Generative AI – what is it and where is it taking us?

Themes at the 2023 AICPA & CIMA Conference covered use cases, governance and the future of the accounting profession.

The last two panels of the Conference did not disappoint. The investor considerations panel covered the pace of AI adoption and the expected investor-driven impact on reporting and disclosures. The technology panel – moderated by KPMG Partner Brian Fields – continued the same theme, discussing the potential impact of GenAI on financial reporting and the related governance needs. These panels, and the Conference, ended with a look forward by discussing the impact of GenAI on the accounting profession.

“Generative AI is different in important ways from technologies previously used in finance and accounting because of its probabilistic nature. It promises amazing improvements in financial reporting speed, quality and insights. But it comes with new demands on corporate governance, internal control and auditing techniques to ensure it is used responsibly.”

— Brian Fields, KPMG Partner

In a recent KPMG survey, three out of four business leaders indicated they will give GenAI priority over other emerging tech during the coming year.

What is AI and GenAI?

AI covers a broad range of computer programs intended to mimic human behavior. It is an umbrella term that encompasses a range of interrelated technologies – from simple rule-based logic (deterministic) to more advanced and complex algorithms (probabilistic). GenAI is a specific subset of AI based on probabilistic technology that synthesizes large amounts of data using an artificial neural network. In essence, it represents an artificial brain.

GenAI does not require the heavy customization that the deterministic models (like robotic process automation, or RPA) require. Although they can also analyze large amounts of data, deterministic models require specific instructions and can break down when even minor things change in their environment.

In contrast, finance professionals will be able to use GenAI to execute complex business processes across applications. Further, GenAI will improve in accuracy and intelligence over time.

With GenAI technology to enable more streamlined communication between humans and heir technology, a new stage of automation may be on the horizon.

What impact will GenAI have on financial reporting?

There was consensus among the panelists that the stakeholders in the financial reporting ecosystem are well-positioned to scale their use of GenAI. Breanne Dougherty, Head of Thematic Research, Bloomberg LP, highlighted that investors will have the ability to analyze disclosures and translate that into research to inform investment decisions.  

Key benefits of this technology include anomaly detection, efficiency and accuracy in processing and analyzing financial data. This can result in:

•      increased productivity through cognitive automation;

•      increased accessibility of data for a range of uses;

•      increased ability to identify data outliers;

•      increased visibility into end-to-end processes and controls;

•      real-time insights into areas of risk or control weaknesses; and

•      faster quarterly and year-end reporting.

Panelists noted a number of examples of GenAI in use today.

Managing supplier and vendor activity

Rohit Gupta, founder and CEO of Auditoria.AI, explained that GenAI can aid in revenue accounting, including billing and gaining insights about customer behavior, invoice digitization within the procure-to-pay processes, and supplier and vendor management.

Reviewing documents

Joshua Waldron, VP Finance & Accounting, Scale AI, explained that GenAI can summarize the important data in documents. This can be particularly effective with long documents (e.g. 200-page leases) and those in a foreign language. This was echoed by Jason Cuomo, Senior Vice President, Moody’s, who was impressed by the ability to quickly summarize long documents, including earnings transcripts. 

Waldron also mentioned that GenAI can map the disclosures in draft financial statements with items in a disclosure checklist. 

Reviewing completeness and accuracy of financial statements

Adaeze Egwuatu, Senior Director Data & AI at Microsoft, explained how companies can use GenAI to analyze financial statements in natural language instead of using time-consuming keyword searches. One company reported a 10% time saving after using GenAI in this manner for only a few weeks.

What are the governance challenges with GenAI?

The panels discussed good governance practices and ICFR considerations. At the heart of good governance is having a clear understanding of what GenAI can and cannot do, establishing guidelines for its use and ensuring compliance with regulations and corporate policies.

To that end, panelists were focused on the importance of quality control and the reliability of information generated by AI tools. Analogizing to driving, Patricia Wilson, Global Head of Research, CFRA, stated that AI may get you there faster, as long as the driver of the car (i.e. the developers) actually knows where you want to go.

Panelists mentioned a few specific areas that present governance and internal control challenges, including adapting and changing related general IT controls. 

Data security

A company should ensure that third parties it uses have proper controls in place so the data it provides to those parties is secure. It’s important to consider how it is working, who is actually hosting that computing, and understanding who has access to that information and data.


Large language models (LLMs) can hallucinate, which occurs when an LLM returns an answer that is misleading or simply false. Certain safeguards may help – e.g. content filters to monitor topics that may be offensive or pose a security risk – but human-in-the-loop controls are still needed in most current GenAI use cases. 

Legal considerations

A company will have to determine what GenAI output to share and how to share it. It also will have to determine who has the ultimate rights to the output and how its partners are using that output and any underlying data.

Ultimately, companies have a responsibility to make sure AI is deployed in a safe, ethical and fair manner. As AI regulations continue to be developed, companies should be mindful of how existing and developing regulations may apply to their application of AI.

Fields mentioned the KPMG Trusted AI Framework, which is based on 10 ethical pillars for managing AI systems throughout the AI lifecycle.

What is the potential future of the accounting profession in a robust AI and GenAI environment?

While some have expressed fear that AI will replace workers in the finance and accounting professions, the panelists tended to have a more optimistic view, believing that AI is likely to ‘raise the floor’ – allowing accountants to focus more on the work of synthesizing results versus ‘number crunching’.

It has the potential to increase productivity and change the way people work, innovate and collaborate. For example, Gupta predicts that every finance professional will have their own AI-bot assistant to execute repetitive tasks, understand and discover anomalies and prepare them for for the work to be done. Gupta believes AI assistance could be leveraged in 35-40% of transactional work. Egwuatu agreed, stating that GenAI tools will be prevalent in the finance and accounting funcations.

As the ability to work seamlessly with advanced technology becomes critical, panelists believe we will be able to attract more talent to the accounting profession. However, panelists noted that when technology like GenAI is used by those without the right skills and training, it is much less powerful and potentially even dangerous. For accountants who embrace the future, this technology could profoundly change their ability to deliver impact.

Learn more

To learn more, check out KPMG resources on responsible AI and the KPMG Generative AI resource page, which includes featured AI insights, AI events, and AI webcasts and replays. This includes our recent AI in Audit Survey and newly launched Trusted AI Framework.