AI in Merchant Underwriting: A TrueBiz White-Paper

AI in Merchant Underwriting: A TrueBiz White-Paper

The world of underwriting and risk analysis is on the cusp of a significant transformation driven by the rapid advancement of artificial intelligence (AI). This technological revolution, while offering significant benefits, also introduces a new set of challenges with fraud detection and prevention. The ease with which AI can be used to fabricate convincing online merchant presences and business operations is a particular concern that professionals in this space must urgently address.

We wrote this white paper to help those in underwriting and risk roles across the payments industry better understand this new innovation. There are many risks and opportunities we think this space presents that are worth sharing.


  1. How does AI work?
  2. What does this mean for fraud risk? 
  3. How AI can upgrade merchant risk assessment
  4. Benefits of using AI powered tools in risk assessment
  5. How TrueBiz leverages AI
  6. Are AI-powered tools going to replace my underwriting team?
  7. Conclusion

1. How Does AI Work?

AI is like a highly efficient assistant capable of recognizing patterns, learning from data, and making informed predictions. Generative AI, a subset of AI exemplified by large language models like ChatGPT, takes this a step further. It not only understands language but also generates new text and content that is remarkably human-like. 

Learning from Data: Just like humans learn from experiences, AI learns from data. In the payments industry, this data could be transaction histories, merchant-submitted information, or even global financial trends. The more data AI has, the better it ‘learns’ what normal looks like.

Recognizing Patterns: AI is particularly good at noticing patterns and inconsistencies. For example, it can identify behaviors that are typical for a certain customer or detect unusual behavior that might indicate fraud based on its analysis of what ‘normal’ patterns look like across its vast data set.

Making Predictions: Based on what AI learns from data it can make predictions. In risk and compliance, this might mean deciding whether a merchant is legitimate or flagging a transaction as potentially fraudulent.

One challenge with large language models is their tendency to provide inaccurate information, known as “hallucinations”. However, this risk can be mitigated by setting clear boundaries and controls for the AI. By defining the scope of its analysis and providing guidelines for its decision-making processes, we can ensure that the AI's predictions are grounded in reality and relevant data. Additionally, regular monitoring and adjustments to the AI's parameters can help maintain its accuracy and reliability in the ever-evolving landscape of merchant fraud.

This technology represents a significant shift from the traditional search engine model that we've grown accustomed to. With search engines, the expectation is to receive factually correct answers based on a corpus of web data. However, this approach has its limitations, especially when it comes to predicting future events or analyzing complex patterns in data.

In contrast, AI technologies like large language models excel at providing predictions and analyzing vast amounts of data to come to a conclusion. They are not just looking for direct matches in a database but are instead identifying patterns, trends, and anomalies that might not be immediately apparent. This capability is particularly valuable in the context of underwriting and risk assessment, where the ability to anticipate and respond to potential risks can make a significant difference in mitigating fraud and improving the merchant experience.

2. What does this mean for fraud risk? 

In the past, setting up a fake merchant website or business operations required a degree of effort and skill, making it somewhat easier to spot and thwart. AI has changed the game.

Today, a fraudster can use generative AI tools to quickly create a realistic-looking online storefront, complete with detailed product descriptions, customer testimonials, and a seemingly legitimate corporate backstory. This level of sophistication in fake digital entities makes it increasingly challenging for underwriters and risk analysts to distinguish between genuine and fraudulent operations.

For those tasked with safeguarding their company’s compliance, this new AI-driven reality requires a reevaluation of existing risk assessment methods. The traditional tools and techniques, while still important, need to be supplemented with new strategies tailored to combat AI-generated fraud.

3. How AI Can Upgrade Merchant Risk Assessment

Professionals tasked with preventing fraud, protecting consumers and the overall payments ecosystem can also leverage AI tools to stay one step ahead of malicious actors. These AI-driven tools provide enhanced capabilities in detecting and preventing fraudulent activities, offering a crucial layer of defense against the evolving techniques employed by fraudsters. By integrating AI into their arsenal, those safeguarding against fraud not only adapt to the changing landscape but also strengthen their ability to protect both consumers and the integrity of the payments ecosystem.

Deep Data Insights
AI systems excel at collecting, processing, and analyzing vast amounts of data from diverse sources such as merchant websites and their broader web presence such as social media pages and reviews. AI systems can analyze complex patterns in merchant data, far beyond what's humanly possible. For example, they can detect subtle changes in review sentiment that may indicate emerging credit risks, or identify correlations between seemingly unrelated data points that could reveal hidden risks.

Automated Risk Assessment
AI-supported risk assessments can leverage algorithms to predict future risks based on historical data patterns. By analyzing what “normal” looks like across your portfolio, systems can identify trends and vulnerabilities. This can help organizations preemptively mitigate risks before they materialize.

Real-Time Monitoring and Detection
An AI-driven web-monitoring solution  continually checks for changes and business trends across the internet. This approach allows for continuous risk and compliance monitoring on an ongoing basis, reducing the need for ad-hoc, periodic assessments. This continuous monitoring ensures that any deviations from established risk protocols or compliance standards are promptly detected and addressed, rather than waiting for a periodic review to occur. By promptly identifying and responding to such incidents, organizations can minimize potential damage and loss, enhancing their overall resilience.

4. Benefits of Using AI Powered Tools in Risk Assessment

Faster Merchant Onboarding & Verification
AI-powered tools significantly outpace human capabilities in terms of response time, allowing for almost instant identification and addressing of potential risks. This prompt response not only helps in mitigating immediate risks but also helps streamline the onboarding and verification process. In a world where merchants often prioritize fast onboarding processes, this speed ensures that legitimate merchants can be integrated into platforms quickly, without being bogged down by lengthy risk assessment processes.

