Businesses that implement artificial intelligence in accounting without understanding the associated

challenges face significant risks.

Key artificial intelligence (AI) technologies, such as computer vision, natural language processing (NLP),
speech recognition and machine learning, help organizations to improve efficiency and derive insights
for more competitive customer and talent strategies. They can also allow the accounting function to play
a more strategic role.

According to the 2018 EY Global Financial Accounting and Advisory Services (FAAS) corporate reporting
survey, close to three-quarters (72%) of finance leaders around the world believed that AI would have a
significant impact on the way finance drives data-driven insight. However, businesses that dive into
implementation of AI technologies without understanding the associated challenges face significant
risks.

Chapter 1

Impact of AI on accounting

Accountants play an important role in many aspects of the business — from recording transactions to
storing, sorting, summarizing and presenting financial information in a form and manner required for
regulatory compliance and decision-making.

Besides managing a huge volume of diverse financial and nonfinancial data, accountants also need to
ensure that financial reporting is timely, accurate and consistent, while meeting the different needs of
internal and external stakeholders. Given the pace of regulatory change, this is no easy task, especially
for companies with complex operations spanning different jurisdictions.

Organizations have been combining AI with other technologies, such as robotic process automation, to
automate many mundane tasks performed by accountants today. This allows accountants to redirect
the time that they used to spend on such tasks toward performing high-value, high-impact tasks.
Combining AI with other technologies, such as robotic process automation, can allow accountants to
redirect the time that they used to spend on mundane tasks toward performing high-value, high-impact
tasks.

Adding AI to accounting operations can also increase output quality by minimizing human errors. AI can
mimic human interactions in many cases, such as understanding inferred meaning in communication
and using historical data to adapt to an activity. It can often provide the real-time status of financial
matters by processing documents through NLP and computer vision faster, making daily reporting
possible and inexpensive. This promotes better and more timely insights required for swift decision-
making.

Auditors also use machine learning tools in audits that can “read” documents, such as sales and lease
contracts, perform trend analysis and identify outliers. This reduces administrative time spent on
reviewing audit documents and allows auditors to spend more time on areas that involve significant
estimates and judgment.

EY Digital Audit is a connected data-driven audit that takes advantage of companies’ digitization journey
by tapping into the sheer volume of data generated by new technologies to effectively deliver high-
quality audits, allowing them to put greater emphasis on risk identification and deliver better business
insights.

As part of digitalizing the audit process, EY launched an AI proof of concept using computer vision to
enable airborne drones to monitor inventory during the auditing process. For example, the drone can
count the number of vehicles in a production plant under audit and communicate the data directly into
the EY global audit digital platform.

While AI-enabled systems can support compliance and their related audits by monitoring documents
against rules and flagging issues, there are hidden dangers. There is currently little visibility into how AI
and machine learning technologies come to their conclusions in problem-solving, leaving practitioners
exposed to various significant business risks.

Chapter 2
Barriers to implementation

Although AI has clear benefits, the implementation of such technologies in finance functions can be
challenging. Resistance to change from teams within the organization is a key risk as many may choose
to adopt a “wait-and-see” attitude.

In the 2019 EY Global FAAS corporate reporting survey, 60% of Singapore respondents said the quality of
finance data produced by AI cannot be trusted as much as data from usual finance systems. The top risks
cited in relation to turning nonfinancial data into reporting information are maintaining data privacy,
data security and the lack of robust data management systems.

AI relies on access to vast volumes of data to be effective. Significant efforts are therefore needed to
extract, transform and house the data appropriately and securely. The advantage of AI systems is their
ability to analyze and independently learn from diverse data and generate valuable insights. However,
this can be a double-edged sword where a lack of proper data management or cybersecurity systems
can predispose organizations to significant risks of inaccurate insights, data breach and cyberattacks.

Further, smaller organizations may face the issue of insufficient data to build models surrounding
specific areas for analysis. Obtaining such data will also require systems and processes to be established
and integrated to ensure that external data harnessed will complement existing data. This requires
significant financial and time investments. Hence, most companies that implement AI applications in
their accounting systems will likely focus on areas that will have the most significant financial and
business impacts. This can be challenging as more sophisticated AI technologies are still in the infancy
stage and the first implementations will therefore be unlikely to reap immediate benefits.

