Artificial Intelligence (AI) technologies are being adopted in pockets across the financial services industry for a wide variety of purposes. AI applications are now widely used in areas like fraud detection and more recently for portfolio trading. Another up and coming area for AI is wealth management where there’s been an explosion in powerful new AI investing tools that are favored by millennials – with the potential to displace the current generation of wealth management professionals.
These use cases are based on analysis of transactional data, with AI being used to analyze thousands or millions of data points to uncover hidden patterns and anomalies. But AI offers powerful capabilities that are now being directed at historic data, an essential element in making strategic decisions. A good example is how AI is being applied to analyze company financial reports and extract insights that impact corporate valuations, growth prospects and market perception.
Financial footnotes are an important source for company insights that accompany each financial filing for public companies. Policies and disclosures included in footnotes provide guidance on accounting treatment for material items that reflect operating results and impact a company’s market value.
However, looking for insights in footnotes is a challenge due to the ever increasing complexity and length of footnote sections. Ernst & Young projects that companies will devote more than 500 pages in their annual reports to footnotes and Management’s Discussion and Analysis by 2032 if the current pace continues. It’s unclear how this level of disclosure will be truly helpful to analysts.
Another complication is the increased use of non-GAAP accounting that requires analysts to spend inordinate amounts of time sifting through footnotes for policies and disclosures. Financial statements typically need to be adjusted or “normalized” to account for items that management believes are not reflective of normal operations. These items include exceptions and non-recurring items like the one-time sale of a major asset or the financial impact of a legal settlement. Normalizing statements requires an intimate knowledge of how the policies and disclosures contained in the footnotes impact financial statements.
The current method requires analysts to pore through screens of data from corporate filings and place numbers into spreadsheets – a process that takes hours for a single statement, even for the most experienced analysts.
Innovative AI methods are being applied to streamline this process which is also aided by SEC reporting guidelines that now require results to be provided in XBRL format. eXtensible Business Reporting Language (XBRL) is a mark-up language for financial reporting that’s similar to HTML. Financial data coded with XBRL uses identifying tags. This allows data (e.g. total revenue for Company X in 2Q 2016), to be searched, extracted and formatted easily and accurately.
Because there’s no standard financial reporting format, text analytics and natural language processing are needed to parse and interpret filing information to make results consumable for analysts. As a first step, footnotes are matched to financial statements. This simple innovation saves enormous time by avoiding lengthy “control-f” searches to track down policies and disclosures associated with each of the major accounting statements.
Next, any items likely to be needed in adjusting financial statements are identified automatically from the footnote policy and reported numbers. This capability enables analysts to quickly apply adjustments to financial statements and gain a more realistic understanding of the company’s results, including key valuation metrics like EBITDA. This insight provides the foundation for fundamental equity analysis and corporate valuations.
Footnote analysis is company specific. Analysts also need the capability to ask higher order questions at the industry level to understand trends and averages. Companies with adjusted or “normalized” financials outside a defined range may be red flags that require deeper analysis. AI based approaches enable complex industry analyses that’s been too time consuming up till now.
For instance, non-GAAP items and policies can be catalogued to create an extensive library of terms used by companies to explain common items in very different ways. And every industry has its own parlance. For instance, technology companies in dynamic markets often incur restructuring charges as divisions are bought and sold. IBM refers to employee related restructuring costs, i.e. severance for layoffs as “workforce rebalancing” charges. The value of having a library of these related terms enables analysts to make apples-to-apples comparisons across the industry, removing barriers created by company unique terms.
What’s next for AI? A future sweet spot for AI in financial reporting is likely to be in the area of the Management’s Discussion and Analysis section of company filings. This long narrative of company operations contains nuggets of insight that are still unmined. Subtle changes in wording over time may be indicators of potential threats or internal challenges perceived by management but difficult to capture without sophisticated natural language processing capabilities. AI has the potential to finally bring real transparency to financial reporting and analysis.