CLICKFORCE is one in all leaders in digital promoting companies in Taiwan, specializing in data-driven promoting and conversion (D4A – Information for Promoting & Motion). With a mission to ship industry-leading, trend-aligned, and revolutionary advertising options, CLICKFORCE helps manufacturers, companies, and media companions make smarter promoting selections.
Nevertheless, because the promoting {industry} quickly evolves, conventional evaluation strategies and generic AI outputs are not adequate to offer actionable insights. To stay aggressive, CLICKFORCE turned to AWS to construct Lumos, a next-generation AI-driven advertising evaluation resolution powered by Amazon Bedrock, Amazon SageMaker AI, Amazon OpenSearch, and AWS Glue.
On this submit, we exhibit how CLICKFORCE used AWS companies to construct Lumos and rework promoting {industry} evaluation from weeks-long handbook work into an automatic, one-hour course of.
Digital promoting challenges
Earlier than adopting Amazon Bedrock, CLICKFORCE confronted a number of roadblocks in constructing actionable intelligence for digital promoting. Giant language fashions (LLMs) have a tendency to provide generic suggestions quite than actionable industry-specific intelligence. With out an understanding of the promoting setting, these fashions didn’t have the {industry} context wanted to align their strategies with precise {industry} realities.
One other important problem was the absence of built-in inside datasets, which weakened the reliability of outputs and elevated the danger of hallucinated or inaccurate insights. On the similar time, advertising groups relied on disconnected instruments and approach reminiscent of vibe coding, with out standardized architectures or workflows, making the processes troublesome to keep up and scale.
Getting ready a complete {industry} evaluation report was additionally a time-consuming course of, usually requiring between two and 6 weeks. The timeline stemmed from a number of labor-intensive phases: one to a few days to outline targets and set the analysis plan, one to 4 weeks to collect and validate information from completely different sources, one to 2 weeks to conduct statistical evaluation and construct charts, one to 2 to extract strategic insights, and eventually three to seven days to draft and finalize the report. Every stage usually required back-and-forth coordination throughout groups, which additional prolonged the timeline. Because of this, advertising methods have been often delayed and based mostly extra on instinct than well timed, data-backed insights.
Options overview
To deal with these challenges, CLICKFORCE constructed Lumos, an built-in AI-powered {industry} evaluation service, utilizing AWS companies.
The answer is designed round Amazon Bedrock Brokers for contextualized reasoning and Amazon SageMaker AI for fine-tuning Textual content-to-SQL accuracy. CLICKFORCE selected Amazon Bedrock as a result of it gives managed entry to basis fashions with out the necessity to construct or preserve infrastructure, whereas additionally providing brokers that may orchestrate multi-step duties and combine with enterprise information sources by way of Data Bases. This allowed the workforce to floor insights in actual, verifiable information, decrease hallucinations, and shortly experiment with completely different fashions, whereas additionally lowering operational overhead and accelerating time-to-market.
Step one was to construct a unified AI agent utilizing Amazon Bedrock. Finish-users work together with a chatbot interface that runs on Amazon ECS, developed with Streamlit and fronted by an Utility Load Balancer. When a person submits a question, it’s routed to an AWS Lambda perform that invokes an Amazon Bedrock Agent. The agent retrieves related data from a Amazon Bedrock Data Bases, which is constructed from supply paperwork—reminiscent of marketing campaign stories, product descriptions, and {industry} evaluation information—hosted in Amazon S3. These paperwork are mechanically transformed into vector embeddings and listed in Amazon OpenSearch Service. By grounding mannequin responses on this curated doc set, CLICKFORCE made certain that outputs have been contextualized, diminished hallucinations, and aligned with real-world promoting information.
Subsequent, CLICKFORCE made the workflows extra action-oriented through the use of Textual content-to-SQL requests. When queries required information retrieval, the Bedrock Agent generated JSON schemas by way of the Agent Actions API Schema. These have been handed to Lambda Executor features that translated requests into Textual content-to-SQL queries. With AWS Glue crawlers repeatedly updating SQL databases from CSV information in Amazon S3, analysts have been capable of run exact queries on marketing campaign efficiency, viewers behaviors, and aggressive benchmarks.
Lastly, the corporate improved accuracy by incorporating Amazon SageMaker and MLflow into the event workflow. Initially, CLICKFORCE relied on basis fashions for Textual content-to-SQL translation however discovered them to be rigid and sometimes inaccurate. By utilizing SageMaker, the workforce processed information, evaluated completely different approaches, and tuned the general Textual content-to-SQL pipeline. As soon as validated, the optimized pipeline was deployed by way of AWS Lambda features and built-in again into the agent, ensuring that enhancements flowed straight into the Lumos utility. With MLflow offering experiment monitoring and analysis, the cycle of knowledge processing, pipeline tuning, and deployment grew to become streamlined, permitting Lumos to attain increased precision in question technology and ship automated, data-driven advertising stories.
Outcomes
The impression of adopting Amazon Bedrock Brokers and SageMaker AI has been transformative for CLICKFORCE. Business evaluation that beforehand required two to 6 weeks can now be accomplished in underneath one hour, dramatically accelerating decision-making. The corporate additionally diminished its reliance on third-party {industry} analysis stories, which resulted in a 47 % discount in operational prices.
Along with time and value financial savings, the Lumos system has prolonged scalability throughout roles throughout the advertising setting. Model house owners, companies, analysts, entrepreneurs, and media companions can now independently generate insights with out ready for centralized analyst groups. This autonomy has led to better agility throughout campaigns. Furthermore, by grounding outputs in each inside datasets and industry-specific context, Lumos considerably diminished the danger of hallucinations and made certain that insights aligned extra carefully with {industry} realities.
Customers can generate {industry} evaluation stories by way of pure language conversations and iteratively refine the content material by persevering with the dialogue.
These visible stories, generated by way of the Lumos system powered by Amazon Bedrock Brokers and SageMaker AI, showcase the platform’s potential to provide complete market intelligence inside minutes. The charts illustrate model gross sales distribution, retail and e-commerce efficiency, and demonstrating how AI-driven analytics automate information aggregation, visualization, and perception technology with excessive precision and effectivity.
Conclusion
CLICKFORCE’s Lumos system represents a breakthrough in how digital advertising selections are made. By combining Amazon Bedrock Brokers, Amazon SageMaker AI, Amazon OpenSearch Service, and AWS Glue, CLICKFORCE reworked its {industry} evaluation workflow from a sluggish, handbook course of into a quick, automated, and dependable system. On this submit, we demonstrated how CLICKFORCE used these AWS companies to construct Lumos and rework promoting {industry} evaluation from weeks-long handbook work into an automatic, one-hour course of.
Concerning the Authors
Ray Wang is a Senior Options Architect at AWS. With 12+ years of expertise within the backend and advisor, Ray is devoted to constructing trendy options within the cloud, particularly in particularly in NoSQL, huge information, machine studying, and Generative AI. As a hungry go-getter, he handed all 12 AWS certificates to extend the breadth and depth of his technical information. He likes to learn and watch sci-fi motion pictures in his spare time.
Shanna Chang is a Options Architect at AWS. She focuses on observability in trendy architectures and cloud-native monitoring options. Earlier than becoming a member of AWS, she was a software program engineer.