Even with the right data, there could still be a risk of machine learning algorithm bias. If the patterns
reflect existing bias, the algorithms are likely to amplify that bias and may produce outcomes that
reinforce existing patterns of discrimination.

Another major concern is the potential overexposure to cyber-related risk. Hackers who want to steal
personal data or confidential information about a company are increasingly likely to target AI systems,
given that these are not as mature or secure as other existing systems.
While the legislation governing AI is still considered to be in its infancy, that is set to change. Systems
that analyze large volumes of consumer data may not comply with existing and imminent data privacy
regulations and therefore, pose risks to organizations.

As with any transformation initiative, the human factor is critical to ensuring its success. The evolution in
AI technologies is changing the roles and responsibilities of accountants, requiring competencies beyond
traditional technical accounting that also include knowledge of business and accounting processes,
including the systems supporting them. These competencies are important to effectively identify and
apply use cases for AI technologies, and facilitate effective collaboration with other stakeholders,
including IT, legal, tax and operations, during implementation.

Despite these challenges, the benefits of AI technologies remain compelling. The competitive economic
environment and rapid technological advances will drive adoption. Over time, slow adopters will be
disrupted and risk becoming obsolete. With the potential of AI technologies to be a game changer for
accounting and finance, adoption is inevitable and a sound AI strategy is paramount to successful
adoption.

Chapter 3

Understanding and managing risks

While harnessing disruptive technologies brings great opportunities, managing new risks that come with
them is just as important. Although the risks depend on each finance function and individual application,
organizations should begin by assessing their situation against a spectrum of possible risks:

Data quality and management

This is key to transforming volumes of data into an organization’s strategic assets. Organizations should
commit to building trust proactively into every facet of the AI system from the start. Such trust should
extend to the strategic purpose of the system, the integrity of data collection and management, the
governance of model training, and the rigor of techniques used to monitor system and algorithmic
performance.
Cyber and data privacy

Considerations should be made when designing and embedding AI technologies into systems.
Developing proper system separation and understanding how the system handles the large amounts of
sensitive data and makes critical decisions about individuals in a range of areas, including credit,
education, employment and health care are critical to managing this risk.

Legal risks and liability

At the most fundamental level, organizations need a thorough understanding of AI reasoning and
decisions. There should also be mechanisms to allow a clear audit trail of AI decisions and extensive
testing of the systems before deployment. Risk mitigation should also include assessing the acceptable
costs of error. Where the costs of error are high, a human supervisor might still be needed to validate
the output to manage this risk. As the technology matures further, the acceptable risk level can be
adjusted accordingly.

Prioritizing use cases and culture transformation

Developing a successful AI implementation road map requires identification and prioritization of use
cases, with the understanding that the human element is a fundamental piece of the equation. This is
because the uniquely human soft skills, such as creativity and leadership, as well as human skepticism
and judgment, are needed to address the new risks that come with the adoption of emerging
technologies.

To overcome resistance to change and drive sustainable culture transformation, organizations should
inject new ideas and fresh impetus into the team. One way is to identify “change ambassadors” who are
empowered by management to embark on new technology initiatives and successful proofs of concept
that would then be commissioned for roll-out to the organization. Such similar efforts will be critical to
overcome inertia and resistance.

Changing the finance and accounting talent mix may provide an important lever for culture change. By
changing recruitment criteria to favor openness and innovation, finance leaders can seek to attract
people from different sectors and backgrounds who come with new perspectives and without the
ingrained assumptions and biases of typical accounting talent. Upskilling the existing accounting
workforce beyond traditional finance and accounting skills and redefining the profile for talent
acquisition are key considerations in driving an effective digital-enabled workforce.
The benefits of adopting AI technologies are evident. While it is impossible to predict how AI
technologies will ultimately affect the accounting industry and profession, one thing is clear: companies
and accounting professionals need to invest time — sooner rather than later — to understand AI
technologies and ecosystems, embark on proofs of concept to validate use cases, and drive behavioral
changes that effectively build a truly digital workforce and organization for competitive growth

Summary

Adding artificial intelligence to accounting operations can improve output quality and decision-making.
It can also allow accountants to focus on performing high-value, high-impact tasks. However, businesses
need to effectively manage the risks of implementation and drive behavioral changes that build a truly
digital workforce to truly benefit from such technologies.