Reference architectures

Patterns, not promises.

These are the architectures we build, described in full: the components, how they fit together, and the metric each one is designed to move. They are illustrative reference patterns rather than accounts of specific engagements.

For attributable customer results on these same patterns, Google publishes detailed stories from organizations running them in production. We'll walk you through the ones closest to your environment on a call — including the parts that didn't go smoothly.

Book an architecture review Read the patterns

01 / Retail & Consumer Goods

Demand Intelligence and Agentic Inventory

Illustrative scenario: A consumer goods manufacturer running regional warehouses with inventory data split across systems that don't talk to each other.

Automated retail manufacturing warehouse with robotic arms and connected inventory systems

The challenge

Fragmented inventory data across regional warehouses and disconnected customer interaction channels led to inventory misallocations, stockouts during peak seasons, and slow time-to-market for new product deployments.

Google Cloud deployment

Gemini Enterprise Agent Platform & BigQuery: Centralized real-time supply chain analytics and customer behavioral data into BigQuery, applying custom forecasting models on the Gemini Enterprise Agent Platform.
Gemini Enterprise Agent Platform: Deployed autonomous inventory tracking agents to notify warehouse managers of demand surges in real time.
Cloud Run & Dataflow: Built serverless microservices for dynamic web search and automated order processing.

Designed to move

Forecast error: The metric this pattern is built to move
Peak throughput: Serverless scaling instead of provisioned capacity
Carrying cost: Overstock and stockout exposure

02 / Financial Services

Real-Time Risk and Agentic Defense

Illustrative scenario: A commercial bank modernizing legacy transaction monitoring, where false positives cost more in analyst hours than the fraud they catch.

Wall Street stock exchange building in the New York financial district

The challenge

Increasingly sophisticated financial crime required the bank to modernize legacy transaction monitoring. High false-positive rates strained compliance teams and impacted legitimate user transactions.

Google Cloud deployment

Agent Platform fraud detection models: Pre-trained and custom machine learning pipelines evaluating high-volume transactions in milliseconds.
Google SecOps & Security Command Center: Enterprise-grade threat detection, data residency controls, and VPC Service Controls across sensitive banking infrastructure.
Cloud Spanner: Migration to a globally distributed database guaranteeing zero-downtime ledger processing.

Designed to move

False-positive rate: Analyst hours reclaimed from noise
Decision latency: Scoring inside the transaction window
Audit traceability: Automated logging for examiner review

03 / Enterprise Modernization

Migrating Off a Legacy Estate

Illustrative scenario: A multi-division enterprise carrying multi-cloud overhead, siloed internal knowledge, and a slow path from commit to production.

Enterprise engineering team working across multiple monitors in a modern office

The challenge

Escalating multi-cloud infrastructure overhead, siloed internal knowledge repositories, and slow internal software development lifecycles across engineering divisions.

Google Cloud deployment

Google Distributed Cloud (formerly Anthos): Consolidated hybrid and multi-cloud Kubernetes cluster management into a single control plane.
Gemini Enterprise for Workspace & developers: Scaled generative AI code assistance across engineering teams and internal documentation search across all business units.
BigQuery enterprise data lake: Unified enterprise telemetry, financial operations, and HR metrics.

Designed to move

Total cost of ownership: Consolidated control plane and committed-use posture
Cycle time: Commit to production, with AI-assisted review
Time to answer: Grounded search across internal knowledge

Get started

Want the real numbers?

Google publishes verified customer results for each of these patterns, with named organizations. Bring us your environment and we'll show you the stories that actually match it — and give you an honest read on which parts of the pattern apply to you and which don't.

Email support@softpool.com Call 786-422-1697