Moreover, AI not only accelerates automated decision-making but also speeds up manual review cases. By quickly analyzing and presenting relevant data, AI enables underwriters to make informed decisions more rapidly, reducing the time spent on each manual review and allowing for a more efficient overall process.

Enhanced Accuracy and Reliability
AI algorithms can process vast varieties of merchant data with meticulous attention to detail, significantly reducing the inconsistency that often accompanies manual risk assessments. With the right guardrails and oversight in place, this enhanced accuracy ensures that organizations receive a more detailed and consistent understanding of their risk and compliance posture.

Cost Efficiency
Automated risk assessments powered by AI can yield substantial cost savings. While traditional manual assessments often demand significant time and resources, AI-driven assessments can be executed efficiently and at lower cost. This translates to reduced manpower requirements and lower operational expenses.

Not only does this approach reduce costs, but it also enables your team to be more productive by allowing analysts to shine in the cases where human intervention is required. By automating routine tasks, AI-powered analysis frees up the team to focus on more complex and nuanced aspects of risk assessment, where their expertise and judgment are most valuable. This shift in focus allows for a more effective allocation of human resources, enhancing the overall productivity of the team.

AI-driven risk assessments are inherently scalable, making them ideal for organizations experiencing growth or dealing with complex operational environments. Whether an organization’s data volume increases or its risk landscape becomes more intricate, AI can adapt to the evolving demands without compromising the quality or speed of assessments.

5. How TrueBiz leverages AI

TrueBiz leverages AI to revolutionize the process of reviewing a merchant’s web presence, a crucial aspect of underwriting risk assessment and ongoing risk monitoring. Our AI-driven system employs advanced algorithms to automatically scan and analyze the merchant’s web presence in real-time. This process involves evaluating hundreds of elements including content on the merchant’s website, data from social media, reviews together with local and industry profiles for the merchant. This analysis can usually take a manual analyst 10-30 minutes to complete. With TrueBiz it is completed in under 30 seconds.

By leveraging natural language processing and image recognition technologies, our AI models can discern patterns and inconsistencies that might indicate fraudulent activities or misrepresentations by the merchant. The models cross-references the information obtained from these websites with other data sources to verify their validity. Large language models enable us to accurately classify the merchant’s MCC codes, analyze their terms and conditions and assess whether their web presence appears ‘normal’.

We heavily invest in the oversight and QA process for each of these models to ensure the accuracy of the results and conduct retraining where needed.

This automated, comprehensive review enables underwriters to rapidly identify potential risks without the need for time-consuming manual analysis. Often equally important however is the ability for TrueBiz to rapidly identify clearly legitimate merchants to help streamline their underwriting experience. As a result of this approach, TrueBiz not only enhances the efficiency and accuracy of merchant assessments but also significantly accelerates the decision-making process during underwriting. 

6. Are AI-powered tools going to replace my underwriting team?

The anticipated evolution in the underwriting field is not about completely replacing human expertise with AI-powered tools, but rather about creating a powerful synergy between the two. Here's how this integration we expect this to unfold:

Augmentation, Not Replacement: AI tools like TrueBiz are designed to augment the capabilities of underwriting teams, not replace them. These tools provide underwriters with advanced analytics, data processing capabilities, and predictive insights that humans alone cannot match in scale or speed. However, the human element in understanding context, applying judgement, and making final decisions remains difficult to replace entirely. AI ‘bots’ can be treated somewhat like an employee that needs guardrails, training and fine-tuning over time.

Enhanced Decision-Making: With AI tools, underwriters can access more comprehensive analysis, leading to more-informed decisions. This doesn't diminish the role of underwriters but instead empowers them with enhanced capabilities to evaluate complex risk scenarios more effectively.

Efficiency and Productivity Boosts: AI-powered tools can handle large volumes of data and perform routine tasks much faster than humans, thereby freeing up underwriters to focus on more strategic and high-level tasks. This shift in focus can lead to higher productivity and more time for underwriting teams to concentrate on complex and nuanced aspects of risk assessment. This, in turn, creates accelerated career growth opportunities for your employees, driving retention and greater business outcomes.

7. Conclusion

As we embrace the era of AI in merchant fraud detection and underwriting, it's clear that this technology brings a wealth of benefits to the table. AI isn't just a new tool in our arsenal; we think it's a transformative force reshaping how we approach risk analysis and fraud prevention.

The real power of AI lies in its ability to rapidly sift through data and identify patterns, providing insights at a speed and depth unmatched by traditional methods. This capability is invaluable in a landscape where speed is of the essence, enabling us to quickly onboard legitimate merchants while effectively weeding out fraudulent activities. The efficiency and accuracy brought by AI not only streamline our workflows but also significantly enhance our ability to protect our companies and users.

In the context of underwriting teams, the difference between those relying solely on manual review and those enhanced with AI is becoming increasingly stark. AI-enhanced teams are pulling ahead in terms of productivity, accuracy, and overall effectiveness. The question arises: which team would you rather be on?

We have already seen evidence of fraudsters using generative AI to bypass traditional checks. In this emerging "arms race" teams that are 100% manual review are likely to fall behind. The integration of AI into risk assessment processes is no longer just an advantage; it's becoming a necessity to stay ahead of sophisticated fraudsters.

In summary, the journey with AI at the forefront promises a more robust, efficient, and secure approach to underwriting and risk analysis. As we move forward, it's the intelligent application of this technology that will continue to drive our success and safeguard the integrity of our ecosystem.

This journey is just starting, and we invite you to continue the discussion with us.

Thanks to ChatGPT, Angela Ross, Gilad Cohen, Angie Dobbs and Charle Kline for reading drafts of this white paper and contributing your thoughts.

Further reading: