Artificial Intelligence and Corporate Reporting

Financial Reporting Council report considers the impact

The Financial Reporting Council (FRC) is the regulator of auditors in the United Kingdom, as well as the standard setter for corporate governance. The FRC’s goal is to promote transparency and integrity in business for the investors and others who rely on company reporting.

In response to the call for improved corporate reporting, the FRC established the Financial Reporting Lab (Lab) in 2011, as a means for investors and companies to together develop pragmatic solutions. To address technology related challenges, the Lab initiated two projects. Digital Present looked at the current state of technology tools and how to best employ them. Digital Future, building on the results of Digital Present, “is designed to identify what benefits the new mediums and technologies should offer, consider which technologies might do this and how companies can make the most of the digital opportunity.”

One of the findings of the Digital Present project report released in 2015 was that companies were not making the most of the available technology in the corporate reporting process. In what may sound a bit quaint today, the report also found that the use of the PDF format was the desirable digital medium for corporate reporting, and that expanded use could take advantage of “a series of beneficial features and attributes.”

When the Lab launched Digital Future in 2017, two approaches were considered: 1) look at what available technology offered, or 2) look at the desired outcomes and determine how technology could help attain those outcomes. The second approach was adopted. From there a framework was laid out for use in future reports that focused on the three stages of corporate reporting: production, distribution and consumption.

Thus far, three reports have been issued: XBRL deep-dive – Digital future of corporate reporting (December 2017), Blockchain and the future of corporate reporting – How does it measure up? (June 2018), and Artificial Intelligence and corporate reporting – How does it measure up? (January 2019). Still to come are reports on Artificial Reality, Virtual Reality and Video.

The newly issued report on artificial intelligence (AI) first explains in basic terms what AI is. While stating that a standard definition is “a branch of computer science that deals with the creation and research of human-like intelligence in machines,” the report then focuses on the currently accepted idea of AI as the “leading edge in computerization and automation,” which also includes areas such as robotics, machine language, and natural language processing. Two basic principles are that AI is 1) highly efficient for processing highly routine items where a clear cut set of rules can be employed, and 2) and highly effective for working with information that is too voluminous for objective and practical human interaction.

While sometime in the distant future, AI may be able to take totally raw data and create meaningful reports, the current status of corporate reporting is such that data needs to be structured to some extent for AI to work well and objectively.

Even so, for AI to replace human judgment in a significant way, there also needs to be trust among stakeholders, which will require those stakeholders developing a much deeper understanding of AI, regardless of how advanced AI processes become. Financial professionals who have been in their careers for several decades may find this hurdle insurmountable for them. Those who now often turn to their grandchildren to operate modern computerized devices may be content allowing this next generation to carry forward the AI mantle.

Trust and transparency are expectations in corporate financial reporting. The report listed three challenges that were dealt with in considering AI:

  1. The efficiency of recording and aggregating transactions, across multiple entities, and then turning that data into an external communication
  2. The efficiency and effectiveness of providing internal or external assurance over the resulting communication
  3. the effectiveness of consuming the information reported by, and about, the company and translating information into insight and ultimately into action.

While discussing the utmost importance of having quality data available for meaningful AI applications, the report offered these questions as ones needing attention:

  1. What data was used to train the system, where did it come from, how was quality and lack of bias ensured?
  2. Are there opportunities in accounting changes (such as IFRS 15 and 16) to gather and optimize information for AI?
  3. If the system is external to the company, what is happening to the company’s data and the learning derived from that data? Is it portable?
  4. What controls and processes need to be changed or modified for AI?

After dealing with these challenges and data considerations, the report concluded that AI clearly has the potential to be important for corporate reporting, and that it is just a question of how long it takes.

I asked an Integra International member who is an active observer of AI to comment on this article. His observations confirm the findings of the report. With regards to a sizable current project recently completed, he noted that “1) the abstracting process requires teaching the software what to look for and letting it get better over time, 2) the human review of a complex agreement…like a complex 100 page lease …is still critical, and 3) the need to have documents in order and complete is very important…such that many companies have not maintained their lease records in the best manner to facilitate machine leaning.”

Looking ahead, the AI report takes some “moon-shots” to theorize what the future could hold. For example, say by 2034, the complete annual reporting cycle could be replaced with an any-time, any-period “perennial report” completely produced by AI. To conceptualize this prospect, consider today’s sports reports. Some of these are already produced by AI. The statistics and graphs produced online for games or matches are readily recast into narrative reports. The data is structured and similar in format, so analyzing team and individual statistics and commenting on which teams and individuals stood out and why is not hard for a computer to learn and describe in a report.

Stepping out another decade beyond 2034, potentially using tools like machine learning (ML) and natural language processing (NLP), a robot advisor could learn from available underlying data and provide accounting interpretations on demand, or eventually even employ natural language generation (NLG) to produce drafts of standards themselves, after analyzing massive amounts of data, history and actual practices, for areas where standards are lacking or incomplete.

Another “moonshot” goes so far as to theorize a corporate board of directors that is made up completely of AI robots. Apparently, according to the report, some companies already have an AI board member. After all, since corporations, themselves, are already entities that are not human (except possibly in the case of US election campaign law), an AI board may not be that far-fetched. To which the report asks the question: “Would investors own the entity or would it own itself?” Stay tuned!

Further details can be found at the Artificial Intelligence – How does it measure up?.