Case study
Vaden Automotive gets faster reports and AI answers at a fraction of the cost with GizmoSQL
Vaden Automotive runs GizmoData Cloud at Customer inside its own AWS account. Data replicated from Snowflake into GizmoSQL powers two things: an MCP server that lets employees ask business questions in plain English through Claude, and internal real-time analytics apps built with the GizmoData App Factory. Reports and AI answers are faster, and costs have dropped dramatically.
- Customer
- Multi-franchise new and used car dealership group serving Savannah, Georgia and the South Carolina Lowcountry
- Website LinkedIn
- Industry
- Automotive retail
- Headquarters
- Savannah, Georgia
- Deployment
- GizmoData Cloud at Customer on AWS
- Products
- GizmoData Cloud at Customer
- GizmoSQL Enterprise
- GizmoSQL MCP Server
- GizmoData App Factory (early access)
How the data flows
One GizmoSQL deployment in Vaden Automotive’s own cloud account serves both Claude and the internal apps.
Snowflake
System of record. Data is replicated into GizmoSQL.
Vaden Automotive’s AWS account · GizmoData Cloud at Customer
GizmoSQL
DuckDB-powered Arrow Flight SQL clusters on Graviton nodes with NVMe, managed by the GizmoData Nexus control plane in the same account.
Claude + GizmoSQL MCP Server
Employees ask questions in plain English, and Claude answers them by querying GizmoSQL.
App Factory apps
Internal web apps with single sign-on and per-app read-only credentials, showing real-time analytics.
The goal: fast answers from data already in Snowflake
Vaden Automotive’s data lives in Snowflake. The team wanted employees to get answers from it in two ways: by asking questions in plain English through an AI assistant, and through internal web apps built for specific jobs.
Both workloads send many small queries and need answers in a fraction of a second, and running all of them on the warehouse is slow and expensive. Vaden wanted a fast, low-cost SQL engine for them, running inside its own AWS account.
GizmoData Cloud at Customer, inside Vaden’s AWS account
The GizmoData Nexus control plane and Vaden’s GizmoSQL clusters run in Vaden’s own AWS account, on the Kubernetes (EKS) cluster that already hosts its internal apps. Data sits in Vaden’s S3 buckets, encrypted with Vaden’s own KMS keys.
Data is replicated from Snowflake into GizmoSQL. Snowflake remains the system of record, and GizmoSQL, which is built on DuckDB and speaks Arrow Flight SQL, serves the interactive queries. The GizmoSQL clusters run on AWS Graviton nodes with local NVMe storage.
Plain-English questions through Claude
Employees ask business questions in Claude. Claude answers them by querying GizmoSQL through the GizmoSQL MCP server: it lists the catalogs, inspects the tables and runs SQL, then answers with the results.
Because GizmoSQL answers quickly, Claude can run several queries for one question and still reply promptly.
- Read-only by default: the MCP server will not run writes unless they are explicitly enabled.
- Row caps and query timeouts are enforced on the server, so one question cannot run away with the cluster.
- Each connection uses its own GizmoSQL credentials, so GizmoSQL’s permissions decide what Claude can see.
Real-time analytics apps from the GizmoData App Factory
Vaden also builds internal analytics apps with the GizmoData App Factory, which is in early access. An employee requests an app by opening an issue in GitHub. After a platform owner approves it, the factory creates the app’s repository from a template web app with Microsoft single sign-on, sets up development and production environments, and posts the development URL back on the request.
Each app connects directly to GizmoSQL, so its dashboards show live data. The factory gives each app its own GizmoSQL credentials.
- Each app gets separate credentials for development and production. They are read-only and limited to one catalog.
- Credentials expire after 90 days and are rotated automatically every month.
- The key that signs GizmoSQL tokens is never stored in GitHub, and a person still approves each new app.
Full visibility into the engine with GizmoData Nexus
Vaden’s team sees how the database engine is performing in GizmoData Nexus. For each cluster, Nexus shows CPU and memory use against the cluster’s limits, along with the queries the engine runs: how many, how long they take and which ones fail.
With these numbers, the team can size clusters to the actual workload and find slow queries before users notice them.
The results
Since moving these workloads to GizmoSQL, Vaden has seen clear improvements:
- Faster reports: dashboards and internal apps query GizmoSQL directly and return quickly.
- Faster AI answers: Claude gets query results back from GizmoSQL quickly, so plain-English questions are answered faster.
- Dramatically lower cost: the interactive queries from people, apps and Claude run on GizmoSQL instead of the warehouse.
- Full visibility: Nexus shows CPU, memory and query performance for every cluster.
Frequently asked questions
What does Vaden Automotive use GizmoSQL for?
Vaden Automotive replicates data from Snowflake into GizmoSQL, which it uses as a fast SQL engine for two workloads: natural-language questions from employees through Claude, using the GizmoSQL MCP server, and real-time internal analytics apps built with the GizmoData App Factory.
Where does Vaden’s GizmoSQL deployment run?
In Vaden’s own AWS account, through GizmoData Cloud at Customer. The Nexus control plane and the GizmoSQL clusters run on Vaden’s existing EKS cluster, and data stays in Vaden’s S3 buckets, encrypted with Vaden’s own KMS keys.
Why replicate Snowflake data into GizmoSQL instead of querying Snowflake directly?
An AI assistant and real-time dashboards send many small queries that each need a fast answer. Serving them from GizmoSQL, a DuckDB-based engine running on dedicated clusters, keeps them fast and keeps that traffic off the warehouse. Snowflake remains the system of record.
What results has Vaden Automotive seen with GizmoSQL?
Reports and AI answers through Claude are faster, and costs are dramatically lower, because the interactive queries run on GizmoSQL instead of the warehouse. GizmoData Nexus also gives the team full visibility into the engine’s CPU, memory and query performance.
What is the GizmoData App Factory?
The App Factory, currently in early access, turns an approved request into a deployed internal web app. It creates the app’s repository from a template with single sign-on, sets up development and production environments, and gives the app its own read-only GizmoSQL credentials, which are rotated automatically.