Contemporary Case Studies

To illustrate the current implementations of Al in the accounting profession, we conducted in-depth interviews with partners, principals, and Al technology leaders from the large professional services firms, including KPMG LLP, the US member firm of the KPMG network of independent member firms affiliated with KPMG International, and its global network of affiliated member firms, Ernst & Young LLP, Deloitte & Touche LLP, and Grant Thornton LLP. The interviews were semi-structured and conducted via phone conference during the first half of 2020. The length of the interviews ranged from thirty minutes to one hour, followed by correspondence via email for clarification and validation. The interviewees mentioned later in this chapter shared their perspectives, experiences, and lessons learned from Al implementations for various accounting applications.

Following these interviews, we develop and present five case studies for more in-depth analysis. The case studies include using natural language processing for risk analysis, utilizing artificial intelligence for transfer pricing services, deploying autonomous audit drones for inventory management, applying Al models to augment auditor judgment, and implementing data transformation tools and RPA to assist with monthly tax reporting needs. These case studies provide useful insights and practical guidance on how firms can leverage the power of Al to solve business problems.

Case Study #1: Use of NLP for Risk Analysis (KPMG)

Background

KPMG LLP began piloting Al in their Audit, Tax, and Advisory practices as early as 2015, according to their Global Head of Innovation, Steve Hill*. Hill notes, “We started an effort in collaboration with one of our strategic alliances to effectively reframe and reinvent the accounting industry using Al, cloud, and data as the three levers of transformation.” Hill describes the initial goal was to use Al to supplement professional judgment to “augment the knowledge and insights of our KPMG practitioners and enhance decision-making.”

Vinodh Swaminathan, Global Lead Partner and Principal with KPMG LLP’s Advisory Management Consulting practice focusing on Digital Transformation, stated that KPMG successfully leveraged natural language processing (NLP) in providing Audit and Advisory risk assessment services for clients engaged in mortgage-backed securities, commercial mortgages, and other loans. Swaminathan described the following scenario: “imagine, a typical loan file for a mall in Philadelphia is probably a few gigabytes, maybe 50 different documents ranging from promissory notes to rent rolls and rental contracts with the occupants. The file has environmental surveys and engineering surveys. Think of a portfolio of documents that comprise a single loan file in which a client may be holding the loan in the portfolio to issue some sort of financial assistance to the third party.” Before Al, Swaminathan explained that a KPMG professional performing this type of risk analysis would have to “literally sift through each of these 50 documents that constitute a single loan.”

Today, KPMG believes that Al will eventually permeate almost every aspect of a business. On KPMG.com/us, KPMG LLP states that “Al, Automation, and Analytics are central to the success of the enterprise and will pervade critical business areas, including data, business processes, the workforce, and risk and reputation” (KPMG, n.d.).

Results

The first benefit of using NLP for this use case is that it resulted in significant efficiency gains. Swaminathan explained that “what normally took our loan professionals two weeks now can take only two hours.”

A second benefit is improved quality, as KPMG loan professionals could now spend more time focusing on other areas that require judgment skills. Swaminathan remarked, “given that we reduced the two weeks down to two hours using the same staff that we have, we can start to look for risk across broader parts of the portfolio.”

Finally, a third benefit of using NLP is the codification of institutional knowledge. For example, before the NLP solution was deployed, if an experienced senior manager would leave, so would their knowledge. However, with NLP, Swaminathan explained, “I have codified

Contemporary Case Studies 73 that institutional knowledge through these machines. 1 am able to, over very long periods of time, do new kinds of analytics. I have hard data now collected and codified because machines are involved over long periods of time. Again, we’re able to go in and offer deep insights to clients, possibly even sometimes changing the nature of the value proposition our member firms are offering to clients.”

Lessons Learned

Providing Al services involves significant investments in technology infrastructure and human capital. The focus of firms providing Al-related services should center on long-term productivity gains instead of short-term cost savings.

For KPMG firm’s clients, the impact on the fee structure depends on the type of engagement and scope of services. According to Hill, “[the fee structure] varies based on the value proposition. Some of the things we’re doing, some of these rates are packaged in fixed fees. We’re giving our clients value back in a number of ways, either additional value in engagement or quite frankly, we get better, faster; the hours go down. It really kind of varies across the footprint.”

For KPMG, Hill explained “the reality is, we’re trying to extract some of the investment out of these innovations because they’re difficult to build, and our intention is to evolve the economics of the business so that where we can provide efficiencies ourselves, the clients get those efficiencies quickly. We have varying kinds of pricing to help us do so.” Further, Hill notes that “you can see the entire industry, especially on the advisory side, moving more toward value-based and performance-based pricing rather than just hours and rates.”

Swaminathan adds that “if we view that Al projects should be done purely to recover the investment based on cost savings, a good seven out of 10 times they’re probably going to fail.”

Hill provides the following advice for organizations seeking to invest in Al. “The other thing a lot of people don't understand is what we call organizational capital required to make these things successful over time. You have to build this very strategically. The intelligence and the value of these appliances become greater over time, and they can be leveraged in different ways. The fungibility of this technology, when applied in patterns that can be reused - some of them proprietary to your brand, proprietary to your analytics, proprietary to your data -can be extremely powerful”.

“But if you think you’re going to develop a compelling investment case for the business with a return narrowed to just a single project,

you are likely to be disappointed. RO1 needs be appropriately and responsibly associated with the institutional knowledge, asset or component reusability, and infrastructural value that you create in the organization, again what we call organization capital, is a huge throw off of value associated with this work,” said Hill.