Silvr: Automating business owners' access to financing with Silvr Classif.ai
About Silvr
Silvr—a tech scaleup focused on financing—is creating a world in which the leaders of European VSB/SMB to make business financing more accessible. With a presence in France and Germany, the company analyzes business data to produce a financing offer most adapted to a business's needs, offering financing in 48 hours.
Tell us your challenge. We're here to help.
Contact usBy adopting Google's open model, Silvr built a cutting edge tool to automate and considerably accelerate its credit application analysis process, while reducing its operational costs. Now more competitive and responsive, the company intends to capitalize on this transformation to realize its ambitions of expanding and developing new services.
Google Cloud results
- Significantly accelerated decision-making processes
- Drastically reduced operational costs associated with transaction classification
- Reduced environmental footprint thanks to a more targeted and less resource-intensive AI model
- Increased capacity to process a larger number of credit applications, thus supporting the company's expansion
- Contributing to innovation by making business financing more accessible and faster
Analyzing businesses' financial health 600x faster and 300x cheaper
Cofounded by Greg Tappero and Nima Karimi in 2020, Silvr set itself the goal of simplifying the financing of European VSB/SMB. Addressing the difficulties in securing traditional financing encountered by business owners, Silvr offers them a simple and flexible alternative. This innovative approach enables businesses to secure financing quickly and without personal guarantees from business leaders.
Since its launch, Silvr has chosen to rely on Google Cloud to construct its IT system. Most notably, the company developed a rating algorithm to evaluate credit applications based on cash flow underwriting, nurtured by a few data sources like various data sources provided by businesses, such as professional bank account statements or balance sheets. The solution's technical architecture is based on a combination of services, including Google Kubernetes Engine (GKE), Vertex AI, BigQuery, and Cloud SQL.
Accelerating application processing to unlock growth
Reliable, high-performing, and scalable, the existing solution already enabled the rapid analysis of a wide variety of customer data, as well as informed, fast, and precise decision-making on financing. To go even further and become even faster to support yet more businesses, the company decided to strengthen its analytical processes with an open model, a deep learning platform developed by Google and specifically trained for Silvr's specific use.
In practice, Silvr mainly relies upon the analysis of a business's bank transactions for decision-making, extracting data from bank statements using OCR when they arrive in PDF format or using the open banking standard, which enables it to directly obtain data from APIs. The data is then centralized in BigQuery for classification purposes, the objective being to automatically categorize each bank transaction according to type (revenue, expenditure, salaries, taxes, and more) to get a clear view of the business's financial situation and to project its potential.
Until recently, part of this classification was still carried out manually and then from a rules engine, which slowed the decision-making process and limited the scope of the credit applications that Silvr's teams could process and also very specific to a country. To overcome these obstacles and provide itself with resources in line with its growth ambitions, the company turned to AI. "Our aim is to automate the classification of bank transactions using AI in order to increase our processing capacity so we can handle more credit applications and implement new services, such as cash flow management," explains Greg Tappero, Silvr Cofounder and CTO.
"Our aim is to automate the classification of bank transactions using AI in order to increase our processing capacity so we can handle more credit applications and implement new services."
—Greg Tappero, Silvr Cofounder and CTOSpecifically trained Google's open model for faster, more relevant, and more economical classification
To begin with, Silvr tested generative AI available on the market to create its tool, known as Silvr Classif.ai. But as Habib Dhif, Silvr's Data Engineer, explains, it was too big for the company's needs: "Our need was limited to understanding the data, then classifying it into our 18 predefined categories. We didn't need infinite parameters, as the diversity of bank transactions is relatively limited, nor did we need creative or conversational capabilities. In other words, the first model tested, an LLM, was both too complex and unnecessarily large and creative in terms of our needs."
In fact, despite a simpler architecture, with fewer parameters compared to more recent LLMs, Google's open model has turned out to be more effective for Silvr's classification needs, thus demonstrating that a less complex but more targeted model was the optimal solution for its specific application.
Thanks to this very pragmatic approach, Silvr was able to considerably reduce its costs and the environmental footprint of Silvr Classif.ai—a targeted, simpler model that is less resource-intensive by nature. This is especially true since the company covers French and German credit applications with a single model, trained with one million certified transactions. "As a result, in relation to the first LLM tested, we divided the costs associated with transaction classification by five with Google's open model," states Greg Tappero. "And compared to the manual approach we started with, we can perform our classifications 600x faster and 300x cheaper."
"Being less complex to manage, Google's open model also allows us to iterate more often, thus promoting accelerated development cycles that benefit operational efficiency," adds Habib Dhif.
Google's open model—an open-source model—also enables Silvr to ensure data processing and analysis are carried out in accordance with the regulations in force, a fundamental point for a financial institution subject to strict legislation. "Plus, unlike other competing proprietary models, Google's open model doesn't limit the number of queries or the volume of exchanges a business can perform in a given time frame, and the underlying infrastructure isn't constrained. We model the resources according to our needs," remarks Habib Dhif.
"In relation to the first LLM tested, we divided the costs associated with transaction classification by five with Google's open model. And compared to the manual approach we started with, we can perform our classifications 600x faster and 300x cheaper with Google's open model."
—Greg Tappero, Silvr Cofounder and CTOIntelligent automation that makes a difference
A compelling example of how artificial intelligence can transform business and technical operations in the financial sector, Silvr now relies on an entirely digitized and automated process that enables it to respond to financing applications in near real time. "Subsequently, businesses can complete an application online and get a response very quickly," explains Greg Tappero. "Similarly, thanks to automation, we can perform the real-time monitoring of our customers' financial health and continuously reevaluate their situation, for example, offering them new financing when their existing loans mature, if their financial creditworthiness is sufficient. In fact, we've reached an important milestone in terms of operational efficiency that we're going to leverage in order to innovate and offer new services to businesses."
Adopting Google's open model to bring Silvr Classif.ai to life, and thus being able to digitize and automate the entire credit application analysis process, marks a huge turning point in Silvr's history. Improving risk management along with reducing operational costs provides the company with a competitive edge in a very competitive market. At the same time, its approach also demonstrates that to stand out, innovation is not always enough. Ultimately, the way in which services are designed and delivered counts just as much as the ability to create services. In other words, AI creates new potential, but you still have to know how to choose the models most adapted to your needs to find an effective, profitable solution that benefits the company and its customers.
"Thanks to AI, we've marginalized the production costs of short-term loans. In fact, we've reached an important milestone in terms of operational efficiency that we're going to leverage in order to innovate and offer new services to businesses."
—Nima Karimi, Silvr Cofounder and CEOTell us your challenge. We're here to help.
Contact us
About Silvr
Silvr—a tech scaleup focused on financing—is creating a world in which the leaders of European VSB/SMB to make business financing more accessible. With a presence in France and Germany, the company analyzes business data to produce a financing offer most adapted to a business's needs, offering financing in 48 hours.