In this summer’s KT Addition, Brian Pereira wrote about “The Pitfalls of Using AI to Prepare Taxes.” This article will expand upon Artificial Intelligence (AI), in specific Generative AI (Gen AI). KT has a limited scope of Gen AI use. This article is one way to openly communicate, educate, and inform our clients about products used by the firm. In this article, I dig deeper into what Gen AI is and isn’t, the rewards it offers, and some of the obstacles that it can present.

What is Gen AI?

Gen AI is a continual learning tool/model that scours information sources to generate text, images, and content based upon data it locates and what it is asked to produce. Gen AI actions are defined by research and constraints, controlled, or entered by the person executing the search. In simple terms – Gen AI pulls data from multiple sources to create the intended outcome as requested by the end user, following the variables and constraints used to define the search.

Gen AI cannot discern fact from fiction on its own, and therefore, uses what information it has access to when creating the final output. Because variables, content, and output may vary from the intended result, human review and revision is used to validate the response.

Gen AI operates in one of two environments: Open or Closed. Open models are publicly accessible, allowing anyone to use, modify, and improve them and capitalizes on a depth of content that is not unique to a topic but inclusive. Closed models are controlled by a specific organization or group and usually contain proprietary and confidential information not accessible to the public without a license or agreement.

Rewards Offered by Gen AI

  • In an Open model, whatever data the Gen AI tool finds is usable and that often includes personal identifiable information.
  • In a Closed model, the data source(s) are usually proprietary and constrained, keeping data secure and relevant to a specific industry.
  • Open models are publicly accessible, anyone may use, modify, and improve upon them.
  • Closed models are usually proprietary and controlled by an organization who is an expert in the content.
  • The Open model fosters collaboration, transparency, and rapid innovation.
  • The Closed model allows for industry specific resources, and training materials or data to be used without external data confusing the context of the search.

Examples of Gen AI

Open Gen AI products: Grammarly, ChatGPT, and Gemini (there are many more).

Closed Gen AI products: TaxGPT and CoCounsel (there are many more, but these are CPA specific).

Obstacles Presented by Gen AI

Lastly, the importance of efficiency and access to information are important to understand. Two studies conducted within the last year validate this perspective:

No matter what society believes about technological change, the rate at which it changes is exponential and the impact on work and outcomes is equally changing. At KT, the firm leverages both Open and Closed Gen AI technologies to be more effective and efficient, benefiting our clients with research that drives quantifiable results without compromising client data security.

KT will never input client information into an Open Gen AI product. Open Gen AI products have limited use at KT. Closed Gen AI products, on the other hand, are gaining a presence at the firm as the information is specific to tax, audit, and accounting. Information searches will never include client information as these products are designed to provide details related to changes in regulation or code.

I hope this helps you understand a little more about Gen AI and how KT may leverage it to benefit our clients and staff.