“Artificial intelligence is here to stay. It’s changing our lives and creating new opportunities for businesses and engagement. It’s now up to our profession to embrace these new opportunities,” says Chris Thatcher, Partner & Leader of Global Innovation for the Audit Practice at Deloitte (Canada). Thatcher spoke at SAICA’s recent Cloud in Practice conference on artificial intelligence (AI) and the future audit firm.
While admitting that the exponential rate of digital disruption in all industries – including the audit profession - signals potentially scary times, Thatcher argued it is also the perfect time for reinventing the audit profession. “We need to understand what the different technologies can do for us, and start preparing for the change that is already happening. In fact, it is now widely accepted that we are going through a level of change that is as significant, if not more so, than the industrial revolution.”
He ascribed the disruption to several factors, including Moore’s law. “Moore’s law means that computer processing power doubles every 18 months while the cost of that technology halved during the same period.” But, says Thatcher, while digital disruption is enabled by technology, technology is not the only driving force. “A lot of changes are driven by social changes.”
According to Thatcher, some of the exponentials influencing the audit profession include artificial intelligence, blockchain, crowdsourcing, cloud, RPA and IoT (internet of things). “Many of them are already touching our daily lives and industries. But when they converge and work together, they become truly disruptive.”
But what does all this mean for the auditing profession? Does it mean robots will replace human beings? Not at all, says Thatcher. “But artificial intelligence opens up new opportunities. For example, we are already seeing AI-enabled robo-wealth advisors provide services to individuals that previously would not have ever been able to afford this type of advice.”
Can one really trust a machine above a human? Thatcher says there already are many examples where machines are trusted instead of humans, and cited examples from the medical profession. “In fact, for certain types of tasks, it is almost certain that a machine will perform significantly better due to removal of bias and being much more consistent than is humanly possible.”
Thatcher emphasised that technology for the auditing profession is still in its early stages, and that there still are some challenges. He explained various examples of AI technology already being used in the auditing profession, including a digital audit assistant and the Argus application. “Argus leverages the ability for a machine to learn and understand language by providing auditors with a digital workflow. This can assist when looking through long complex documents, like purchase agreements and lease agreements. Argus can help auditors look for key terms or information that will lead to a particular accounting or be of audit interest. So, basically a user can tag information that is relevant in the document and this can be reviewed online and then pulled into a working paper.”
But, says Thatcher, the real AI power of Argus is that the more times information of interest is tagged, the smarter the system will get. “The application is able to self-identify information of interest based on the language and context of the information. This vastly improves both efficiency and quality as now it is easier to identify information of relevance.”
AI and machine learning also helps auditors or accountants to identify unusual trends, outliers or anomalies for further investigation. “Through using these types of algorithms, you can quickly identify where to focus an audit or follow up effort to ensure that questions are asked where it matters most. The magic of AI is that it can start to identify outliers that might be less obvious or trends that may not be immediately apparent to the human eye or to previous experience. AI-enabled analytics is really the only way to deal with this massive volume of data.”
He admits that there still are challenges. “This includes getting access to the quantity of data that is required to get meaningful results. Identifying and defining what is an anomaly or item of interest from an audit perspective is another challenge. We don’t always fully understand why something has been identified by an algorithm, but there are applications that provides a full trail on how the item was flagged, leaving the judgement in the hands of the auditor. So, basically these applications help the auditor to be as smart as they can be, but ultimately leaves the final decision to the auditor.”
Thatcher also spoke about visual imaging and how the use of drones can help auditors to verify assets. “We did this as an experiment, and the level of accuracy was higher than expected. One of the most important things we learnt, however, was that observing inventory with drones is not so much about drones as it is about computer visual recognition. In other words, leveraging AI to interpret what the image is showing and turning that into data that can be analysed.” In practice, this means that unstructured data, such as fixed video logs and satellite images can be turned into data that can assist in audits.
Thatcher emphasised that implementing AI is tough. “It is not because of the technology, but because the audit profession hasn’t been organised to be AI ready.”
One of the big challenges in getting AI ready, is the storage of data. “To train algorithms, you need large amounts of data to be kept in a way allowing for machine learning. Also, it is not just the data we traditionally think about, but also metadata about users, where people clicked, what was flagged, etcetera. This is where cloud can become a very important enabler. It not only ensures the data is aggregated to allow for the volumes needed, but also allows for sharing of learning across your firm and also potentially across the profession.”
Thatcher emphasised the importance of business models where data is shared in a common platform. “This is really critical for unlocking the potential of AI technologies, and one of the design principles of our Auvenir venture. Through our own journey, we learnt that to unlock the value of automation and AI, you really need to start with standardising processes. Machine learning is basically impossible if you do not have a high degree of standardised process and flow. This is not the way audits have traditionally been done and requires an important mind shift. The reality is once information is standardised and digitised, automation is quite easy, and this is the first step on the road to cognitising the process. But while these are in actual fact human questions, and largely come back to our ability to understand and embrace change, I still maintain that AI remains an incredibly exciting opportunity for us to transform the auditing profession.”