Agentic AI impact on CPAs

Agentic AI Impacting Accountants' Work

The next generation of AI is coming less than three years after ChatGPT. It promises to change how accountants work. Agentic AI has agency and can act and make decisions independently. It can define goals, create plans, and adapt with minimal human assistance.

Randy Johnston, CEO and co-founder of Network Management Group Inc. (NMGI), said that while generative AI tools need step-by-step prompts from users, agentic AI uses “sophisticated reasoning and iterative planning to autonomously solve complex, multistep problems.”

Donny Shimamoto, CPA/CITP, CGMA, founder and managing director of IntrapriseTechKnowlogies LLC, said agentic AI is expected to be widely used by the end of the year and incorporated into many AI tools. According to Shimamoto, it will free accountants to focus more on analysis and strategy, which will make their jobs easier and allow them to use their training. He compared it to having quick access to a team of Ph.D.-level researchers who can think through complex projects rapidly.

Johnston, who will present the Technology Update session with Brian Tankersley, CPA/CITP, CGMA, of K2 Enterprises at the AICPA & CIMA ENGAGE 25 conference on June 10, anticipates agentic AI agents will be used for customer service, sales support, supply chain reconfiguration, and IT project management. After looking at several accounting-related agents, he predicts that agentic AI will be able to do certain tasks in the next one to three years.

Both generative and agentic AI use algorithms to predict the next word and simulate human learning and decision-making. They are trained using large datasets and can respond to natural language requests to create content like text and images. Agentic AI is programmed to learn and adapt faster than generative AI and make context-based decisions.

According to Johnston, generative AI responds to prompts such as summarizing meetings and converting reports into presentations. Agentic AI can independently handle multistep queries to find common answers and expand on them. For example, agentic AI could research tax law changes and create a client presentation or produce a year's worth of marketing materials by analyzing strategic documents and prior marketing efforts and then creating content and scheduling its distribution on various channels. It can also manage customer interactions on social media and websites.

Agentic AI solves problems in four steps:

Agentic AI Four-Step Process

Perceive: Input information from text, audio, cameras, and sensors to receive project requests from human users and gather data.

Reason: Use LLM technology to process data and set project goals.

Act: Independently complete tasks to fulfill user requests.

Learn: Improve through machine learning and feedback.

An IBM report comparing Agentic AI vs. Generative AI states that agentic AI can learn and operate on its own, making it a promising technology for streamlining workflows.

Marc Staut, chief innovation and technology officer with Boomer Consulting, said that, like generative AI, agentic AI agents improve with use, and training makes them more effective.

Implementation of Agentic AI

Agentic AI’s entry into our everyday life won’t necessarilycome with signs popping up to announce, “You’re now using agentic AI!” Instead,expect incorporation resembling how search engines such as Google adopted AI platforms. A user wasn’t necessarily told what powered their search for the best tacos in the area, they just received a curated list within moments.

Shimamoto said that agentic AI will likely be phased into existing software for those in accounting and finance management, and users may notice more user-friendly interfaces that can produce more sophisticated results. Byron Patrick, CPA/CITP, the CEO of VERIFYiQ, said that agentic AI is being phased in as access increases to DIY automation tools. According to Patrick, combining agentic AI and DIY automation allows small firms and finance teams to improve processes by easily building automations and incorporating agentic AI in their operations.

Staut suggests accountants and finance professionals become familiar with agentic AI to determine which existing processes could benefit from it. He also said that it’s important to remain strategic about software and vendors to ensure it’s a smart move and not just buying everything with agentic AI. Staut stated that figuring out the best agents to invest in may be a challenge.

Agentic AI vs. Robotic Process Automation

Patrick said that agentic AI can accomplish what robotic process automation (RPA) was supposed to do but often failed to do because of inflexibility and complexity. He said to think of RPA and generative AI as individual musicians and agentic AI as an orchestra that can have multiple agents on their own decision-making missions and then combine all the pieces. According to Patrick, continuous deployment can produce exponential improvements.

Agentic AI could help with the staffing crisis in the accounting profession by taking over tasks that add hours to the workday. Patrick said that these tools will increase productivity, value, and output and possibly give people time to vacation with their families while the bots work.

It is important to remember that AI, whether generative or agentic, can’t do the creative thinking that humans do. AI follows processes and looks for the most common answers, which aren’t always correct. Human oversight is still needed, and more strategic decisions and advice should not be handed off to the bots. According to Shimamoto, human creativity looks outside the box, while AI generally looks inside the box.

Challenges and Considerations

Patrick said that behavioral shifts and a willingness to see what the technology can do are needed to harness agentic AI’s potential. He added that another hurdle will be understanding and accepting the results. While agentic AI is less prone to hallucinations than generative AI, they are still possible. Also, accountants likely won’t know the processes agentic AI uses, which can cause doubt. Patrick expects it will take time for accountants to better understand how AI can help and where caution is needed.

It’s important to determine where a human needs to be involved, including critical decision points and processes that might be difficult to undo or cause issues if done wrong.

Shimamoto said that firms and companies must conduct due diligence when selecting AI product vendors and ensure data is stored securely and privacy regulations are followed. Organizations should also address how functional and accurate the product is and whether it is fit for its intended purpose. Patrick suggests using existing processes to vet vendors and asking for SOC 1 or SOC 2 reports.

Other assurances to look for are proof of compliance with the California Consumer Privacy Act (CCPA) and the European Union’s General Data Privacy Regulation (GDPR). According to Patrick, agentic AI is not a passing phase, and CPAs should plan to use it in the next few years and get curious.

Sarah Ovaska is a freelance writer based in North Carolina. To comment on this article or to suggest an idea for another article, contact Jeff Drew.