By Hitendra R. Patil and Eli Fathi
Artificial intelligence is set to transform many professions. As AI shifts from news and hype, it will significantly change how people work in their jobs.
MORE: How to Be Ready for Real Artificial Intelligence | How Accountants Can Ride the Artificial Intelligence Wave | Automation Makes It Easier for Accountants to Offer Cybersecurity | The Why, What and How of Cybersecurity for Accountants | Blockchain: What You Need to Know Now | The What, Why and How of Artificial Intelligence for Accountants | The Client of the Future | The Rise of the Robot Accountant: Opportunity or Threat?
Exclusively for PRO Members. Log in here or upgrade to PRO today.
The internet and cloud technologies transformed the workplace. Similarly, AI will refine, reshape and even replace many of the tasks we do every day.
Till now, technology transformations mostly made tasks faster or easier. AI is going to be different. Let us understand how AI can be effectively deployed for the accounting workforce.
It is now possible to create AI systems that perform tasks traditionally done by humans. AI systems can have a strong influence on human decision-making. Hence, it is imperative to consider the impact of AI on how we work.
How do we understand what AI is telling us? Should we disclose when AI helped derive a result? Importantly, why should we trust the AI results? There are many such questions that arise in our minds when we think of AI and our work.
The key answer to all such questions? Make AI human-centric.
Human-centric AI makes human users the central part of the process. Design of such an AI system makes it interact with the way humans think, act and feel. Rather than mislead or hide information, human-centric AI helps you understand what’s going on and foster trust in the results.
Accountants and auditors rely on confidence in the analysis and explainability of the results. Many aspects of AI are of critical importance to achieving these objectives. Three important aspects support these aims.
AI That’s Unobtrusive and Verifiable
One of the biggest benefits of AI-based systems is their ability to ingest and classify massive amounts of data. This is also the most common bottleneck in the audit process. Auditors go back and forth with the client to get the data necessary to plan and perform the audit. It would defeat the purpose of the AI system if an auditor had to deal with the time and tedium of telling the AI the specifics of each piece of data, e.g., which items are accounts, which are cash assets, liabilities, non-capital expenses, and so on. Not only would this be counterproductive, but it would also be prone to error.
Human-centric AI systems provide answers to these practical challenges. Such systems leverage language processing and classification algorithms to automate the judgment or context functions, e.g., identification of data items to get an accurate view of a client’s finances. Auditors can perform near-accurate manual text search with simple ledgers having limited transactions. But, with highly complex accounts, text search methods break down completely and accuracy drops to low single digits. In such situations, AI will crunch away the voluminous data with high accuracy and speed until it’s complete.
AI makes complex data classification tasks faster. Human-centric AI means it must also maintain the trust of the user. Such a human-centric AI-based platform should present the results to the auditor for verification and feedback. This helps the AI to learn where it needs adjustments and the auditors to understand what the AI is doing.
AI That Tells You What It Understands
AI should follow human behavior and expectations, not flights of fancy on its own. For example, an AI-controlled vehicle should follow the rules of the road and navigate to its destination as a human would. Imagine the outcome if the computer took the shortest path, crashing into objects and people.
Human-centric AI should similarly reflect the constraints and expected behaviors that human auditors follow. Data analyzed by AI should be considered from a position of domain knowledge and experience, e.g., audit procedures or client history, rather than making random assumptions. Even the best AI will quickly become untrustworthy without inputs from real humans on key decision points and processes.
Natural language processing (NLP) is a popular type of AI that deals with how humans interact with computers. To take some future action, NLP aims to read and understand what a human is typing or speaking. An NLP application could help auditors understand the content of client documentation. Or it could allow the user to build up complex searches within ledger data for audit planning and testing.
In an accounting context, NLP-human interactions must be considered carefully to ensure transparency and trust. NLP used to search client ledger data should offer feedback and confirmation that the auditor's instructions are understood properly. An auditor is more likely to trust an AI-based search that confirms the query first, rather than just presents the results without context.
AI That Explains Its Results
Most people don’t know how AI works. It is a phenomenon that’s referred to as the “black box of AI,” making it a significant barrier to adoption. People place less trust in systems in which the choices made are not well understood.
Let us take an example. Let's say an AI-based auditing tool reports that an area of a client’s ledger is potentially risky but it does not identify the nature of the risk. In such a case, the auditor is less likely to use the tool. If the auditor uses such tools, the client is less likely to trust the auditor as the auditor will not be able to explain the conclusions/opinions.
An important aspect of the human-centric design is the concept of "explainable AI." Explainable AI builds trust in the system and drives better adoption rates by helping the user trace the AI tool's process used to derive a result.
Human-centric AI is as much about the algorithms as it is about the user experience and design. In classic software, from a design perspective, it was highly desirable to keep the complexity of the system invisible to the user. With increased adoption of AI-based systems, to build user trust it is a must to reveal the decision-making process. AI-based solutions with built-in verifiability and explainability will give firms a competitive advantage.
When will your firm leverage human-centric AI?
Co-author Eli Fathi is CEO at MindBridge Ai, developer of the world’s first auditing tool based upon artificial intelligence and machine learning technologies to uncover errors in financial data.