Where and How to Use AI for Lead Verification

Given the current hype around AI, specifically GenerativeAI and ChatGPT, we are often asked, ”How can we use GenAI to improve our lead verification processes?”

Our high-level summary:

Machine Learning algorithms, from supervised to unsupervised learning, should suffice for any decision around fraud, fake leads, or bots or even to prioritize the right leads and signups. There should not be any need for anything further sophisticated.

Beyond these decision requirements, we believe using GenAI (Generative AI, the technology behind ChatGPT) based tools can be a great copilot for analytics and many business operations involved in lead verification and management. These tasks include everything from data analysis to finding patterns, creating rules within your lead management platforms, accelerating manual tasks such as data entry for compliance, and other operational and administrative tasks.

GenAI will need to solve a few critical data quality and governance challenges to evolve beyond copilots. The ideal combination is human + rules + ML + GenAI.

How can Machine Learning be used for Verification and Prioritization?

Machine Learning (ML) is a subset of AI (Artificial Intelligence). Machine learning is specifically used to create algorithms to learn from and make predictions or decisions based on data. There are multiple methodologies here, but supervised learning, a machine learning technique that uses labeled data to train algorithms and predict outcomes, is most relevant in our space. These machine learning-based systems can predict the likelihood of fraud, fake leads, or bots with high accuracy by using a mix of historical and real-time data. For supervised learning, we see the usage of Gradient Boosting and XGBoost decision tree algorithms providing the best value for the use cases above. 

Machine learning models must be trained and refined using “features” to do this effectively. Consider features to include everything from the device, behavior on the site, and third-party data like phone type, demographic data, etc. These attributes and predictors teach the algorithms to recognize patterns and make predictions. Depending on the answer being looked for, these feature sets can change. For example, a simple phone risk score would suffice with phone validity, phone line type, activity (how active it has been), if the name on the signup form matches the phone subscriber name, etc. However, the feature set might be expanded for propensity to buy, including demographics, purchase behavior, and more. Today, machine learning does a lot of the heavy lifting when detecting fraud patterns, spotting anomalies, and determining risk scores. Machine learning-based systems are much more effective at responding to changing threats in real-time, which is why most organizations are either already using or looking into using machine learning for fraudulent leads and bots.

How can Generative AI be Used?

Generative AI is still a relatively new technology for identity verification, fraud prevention, and lead verification. Today, it’s best suited as a fraud and compliance operations copilot. We believe GenAI’s “low-hanging fruit” is the copilot for verification and compliance. Organizations spend 10% to 30% of their headcount on compliance-related activities. Today, the operational and compliance teams are trying to keep up with the ever-higher workload of keeping a good stream of quality leads while adhering to compliance and regulations. Much of the work is manual, and these tasks are often repetitive data summaries or trawling through large data sets. This forces them to focus on only the worst cases. These are the places where GenAI and GPT interfaces can significantly reduce the manual efforts involved and improve organizational productivity. Typical tasks where a copilot can be of great help:

  • Managing compliance through D10LC campaigns
  • Analyzing and summarizing sizeable unstructured data sets
  • Paperwork, reasoning/evidence for returned leads, and tracking for the returns
  • Cleaning up datasets before ingesting them into systems (e.g., from CRMs to automated dialers)
  • Analytics to check if specific suppliers are trending with good versus consistent bad leads
  • Helping you create rules in your lead management system
  • Answering questions about why a rule has been fired

A large language model can perform up to 90% of the dull administrative work, allowing the teams to spend more time improving lead quality and compliance. GenAI is transformative in many domains, such as marketing copy text generation, images, audio, and video generation. However, we are still in the early days of developing this technology, and they are also expensive and slow. Also, GenAI is not as strong at raw number crunching and statistical modeling. However, one of the most common and significant issues of GenAI today is AI hallucination, where these models perceive patterns or objects that are nonexistent or imperceptible to human observers, creating nonsensical or inaccurate outputs. This is true today for almost all the models, from ChatGPT to open-sourced Llama (by Meta). These hallucinations severely limit its use cases feasible today, especially in production, large-scale deployments. 

While all these issues will improve over time, for the foreseeable future, existing model types are the most efficient tools for verification and fraud detection purposes. Supervised and unsupervised machine learning models work well for most lead verification/identity verification needs. They’re naturally much stronger at dealing with large statistical models and giving answers with higher confidence.

The Potential for Machine Learning in Lead Management

Another thing we observe is most of the lead verification efforts done by marketing operations have barely scratched the surface of what machine learning can do. Many businesses are still working with bulky rule systems, and there is a great opportunity to up-level their decisions with machine learning models. Even when models are used, most are small, rely on limited data sets, and the teams are hampered by poor access to external data. Also, while still significant, engineering and data science costs are considerably lower than just a few years ago. With the advent of modern tools, they are more accessible for most organizations. Investments in good quality data at scale, engineering teams with the requisite skills, and access to necessary tools can uplevel an organization’s capabilities in using ML models that can positively impact their lead management and compliance operations.

This, combined with ever-improving GenAI models, will be a tremendous asset for any organization. These technologies will continue to surprise us with the accelerated pace of improvements in creativity and problem-solving. But we don’t think we’ll have to worry about being replaced by AI. These copilots will simply make us more productive and free up our time and resources to focus on higher-value tasks.

It’s clear that machine learning and GenAI will play crucial roles in the lead verification space. What do you think? Where else can GenAI help with lead verification and management efforts for businesses?