In the first post on the power of multi-model databases to lay the foundations for gen AI, we highlighted how Google Cloud Spanner helps organizations overcome some of the challenges presented by traditional approaches to database architecture and management. In this post, we dive deeper on the specific examples, across four common use cases.
We are seeing customers increasingly choose Spanner’s multi-model capabilities to address three key strategic goals:
-
A foundation of scale and reliability: Many specialized databases for graph, vector, or search, are built on traditional, single-machine architectures. As a result, they face fundamental challenges with scalability, availability, and consistency. We see customers migrate off these specialized systems because they have – or are about to – hit a wall. All Spanner’s data models are built on its tried-and-true platform offering 99.999% availability, automatic scaling, and limitless horizontal scale, and they can easily extend to new capabilities. For example, adding a vector embedding column to the existing graph schema.
-
Consolidating database sprawl and eliminating ETL: Managing, securing, and patching multiple disparate databases, each with its own data model, query language and backup policy can be an operational nightmare for users. Extract, transform, load (ETL) pipelines required to sync data, are especially frustrating as they often create inconsistency and delays. Spanner eliminates this complexity by offering multiple data models in a single unified database, eliminating extra data copies, inconsistency and management overhead. Moreover, Spanner’s interoperable multi-model capabilities allow a developer to write one SQL query that joins relational tables, traverses a graph relationship, and filters on a vector or text search function.
-
Future-proofing for evolving application needs: While many customers start with a simple application, they know it will need to get smarter and more complex over time. In Spanner, adding a graph-based recommendation or AI-powered vector search can be an afterthought. A developer can simply turn on graph or search capabilities on their operational data, with a simple data definition language (DDL) command. With Spanner, there is no painful migration, no complex re-architecturing and no growth ceiling. Instead, customers can build on a reliable relational database while seamlessly adding new, advanced data models as their application evolves.
Here’s how customers across industries are already leveraging Spanner’s evolving multi-model capabilities to solve their toughest data challenges and achieve early success:
1. Fraud detection
Fraudsters often exploit complex, non-obvious patterns across multiple transactions and accounts. Traditional relational databases struggle to detect these intricate relationships in real-time. Spanner combines relational queries with graph analytics to enable real-time pattern recognition. This allows businesses to efficiently identify suspicious clusters or unusual connections that might indicate fraudulent behavior, significantly reducing financial losses and enhancing security.
DANA: Anti-money laundering for fast growing customer base
DANA, an Indonesia-based e-wallet app, offering payments and digital financial services including lending, insurance and investments, has adopted Spanner Graph to support its critical anti-money-laundering, or AML, efforts.
-
Challenge: With a massive and still rapidly growing user base, DANA struggled to scale and meet query performance SLAs using existing relational databases to detect money laundering patterns in transactions. Moving to do the analytics in graph databases was obvious, but many graph database providers in the market simply could not handle the scale.
-
Solution: Spanner was selected after an elaborate RFP process due to its high availability, virtually unlimited scale, and external consistency. The ability to use full-text search (FTS) and vector search directly within the Graph model were key differentiators.
Palo Alto Networks: Access graph for SaaS identity
Palo Alto Networks, one of the leading cybersecurity firms, leverages Spanner to provide insights into organizational identity posture, surfacing misconfigurations and over-privileged accounts, dormant accounts, unrotated credentials, over-privileged accounts, and accounts missing in the Identity Provider (IDP).
-
Challenge: The team needed to build a world-class agent security product for the AI era that could innovate quickly while ensuring highly scalability without creating data silos.
-
Solution: They built an “Access Graph” on Spanner to connect user identities, access permissions, and the associated user activities within the SaaS applications. Spanner allows them to achieve massive scale with a single schema for both graph and non-graph use cases seamlessly.
Verisoul.ai: Real-time fake user detection
Verisoul offers a unified AI-powered platform to detect and prevent fake users, ensuring accounts are real, unique, and trustworthy.
-
Challenge: Verisoul previously built and maintained 10 different independent services across Postgres, Cassandra, and Neo4j to handle a variety of types of data, such as network intelligence, device intelligence, behavioral and sensor data, email and multi accounting. This complexity made it difficult to provide zero-latency detection to counter the speed, scale and sophistication of modern-day fraud attacks,
-
Solution: By consolidating onto Spanner, Verisoul now can monitor hundreds of customers with millions of accounts in real time, capturing every login, page view, click, and mouse move. Spanner provided an all-in-one database for Graph, vector search, and seamless integration with BigQuery, allowing them to eliminate maintenance overhead while delivering unlimited throughput all with a simple architecture.
2. Recommendation engines
Personalized recommendations are at the heart of online consumer businesses. Building an effective recommendation engine requires analyzing vast amounts of user behavior data, product and service attributes, and historical interactions. Spanner’s interoperable queries allow you to combine user profiles (relational), interaction history (search), and product similarity (graph) to generate highly relevant recommendations in real time, driving better user engagement and improving conversion rates.
Target: Combining Vector and Graph Search for gift recommendations
Target sought to elevate the holiday shopping experience with a generative AI-powered Gift Finder for highly personalized gift recommendations.
-
Challenge: The application was run on a specialized search database, providing limited gift recommendations. To enhance and personalize the experience, Target needed a sophisticated upgrade.
-
Solution: Target selected Spanner Graph for its versatile hybrid query model. The solution blends graph traversals with vector search with their proprietary embeddings delivering intuitive, real-time product suggestions — all delivered just in time for the 2025 Black-Friday-Cyber-Monday shopping rush.
True Digital Group: Consolidating AI search
True Digital Group, Thailand’s leading telecom-tech company, offers customers a wide array of high-quality digital services, encompassing both streaming and print media, along with customer loyalty tracking. Their AI-driven intelligent search feature ensures accurate content retrieval based on keywords and user intent.
-
Challenge: A fragmented stack with multiple databases resulted in outdated data, inconsistent tokenization, multiple query languages, and poor search quality, causing users to avoid the search feature.
-
Solution: True Digital consolidated all search functionality onto Spanner. By combining keyword and intent-based search results using SQL, they significantly improved search relevancy and accuracy, leading to increased customer engagement and satisfaction.
3. Hybrid search
Information retrieval is the critical bridge that grounds AI models in factual, up-to-date data and enables agentic workflow. Often, users must locate a specific needle in a haystack — searching through a massive corpus of legal documents, financial reports, or research papers. Interoperable multi-model Spanner empowers customers with hybrid search capabilities, ensuring AI models retrieve the most relevant context at any scale with pinpoint accuracy.
Rogo: Financial workflow automation
Rogo connects proprietary internal data with external financial sources like filings, PitchBook, LSEG, FactSet, and S&P Capital IQ to help finance professionals automate their workflows, from building pitch decks to drafting investment memos.
-
Challenge: Rogo needs to ingest and connect data from dozens of sources at once, across both structured and unstructured formats. Finding the right backend to support that wasn’t straightforward.
-
Solution: Rogo chose Google Cloud Spanner for its high performance, scalability, and easy management. It lets them store and query both relational and document-based data in one place, which has made it easier to audit and maintain as the platform has grown.
Inspira: Streaming legal intelligence
Inspira, a leading legal tech company, provides AI-driven solutions tailored for legal research and general workforce optimization. Their platform serves law firms, corporations, and government entities, managing a massive repository of 75 million legal documents and 440 million vectors.
-
Challenge: Before migrating to Spanner, Inspira struggled with a complex, fragmented architecture, relying on a polyglot system consisting of Elasticsearch, BigQuery and Cloud SQL. This led to complicated data synchronization, and complex “two-stage” query filtering to combine keyword and vector searches. The team also needs a path to scale beyond 1 billion vectors without sacrificing latency and high read/write throughput.
-
Solution: To resolve these inefficiencies, Inspira consolidated their entire stack intoSpanner, drastically simplifying a 4.5 TB data pipeline into a unified, high-performance single-source of truth. Leveraging Spanner’s native support for both FTS and vector search, Inspira enabled single-stage filtering for hybrid queries and achieved high-precision snippets for LLM-based legal analysis with RAG workflow.
4. Autonomous network operations
Autonomous network operations (ANO) represents the transition from reactive maintenance to predictive, self-healing networks. By creating a comprehensive digital twin of the network topology and overlaying it with real-time operational data, telecommunications providers can automate root cause analysis, predict anomalies, and resolve network incidents without human intervention.
MasOrange: The digital twin
A temporal digital twin is at the heart of MasOrange’s ANO efforts, replicating its country-wide wireless network topology, alongside operations support systems (OSS), and business support systems (BSS) data.
-
Challenge: MasOrange needed a graph database that was highly available, infinitely scalable with zero RPO/RTO to serve as the foundation of its ANO stack. They required vector, and FTS capabilities without the operational overhead of managing multiple disparate solutions.
-
Solution: MasOrange chose Spanner for its ability to meet strict scalability and availability requirements while offering fully interoperable Graph, vector and FTS capability. Today MasOrange’s digital twin is live on Spanner, powering end-to-end anomaly detection and root cause analysis.
Looking Ahead
With the scale insurance, high reliability, global consistency, and versatility in handling different data models interoperably, Spanner is a future-proof database for your agentic workload.
We envision a future where the database becomes a simple implementation detail, allowing you to focus purely on accelerating developer productivity, improving operational efficiency and delivering your business goals. Visit our Spanner page to learn more and get started today.
