Alteryx One Live Query integrates with Google Cloud BigQuery to handle complex unstructured data workloads at scale by reducing reliance on disconnected tools, minimizing data movement, and enabling warehouse-native execution. Live Query provides a browser-based, low-code environment for building data pipelines without writing code. When paired with BigQuery, it supports secure SQL pushdown, so transformation logic runs inside the warehouse instead of moving data across the network. This is particularly useful for document-heavy workflows, where Live Query can orchestrate Google AI models such as Gemini to extract intelligence from PDFs and images while keeping data governed in BigQuery.

The article’s main use case is vendor invoice processing. Finance teams receive PDFs with inconsistent formats, which traditionally forces manual extraction or a fragmented set of tools to move data into a warehouse. With Live Query for BigQuery, the full invoice lifecycle can be automated: PDFs are processed, key fields extracted, and results standardized directly in the BigQuery ecosystem. The solution compares extracted data against historical invoice records already resident in BigQuery, identifies exceptions such as duplicates or amount mismatches, and keeps governed data within the warehouse using BigQuery fine-grained access policies.

Architecturally, the pattern has three layers. Live Query in the browser is the authoring layer, where users configure extraction, classification, cleansing, validation, and routing visually. Alteryx One Platform Services coordinate request handling, execution planning, and result retrieval, adding traceability and observability across workflow execution. The customer’s Google Cloud environment is the governed data and execution layer: invoice PDFs are staged in Google Cloud Storage, historical invoice data remains in BigQuery, AI-backed extraction runs against that environment, and outputs are written back into BigQuery.

The workflow walkthrough starts with a directory input in Cloud Storage that enumerates PDF locations and returns file metadata. The Document Extract tool builds a temporary external object table from document URIs and calls either ML.PROCESS_DOCUMENT for Document AI or AI.GENERATE for Gemini; this example uses Gemini. During design, only a small subset is processed for preview, while runtime processes the full set. Next, the Classify tool uses zero-shot classification with AI.CLASSIFY to label line-item content into categories such as industrial or office supplies. Invoice numbers are normalized so formatting differences do not hide duplicates, then the workflow joins current records against historical BigQuery data to flag potential duplicates. A formula step applies business rules, such as checking whether invoice_amount matches net_amount plus tax_amount, and flags mismatches. Finally, two operational outputs are produced: one for invoices matched to historical records and flagged as duplicates, and another for unmatched, validated invoices prepared for unpaid-invoice processing.

The stated benefits are higher straight-through processing, faster reconciliation with fewer manual exceptions, stronger auditability, and reduced data movement. The result is a governed document-to-decision workflow rather than a simple extraction tool. Michael Wyant, Vice President of Enterprise Data and Corporate Solutions at Papa Johns, is quoted saying Alteryx One: Google Edition fits naturally into the Google Cloud experience and excites him about making analytics more accessible to business users through an intuitive, governed experience.