During latest joint SME Banking Club and AdviceRobo webinar “From Manual Underwriting to AI in SME Credit & Fraud”, Advicerobo demonstrated how banks and SME lenders can combine behavioral AI, alternative data, fraud analytics, and automation to improve credit decisions. Rather than replacing existing scorecards and underwriting processes, AdviceRobo adds new signals that help lenders better understand borrowers, identify risk and fraud earlier, reduce unnecessary manual reviews, and make faster and more informed decisions.
Diederick Van Thiel (CEO at AdviceRobo) and Rosali Steenkamer (CCO at AdviceRobo) emphasized that traditional SME credit assessment often struggles with limited financial history, fragmented data, manual reviews, and increasing fraud risks.
Below are the four key use cases presented during the webinar and their implications for SME lenders.
-
Synthetic SME Data Lakes for Market Expansion and Model Development
The first use case focused on synthetic SME data. AdviceRobo helps lenders address a common challenge when entering new markets or developing new risk models: limited historical data.
To support model development and testing, AdviceRobo provides a synthetic SME data environment built from a data lake of approximately 120,000 SME profiles, incorporating different types of financial, behavioural and alternative data.
The webinar highlighted the example of Dutch SME lender Finom, which wanted to expand into Germany. With limited historical data available in the new market, Finom used AdviceRobo’s synthetic data environment to support model development and validation before deployment.
The environment incorporated different feature types, including psychometric, biometric, email, mobile-phone and loan-related data. These additional data sources provided an environment for developing and testing risk models before entering the new market.
How this helps banks reduce risk
For lenders expanding into new SME segments or geographies, synthetic data can provide a controlled environment to:
- Test underwriting policies before deployment.
- Support model validation while reducing reliance on sensitive customer data.
- Identify potentially useful variables in unfamiliar markets.
- Test different scenarios and decision thresholds.
- Reduce uncertainty when launching new lending products.
Rather than entering a new market based mainly on assumptions, lenders can test and stress-test models, thresholds and credit policies before taking real credit exposure.
-
AI-Powered Financing Inovice and Cross-Border SME Risk Assessment
The second use case focused on invoice financing, where lenders often struggle to assess both the borrower and the buyer behind the invoice.
This challenge becomes particularly important in cross-border transactions, where traditional credit information may be limited or difficult to verify.
AdviceRobo demonstrated how lenders can use digital and behavioural signals alongside traditional information to build a broader view of both fraud and credit risk. These signals can include:
- Email and digital identity signals.
- Behavioural characteristics and pshycometric indicators.
- Digital footprint and company information.
- Domain and website information.
- Network and connection signals.
For example, the solution can determine whether an email address has been involved in serius data branches, whether a company website was created recently, and whether several digital signals indicate a trustworthy business. These signals are them translated into fraud and credit scores.
How this helps banks reduce risk
In SME lending and trade finance, this approach can help lenders:
- Strengthen the assessment of counterparties in cross-border transactions.
- Identify potential risk indicators before financing invoices.
- Identify fraud risks that traditional credit data may not capture.
- Assess SMEs where traditional financial information is limited.
Together, these additional signals can give lenders a broader view of the borrower and its counterparties, supporting more informed credit and fraud decisions when traditional information alone is not enough.
-
Behavioural Credit Scoring for Higher Approvals and Lower Defaults
The third use case presented a real-life implementation at a lender in the Czech Republic.
AdviceRobo added behavioural variables to the lender’s existing underwriting process for micro-enterprises and sole traders. The objective was not to replace existing risk models, but to enrich them with additional behavioural intelligence.
The results for micro-enterprises were significant:
- Acceptance rates increased from approximately 19% to 42%.
- At the same time, default rates decreased from around 7% to approximately 5%.
The case demonstrates how behavioural data can provide additional insight into borrower characteristics and risk patterns that may not be visible in traditional financial data alone.
How this helps banks reduce risk
This approach can be particularly relevant for underserved SMEs and thin-file borrowers, where traditional information may provide only a limited view.
Behavioural intelligence can help lenders:
- Identify additional creditworthy SMEs that traditional models may overlook.
- Increase approvals while maintaining control over risk.
- Strengthen assessments where traditional credit information is limited.
- Improve the consistency of credit decisions.
Rather than replacing traditional credit models, behavioural scoring can provide an additional predictive layer, helping lenders pursue higher approval rates while maintaining control over credit risk.
-
Fraud Detection and Credit Policy Optimization
The fourth use case combined two important capabilities: fraud detection and credit policy optimization.
First, AdviceRobo demonstrated how Digital Intelligence can help identify potential synthetic identities and other fraud risks during onboarding.
The approach combines application data with digital identity, behavioural and network signals to assess the level of trust and identify anomalies or connections that may indicate higher fraud risk. This can help lenders identify higher-risk or suspicious applications before financial exposure occurs.
The webinar also included a demonstration of the Credit Policy Optimizer, a tool that allows lenders to simulate and compare different lending strategies and evaluate the trade-offs between:
- Approval rates.
- Default rates.
- Risk appetite.
- Expected profitability.
Risk and commercial teams can use these simulations to compare policy scenarios and identify strategies that are expected to perform better within defined risk and business constraints.
How this helps banks reduce risk
By combining fraud intelligence and policy optimisation, lenders can:
- Identify potential synthetic identity and fraud risks before loan approval.
- Support earlier fraud detection and help reduce fraud exposure.
- Test approval thresholds against risk and profitability objectives.
- Support more consistent, data-informed credit policies.
- Reduce unnecessary manual underwriting and focus human review on higher-risk or unclear cases.
Together, these capabilities help lenders evaluate the trade-offs between growth, risk and profitability before putting a credit strategy into practice.
Conclusion
The four use cases presented during the webinar show how AI and additional data sources can support different stages of SME lending:
- Synthetic SME Data Lakes can support model development and testing when historical data is limited, particularly when entering new markets.
- Digital Identity and Behavioural Intelligence for Invoice Finance can provide additional context when assessing borrowers and counterparties, especially in cross-border transactions.
- Behavioural Credit Scoring can complement existing risk models and help identify creditworthy SMEs that traditional data alone may overlook.
- Fraud Detection and Credit Policy Optimization can help lenders identify fraud risks earlier and test different credit strategies against risk, approval and profitability objectives.
The common theme across these use cases is not replacing existing underwriting processes with AI. It is about adding new sources of intelligence to existing data and models to help lenders make more informed credit and fraud decisions.
By combining traditional credit information with behavioural and digital signals, lenders can build a broader view of SMEs, test strategies before deployment, focus manual review where it adds value, and maintain greater control over their lending decisions.
If you’re looking to transform your SME lending strategy today, reach out to AdviceRobo to learn how they can support and accelerate that transformation.
