedrock. End users interact with a chatbot interface developed with Streamlit and running on Amazon ECS fronted by an Application Load Balancer. When a user submits a query, it is routed to an AWS Lambda function that calls Amazon Bedrock Agent. The agent retrieves relevant information from an Amazon Bedrock knowledge base built from source documents such as campaign reports, product descriptions, and industry analysis files hosted in Amazon S3. These documents are automatically converted to vector embeddings and indexed by Amazon OpenSearch Service. By basing the model’s response on this curated set of documents, CLICKFORCE ensured that the output was contextualized, less hallucinatory, and consistent with real-world advertising data.
Next, CLICKFORCE used Text-to-SQL requests to make the workflow more action-oriented. When a query required data retrieval, Bedrock Agent generated a JSON schema via the Agent Action API schema. These were passed to a Lambda Executor function that turned the requests into Text-to-SQL queries. AWS Glue crawlers continuously updated the SQL database from CSV files in Amazon S3, allowing analysts to run precise queries about campaign performance, audience behavior, and competitive benchmarks.
Finally, the company improved accuracy by incorporating Amazon SageMaker and MLflow into its development workflow. Initially, CLICKFORCE relied on an underlying model of Text to SQL conversion, which proved to be inflexible and often imprecise. Using SageMaker, the team processed data, evaluated different approaches, and fine-tuned the entire Text-to-SQL pipeline. After validation, the optimized pipeline was deployed through an AWS Lambda function and reintegrated into the agent, allowing improvements to be reflected directly in the Lumos application. MLflow provides experiment tracking and evaluation, streamlining data processing, pipeline tuning, and deployment cycles, allowing Lumos to improve query generation accuracy and deliver automated, data-driven marketing reporting.
result
The impact of implementing Amazon Bedrock Agents and SageMaker AI has been transformative for CLICKFORCE. Industry analysis that previously took two to six weeks can now be completed in less than an hour, dramatically accelerating decision-making. The company also reduced its reliance on third-party industry research reports, resulting in a 47% reduction in operational costs.
In addition to time and cost savings, the Lumos system has expanded scalability across roles within the marketing environment. Brand owners, agencies, analysts, marketers, and media partners can now generate insights independently without waiting for a centralized analyst team. This autonomy increased overall campaign agility. Furthermore, by rooting the output in both internal datasets and industry-specific context, Lumos has significantly reduced the risk of illusions and ensured that insights more closely align with industry reality.
Users can generate industry analysis reports through natural language conversations and iterate and improve content as they continue to interact.
These visual reports are generated through the Lumos system, powered by Amazon Bedrock Agents and SageMaker AI, and demonstrate the platform’s ability to generate comprehensive market intelligence within minutes. These graphs illustrate a brand’s sales distribution, retail and e-commerce performance, and demonstrate how AI-driven analytics can automate data aggregation, visualization, and insight generation with precision and efficiency.
conclusion
CLICKFORCE’s Lumos system represents a breakthrough in the way digital marketing decisions are made. By combining Amazon Bedrock Agents, Amazon SageMaker AI, Amazon OpenSearch Service, and AWS Glue, CLICKFORCE transformed its industry analytics workflow from a time-consuming manual process to a fast, automated, and reliable system. In this post, we demonstrated how CLICKFORCE built Lumos using these AWS services to transform advertising industry analytics from weeks of manual work to an automated one-hour process.
About the author
Ray Wang I’m a Senior Solutions Architect at AWS. With over 12 years of backend and consulting experience, Ray is focused on building modern solutions in the cloud, specifically NoSQL, big data, machine learning, and generative AI. A voracious worker, he passed all 12 AWS certifications, increasing the breadth and depth of his technical knowledge. He loves reading and watching science fiction movies in his free time.
shana chan I’m a solution architect at AWS. She focuses on observability in modern architectures and cloud-native monitoring solutions. Before joining AWS, she was a software engineer.
