While ChatGPT and other similar generative AI engines are relatively new technologies, bad actors are already using them to automate and perpetuate fraudulent attacks. Scammers can use generative AI to produce more effective phishing emails, for example, without the grammar and spelling inaccuracies that typically give them away. Fraudsters could also use generative AI to produce more authentic-looking fake documents like bank statements, accounting documentation, or other documents required during an audit. 

The actual impact of generative AI in tax-related scams will take time to fully unravel, but it’s critical that tax agencies act now to both understand how this technology could present new threats and how it can be leveraged to prevent them.  

In fact, generative AI has the potential to create tremendous value for tax and revenue agencies. Applied to the right use cases, these capabilities could enable process automation, enhanced decision-making, increased compliance, and better taxpayer experiences.  

Generative AI can help agencies more efficiently identify non-filers, non-compliance, and fraud. The models can analyze large amounts of data from multiple internal systems and external data resources to identify patterns that are indicative of identity theft, for example. Automation can be put into place to help train models on a timelier basis to keep pace with the bad actors that will also use the technology to improve their schemes and tactics. Moreover, natural language models can allow non-technical users to analyze data more efficiently. Users could simply ask questions to a natural language interface instead of wading through data aand then designing summarized reports and spreadsheets. 

Drastically reduce the amount of time agency staff spend reviewing and responding to taxpayer documents by leveraging generative AI to augment rote functions. Many documents submitted by taxpayers, like Identity Theft affidavits, include open ended text-fields that require intensive manual review, resulting in a backlog of cases that need to be reviewed, classified, and followed up on. Generative AI models can automatically extract meaning from documents, summarize information, and help examiners determine the best next steps. They can also help draft responses to inquiries or correspondence from taxpayers. 

Generative AI could also be leveraged to create or enhance virtual assistants (e.g., voice bots, chat bots) to provide more personalized support, answer tax-related questions, guide users through complex processes, and promote voluntary compliance. It could even analyze open-ended text to flag interactions that should be escalated to a higher level of support. By delivering information that is tailored to each taxpayer’s situation and needs, agencies could see higher levels of customer satisfaction, increased compliance, and more efficient issue resolution.  

To ensure that tax agencies begin to unlock the value of generative AI without sacrificing current priorities, we recommend: 

• Starting small. Focus on discrete business processes that will generate tangible ROI. For example, how can we use generative AI to help us better identify remote sellers, non-filers, or fraud? 
• Using AB Testing to ensure that the generative AI solutions you put in place are demonstrably better than what you already have – however if the results are similar don’t discount automation benefits that can free up staff to do higher level analysis. 
• Creating an AI/ML competency center to expand your internal understanding. Like any new technology there is an imminent bubble based on dubious claims. Having a team focused on learning will help your organization sort out what is actually possible with the technology vs. what is just hype.  

-Voyatek Leadership Team