Following on from 'Developments in IT and the impact on performance management – part 1' (see 'related links'), this article looks at some more recent developments in IT; specifically process automation, artificial intelligence (AI), data visualisation and the internet of things.
Process automation is the concept of processes being performed by machines rather than by humans. Machines can perform some repetitive processes, to a consistent standard, quickly and without errors, so are often better at performing these tasks than the human. Robotic process automation implies the use of computer software in the automation.
In manufacturing industries, robots are common, for example in car production lines, where they perform many of the assembly tasks. In services industries too, many processes are being automated, such as bank transactions being processed entirely without human involvement. Robots are even assisting in surgical procedures in hospitals.
Automation is not a new concept. The first completely automated industrial process was a flour mill, developed by Oliver Evans in 1785. Recent developments in computer technology are providing scope for greater automation. These developments include better hardware and software, and developments such as artificial intelligence described below.
A report by the Mckinsey Global Institute 1 published in 2017 claims that in the global economy, 49% of the tasks that are currently performed by humans could be automated using technology that already exists. Only 5% of occupations could be fully automated, but at least 30% of the activities performed could be automated in 60% of jobs.
From a global perspective, Mckinsey see automation as leading to greater productivity and higher economic growth. They predict that since the jobs that are automated would be replaced by new types of employment, automation would not lead to higher unemployment. For businesses, the benefits of automation are not limited to labour savings. Benefits include 'greater throughput, higher quality, improved safety, reduced variability, a reduction in waste and higher customer satisfaction'.
How will automation impact on the work of the management accountant? The tasks most likely to be automated include the more basic accounting work such as the collection and processing of data. Many accounting software packages upload bank transactions from the bank’s systems, and supplier invoices scanned by smart phones can be automatically booked. The work of the management accountant which can’t be automated are the more advisory aspects of accounting – interpreting and analysing information and providing recommendations. This reinforces the role of the management accountant as a consultant and advisor to organisations, in relation to strategy development, decision-making and value creation.
Artificial intelligence (AI) and machine learning
Artificial intelligence (AI) can be defined as 'The ability of machines to exhibit human like capabilities in areas related to thinking, understanding, reasoning, learning or perception'. 2 What this essentially means is machines that can think for themselves, like humans. Science fiction films provide many exciting examples of intelligent machines, such as C-3PO in the Star Wars films, but these do not yet exist in the real world.
Computer scientists talk about two levels of AI – weak and strong. Weak AI means that machines can think for themselves, but only to the extent of doing specific tasks, for example cars that can drive themselves. Strong AI means that machines have a general level of intelligence and can think like humans. Strong AI currently only exists in sci-fi fantasy, but weak AI is an area that is developing very quickly and has innumerable applications to business and society in general.
Early applications of AI included expert systems. Computers were programmed to make decisions that previously experts had made, such as giving quotations for car insurance. Relevant information about applicants, such as their age, gender and driving experience could be entered into a system, which would then evaluate their risk category and quote an appropriate premium, thus performing the role of an actuary. Other examples include systems which support doctors, whereby the doctor can enter information about the patient’s symptoms, and the system will help to provide a diagnosis.
Expert systems can only make decisions based on rules that have been hard programmed into them by experts. Over time, their output for a given set of circumstances does not change, unless they are reprogrammed. With the advent of 'machine learning' machines can 'learn' from their experience, rather like humans do, and therefore provide more accurate output. Some accounting systems 'guess' where the debit side of a payment or the credit side of a receipt should be booked, based on what the bookkeeper did last time a similar receipt or payment arose.
Machine learning is used extensively in data analytics where computers performing the analysis learn more about the data population with experience. To take a simple example, imagine that we wanted to analyse all the companies on the stock exchange to predict which might go bankrupt within 12 months. We might programme a computer to calculate a score, such as Altman’s Z-score model, and use this as the basis of our predictions. This is not AI as the computer is just doing what it is programmed to do. Unfortunately, the Altman Z-score model does not predict perfectly, sometimes classifying companies that are at risk of bankruptcy when they subsequently survive or failing to predict the bankruptcy of others. These errors are referred to as classification errors. Without machine learning, programming the computer to simply calculate z-scores would not reduce the probability of these errors.
Alternatively, we could programme a machine to analyse a sample of historic data and come up with its own version of the Z-score model. Machines can analyse much larger volumes of data than humans, so the machine would come up with a much more reliable version of the z-score, possibly incorporating many more variables than the five used in Altman’s model. Here the machine is learning from the data, so this is AI. There would still be a probability of classification errors, but it is likely that this would be much lower than in the Altman model. What is more, the machine would continuously update its model as it analysed more and more companies. This demonstrates one use of AI, in classification of data.
There are several areas where artificial intelligence and machine learning are relevant to management accountants. Firstly, within data analytics, analysis that incorporates machine learning techniques can provide new insights to management. One of the roles of the management accountant has been to analyse data for the purposes of planning, decision making and cost control. For management accountants to remain relevant in this brave new world it is essential that they understand, and input into the AI which is being used in analysis.
AI can also be used for identifying unusual transactions that may help accountants to become better at detecting fraud. As fraudsters are continually developing new methods to practice their trade, machine learning provides an opportunity for accountants to keep up with them.
AI can also help accountants provide much more accurate forecasts, based on a more thorough analysis of the external environment, using machine learning to identify with greater accuracy the factors that will affect a business’s revenues and costs.
Data visualisation techniques
Data visualisation refers to presenting data using visual techniques such as charts and diagrams so the story behind the data can be seen easily. As the old saying goes, a picture is worth 1,000 words. Providing information visually can assist decision makers to understand data much more quickly, providing that it is presented in a way that helps their understanding.
Charts and diagrams have been with us for hundreds of years, so there is nothing new in the concept of data visualisation. A classic example is the map of the London Underground that was designed by Harry Beck in 1931. What has changed in recent years is the volume of data that is available for use by businesses, from devices such as smart phones, smart devices on the 'internet of things', the explosion of social media, and the use of sophisticated data analytic techniques to assist in analysing all of this data. There is also increased demand from management for analysis of this data.
New technology has also become available to help perform data visualisation. Market leaders are Microsoft Power BI, Tableau and Qlik. Such packages enable users to access data from many different platforms and provide reports that update in real time. They help users provide many different types of visualisation using built in templates with high quality graphics.
From a practical point of view, businesses need to consider where the data for the reports comes from. It is important that the data can be extracted easily, particularly where it comes from several sources. It may be necessary to have a data warehouse where the data from different sources is deposited. The data can then be cleaned if necessary. The reports would extract the data from the data warehouse rather than from the original sources.
Common types of data visualisation techniques are as follows:
Dashboards contain summarised information, typically by showing a small number of key performance indicators. They allow managers to see the big picture quickly, focussing on the critical success factors. They may also include drill down facilities whereby managers can click on the number in the dashboard to see more detailed analysis. This can give managers the ability to answer some of their own questions.
Waterfall charts, also referred to as bridges, can be used to show the components that make up a total. The components are shown as bars, where the length of the bars represents the value of the component. It will also show whether the components increase or decrease the total. A simple waterfall chart showing variances for the month is shown below: