Regulatory agencies are using artificial intelligence for ‘anticipatory regulation’. This type of regulation hyper focuses on subjects’ characteristics for successful efficient regulations. The regulatory agencies’ subjects are businesses. The affect is business risk areas which are prone to regulatory agencies use of artificial intelligence.
Anticipatory regulation uses artificial intelligence to solve problems and find regulatory infractions in today’s complex world. The extend of this regulatory action is witnessed here by the World Economic Forum stating, "....we are beginning to see a radical change in both the theory and practice of regulation with the emergence of a new field of ‘anticipatory regulation’," 10 Ways Governments Can Keep Up.
Anticipatory regulation determines the probability of compliance based on the subjects’ characteristics. For example, the regulatory agency’s anticipatory regulation identifies two groups for probable noncompliance. The groups identified are 1) rapid growth companies and 2) companies with distribution centers. The identification provides a framework for the regulatory agency to tailor reviews, and audits based on the explicit knowledge of areas of compliance risk.
The massive computing capabilities of artificial intelligence provides the means to identify and endless number of groups, characteristics, and patterns. These patterns are represented in a continuum of probable compliance to probable noncompliance. To shed light on this idea, each point is represented by a percentage of likelihood from probable compliance to probable noncompliance.
This builds a framework for regulatory agencies to tailor reviews, and audits based on the explicit knowledge of areas of compliance risk. This hyper focus by regulatory agencies will affect business and many of them will be unsuspecting recipients. Business risk areas are seen in everyday activities and may result in forced correction by regulatory agencies due to their use of anticipatory regulations.
Business Risk Areas Prone to Regulatory Agencies Use of AI
1. departmental silos
2. lack of adherence to policies and procedures, unintentional and intentional
3. notices received for federal or state late filings resulting in fines
4. surprise additional taxes owed and due
5. extension of returns year after year
The root of these business risk areas is frequently systemic. In this context, the root of the problem is intertwined, interrelated, and cross functional throughout the business. The complex nature of the problem is historically the reason it is not corrected. Frequently, this is due to other priorities, as seen in the business case below.
The business case unpacks an example using A Company, Inc. (ACI). The business case illustrates, how business risk areas, translate to awareness by regulatory agencies using artificial intelligence for anticipatory regulation.
ACI exhibits symptoms of the business risk areas of department silos, unintentional lack of adherence to policies and procedures, and surprise additional taxes owed and due. 1, 2, and 4 above. When systemic problems are present, it is common for problems to be rooted in multiple business risk areas.
Notice of tax due received
ACI recently received a tax bill from their primary state of business for a large underpayment of taxes. The tax due comes as a surprise because no other notices were received. The amount due seems unlikely since ACI uses best practices. In support of that, the records reconcile to the tax filings, and all returns are filed timely.
The decision-making failures of the ACI’s employees depicts the intent to make the right decisions but the wrong decisions are made. The problem is rooted in a non-intentional lack of adherence to the company's policies and procedures; a lack of time; and a breakdown in communication cross-functionally, departmentally, and with the technology, in the ACI business case, a credit card portal (input) to regulatory reporting (output).
The identification of the failures resulted from researching the property tax bill due. The problem traced to the credit card process. Capital assets charged on the company credit card are (unintentionally) expensed. Stated differently, the capital assets are recorded to the income statement as an expense, instead of a fixed asset on the balance sheet.
The inaccurate expensing ultimately affects the regulatory filings through omission. The governance standards are in place, evidenced in the policies and procedures; however, the controls failed to stop the wrong decision from being made.
The company grew 200% year over year for the last 3 years. The rapid growth led to international customers and employees.
The growth is driven by 3 cloud applications, and a popular hardware device that is complimentary to the applications. The rapid expansion led to strategically placed distribution centers for the devices. New distribution centers are projected over the next three years.
All supervisory employees and above carry a company credit card that is used extensively to fund the demand of the growth. Employees’ computers and build out costs for the distribution centers are frequently charged to the company credit cards.
The existing credit card system consists of the receipt’s image captured at the time of the transaction. To complete the transaction the employee (the credit cardholder) codes, finds and attaches the receipts’ images, and writes short descriptions. These steps occur in the financial institution’s online portal.
The completion of all transactions is required once per month. The employees are extremely busy and usually complete the transactions, at the urging of accounting.
Incorrect coding and descriptions frequently occur due to the time lag to complete the transactions, the amount of monthly activity, and an overall lack of time by the employees. Corrections in the accounting department are rare due to the same reasons.
Frequently the failures in the process results in capital assets recorded as expenses to the income statement, instead of to the balance sheet. The inaccurate expensing ultimately affects the regulatory filings through omission.
The records used to file the returns do not include the capital assets, because they were expensed. This means under reporting on the regulatory filings. The omission is the reason the records supported the regulatory filings as timely and accurate.
The omission leads back to ACI’s notice of tax due. Anticipatory regulation using artificial intelligence determined the probability of compliance based on the subjects’ characteristics: industry, growth, distribution centers, global. Within this group, the regulatory agency identified 1) rapid growth and 2) distribution centers as significantly underreporting.
This led to the regulatory agency determining the probable amount under reported and sending the tax due notice to ACI. It is determined internal to ACI, significant underreporting exists in prior regulatory filings.
Unfortunately, ACI must pay the taxes due for the years indicated on the notice.
ACI understands the value of data driven business models. The chief executives have been focusing intently on building the business model utilizing artificial intelligence. This affords real-time information allowing for better business decisions.
It is determined the pilot program for the new business model is the correction of the business problem of inaccurate input (coding, image attachment, etc.) to output (regulatory filings). The business case provide focuses on the input, the credit card.
The flip side of the surprise tax amount due is ACI now has insight into the regulatory agency's expectations of reported values. A basis is provided, in addition to, the tax rate and the amount due.
The ‘expected’ reported values will be used as boundaries or parameters within the artificial intelligence. Artificial intelligence’s parameters are discussed below.
Artificial intelligence will be used to correct the business problem. It will properly code the transactions; find images in storage and attach to the correct transaction; and provide correct descriptions. This information is viewable in real-time metrics by the proper designees.
Before discussing the specific technology used, an understanding of artificial intelligence is necessary.
Artificial intelligence is a set of technology tools made up of algorithms which use data sets to solve problems. Two examples are natural language processing technology and machine learning technology, defined and examples provided in the business use case below. The tools frequently work together using statistical methods to provide real-time information such as diagnostic, predictive and prescriptive analytics, among other things.
A few of the data sets, ACI will use are the vendor name, such as restaurant names; vendor details, such as address; amounts paid, and descriptions. A less obvious data set, as mentioned above, is the known expected reported property values by ACI to the regulatory agency.
Artificial intelligence uses the historical data sets to ‘learn’, by processing algorithms. This is known as cognitive computing which is the ability of a computer to simulate human thought processes. Discussed more in this article, What Everyone Should Know About Cognitive Computing.
Artificial intelligence learns from data sets. If the artificial intelligence uses poor data sets, the artificial intelligence learns incorrectly, and then the outcomes are poor. The data must be cleansed and standardized before using artificial intelligence. In the case of ACI, this means correcting all prior credit card miscoding and poor descriptors.
A pillar in ACI’s data driven business model is risk management. It is determined by the chief executives to minimize its compliance risk by using a set of technology consisting of three types of artificial intelligence which are:
1) natural language processing technology
2) machine learning technology
3) robotic process automation technology (RPA)
Natural language processing, machine learning, and RPA are all types artificial intelligence. There are some professionals who do not include RPA as artificial intelligence because it has no cognitive abilities.
Natural language processing technology processes natural human language. Its abilities are written and speech recognition. Natural language processing technology processes speech into text, text into speech, translates language to language, etc. For ease of explanation, disregarding other embedded technology, a few examples are online chatbots and voice phone assistance.
Returning to the ACI business case, the natural language processing technology will process, or read the text on the credit card receipts. In order to do this, the framework for the algorithm is all prior vendor names, details, descriptions, and amounts paid.
Machine learning technology uses algorithms to perform tasks and relies on learning patterns in the data and using inferences. Machine learning technology can use text and numbers. Examples of machine learning technology are the friendly personal assistants', Siri and Alexa, with the ability to provide help. Learning your preferences and providing recommendations are machine learning in action. Other embedded technology is natural language processing which allows the assistant to speak.
ACI will use machine learning technology to correctly code and provide descriptions for the credit card transactions. The framework for this algorithm is all vendor names, details, and descriptions.
RPA is a software bot that performs high-volume, repeatable tasks. RPA can achieve any repeatable tasks performed on a computer. To name a few functionalities, the bots can file, paste, save, and manipulate. For proper computer controls, RPA has its own login identity (i.e. username and password) to perform functions such as logging in to private clouds, virtual private networks, and financial institutions' portals.
ACI will use RPA to access the financial institution's portal. In combination, the natural language processing and machine learning will code; provide descriptions; find and attach images; read handwritten text descriptions from the images and read the amounts on the credit card transactions. This will occur in real-time, as soon as the transaction is posted by the financial institution. Providing a means for better decision making for the company.
Natural language processing, machine learning, and RPA will be used for improved business decisions. ACI will use the real-time information for descriptive, predictive, and prescriptive data analytics, among other things. The historical information, up to and including real-time information, will allow better business decisions by for the immediate and the future.
As mentioned earlier, technology provides for parameters to be set so that the technology functions in the manner intended. This holds true for artificial intelligence, also. The expected value, by the regulatory agency, can be monitored in detail by ACI. The artificial intelligence technology set in use will have parameters established to identify anomalies and patterns in the data.
For example, ACI sets a parameter that provides notification to employees when current activity is inconsistent with historical activity. This information is analyzed in 1) total of all credit card holders, 2) individual credit card holders, and 3) more complex patterns such as bundling of transactions. In the instance of opening a distribution center, a parameter is set to flag missing service providers.
Another example of a parameter is risk management controls for accuracy. A parameter is set that notifies an employee to review. For example, ACI sets its parameters, such that, if there is more than a 20% chance the vendor recognition is wrong, then the artificial intelligence will notify the employee for additional review. Stated from the flip side, when there is an 80% or more chance the vendor recognition is correct, then the transaction is placed into a data set with the vendor identified and the transaction continues through the channels of coding; image identification and attaching; and description writing.
The employee who is the credit card holder is still accountable for the accuracy of the work performed by the artificial intelligence in the context of the credit card transactions such as the coding, vendor recognition, and descriptions.
Thankfully the business problem in ACI is resolved, risk is mitigated. and future returns will be filed correctly. Employees are freed from the timely tasks of finishing the credit card transactions to using real-time metrics.
Unfortunately, ACI must pay the taxes due for the years due on the tax bill because under reporting was found.
Business risk areas, seen in everyday activity, are prone to regulatory agencies use of artificial intelligence. This new type of regulation gives chief executives a good reason to evaluate and correct business problems. Becoming a surprise recipient of anticipatory regulation is avoidable.
Artificial intelligence is a means to understand today’s complex world. Using artificial intelligence is a new business model. Non-strategized digital platforms are witnessed in high costs per employee due to SaaS subscriptions with overlapping functionalities, disparate systems, and a lack of real-time information.
A business model with artificial intelligence provides for real-time information improving business decisions and customers’ experiences. TICH builds data driven business models using artificial intelligence from input to output.
This new business model moves companies beyond a competitive advantage to a business advantage.