Silvr: Automating business owners' access to financing with Silvr Classif.ai

Silvr—a tech scaleup focused on financing in France and Germany—built Silvr Classif.ai to automate credit application analysis and help produce financing offers adapted to a business's needs. The company extracts bank-statement data using OCR for PDF statements or open-banking APIs, centralizes transaction data in BigQuery, and uses a specifically trained Google open model to classify transactions into predefined categories. The solution supports cash-flow underwriting and near real-time financing decisions, with the company reporting that it can analyze businesses' financial health 600x faster and 300x cheaper than the manual approach.

Organization
Silvr
Industry
Finance
Location
France
Published
January 2024

Reported outcomes

600x

timeTime & speed

300xquantified impact5xcost

Strategic outcomes

Speed & agilityAutomated faster financing decisionsScale & capacityIncreased credit application processing capacityCustomer experience & trustEnabled real-time financial health monitoringNew product / capabilityAdded automated transaction categorization

Catalog median for time & speed deployments: −55% across 674 reported metrics. Compare benchmarks →

Primary read

Use case focus

Showing 3 of 3

  • 1Credit Underwriting
  • 2Document Processing
  • 3Transaction Classification
  • Credit application analysis and transaction classification were partly manual and rules-based, slowing decision-making and limiting scale.
  • Silvr needed a more targeted model to classify transaction data into 18 predefined categories for cash-flow underwriting.
  • Silvr uses Google Kubernetes Engine, Vertex AI, BigQuery, and Cloud SQL as part of its architecture.
  • It ingests bank-statement data through OCR and open-banking APIs, then centralizes the data in BigQuery for classification.
  • The company trained and deployed a Google open model on one million certified transactions to automate bank-transaction categorization and support faster financing decisions.
  • Credit application analysis is 600x faster than the manual approach.
  • The company reports classifications are 300x cheaper than the manual approach.
  • Transaction-classification costs were cut 5x compared with the first LLM tested.
  • Automation increased capacity to process more credit applications and enabled real-time monitoring of customer financial health.
Architecture

The architecture combines GKE, Vertex AI, BigQuery, and Cloud SQL. Bank-statement data is extracted through OCR from PDFs or via open-banking APIs, centralized in BigQuery, and classified by a specifically trained Google open model into predefined transaction categories.

Sources & evidence1
Groundedness: 5/5Type: Customer StoryPublished: Jan 1, 2024Publisher: Google CloudEvidence: PrimaryConfidence: High

AI-generated summary. Verify important details with the linked sources before relying on this case.

Explore related AI use cases
This website uses cookies to enhance the user experience. Learn more.