Cloud Data Platform & AI Data Query

With analysis requests piling up, every new report took 3–5 days to build. In two phases, we built an AWS data foundation, integrated six data types, rebuilt the buyer prediction model and introduced an Amazon Quick Chat Agent—so staff ask in plain language and get grounded answers and charts in seconds.

Background & challenges

• Buyer engagement and value prediction models needed to be rebuilt and migrated to a new cloud account.
• Data was spread across existing databases, Google Sheets, ad campaigns and third-party trade data, each with different formats and update frequencies.
• The existing RFM model couldn’t distinguish large enterprise buyers, and every new analysis request meant a new report—about 3–5 days each.

Our approach

Following Datas’ four-step method: get the data right first, then deliver the application, then keep it running.

1. Data audit: Audited six data types—product search volume, buyer classification and lead quality, online exhibitions, import/export trade, ad campaigns and video meetings—including formats, update frequencies and sensitive fields.
2. Build the platform: Set up a new cloud account and rebuilt Amazon Redshift and existing databases; structured data loads directly and form-based data is cleaned first, with dedicated datasets/views for AI queries whose field aliases match business language.
3. AI applications: Rebuilt the buyer engagement and value models, adding a “large enterprise” flag and engagement frequency to RFM, and enabled an Amazon Quick Chat Agent so staff can query data in natural language and get summaries and simple charts.
4. Operate & optimize: Role-based permissions with personal and sensitive fields masked or excluded; acceptance against agreed representative questions, plus table lists, field descriptions and data-flow documentation so every answer is verifiable.

Results

• Both phases are complete: cloud data foundation, model rebuild, six-source integration and AI queries are all live.
• New questions no longer wait for report development—staff ask in natural language and get answers and simple charts within seconds.
• The RFM model now identifies large enterprise buyers, so resources go to the highest-value accounts first.
• Every answer traces back to tables and field definitions, so results are verifiable.

Related services & next steps

• AI Assistant: the Chat Agent in this project—natural-language queries over enterprise data with traceable answers.
• Predictive & Segmentation Models: the rebuilt buyer value model, RFM plus a large-enterprise flag and engagement frequency.
• Data Platform & Data Engineering: the Redshift cloud data foundation and AI-ready datasets.