For example, when you get the latest version of any good data visualization software, such as QLIK, you will be able to use a “bot”. Using the “bot” you can type a question: “what were my sales in 2020”? And before your eyes a chart showing the sales of 2020 will appear. This is really what we mean by machine learning. It is getting the machine to understand the human-formed requirement, and doing the obvious work, that any decent data analyst would know how to do. The machine interprets the sentence and does the work to get the result.
And we can go further. We can then say to QLIK bot: “what about the margin?” In this question, the bot will be “in context”, meaning that it knows the context of the conversation. It knows that we are talking about 2020 and it will give you the margins for 2020. The machine has “common-sense”.
So, these things are coming. And it is true that if you have an army of data analytics people in your office, you may not need so many of them in the future. Instead of asking your data analytics team to produce a dashboard on sales margin per product, you will be able to write the sentence “Sales margin per product” and QLIK will present you automatically with a set of options of dashboards for sales margin per product.
You might be thinking – but how does QLIK know where to get the data? Well, probably someone at some point is going to create standardised data models for each major ERP (Enterprise Resource Planning: aka accounting tool) system, so that it can become an off-the-shelf SAP robot for QLIK.
You might also be thinking – but how does QLIK know what to do if some of the data is from the ERP and some of the data is from elsewhere? So yes, probably you will still need someone to understand what data is in Excel and elsewhere, and correctly give that data to your robot, so that the robot can identify what the data is, what it is used for, etc. QLIK will know all of these things because the data will be set-up in a Data Lake, by your IT department and it will be categorized and organized correctly so that the robot can easily ask questions of it.
So, if you are in data analytics for audit, you may be wondering what the future holds for your particular skill-set right now.
Well, marketing people do like to say that everything works fine before it does, so we still have a few more years before we can truly speak to our computers. But probably only a few.
In the relatively near future it means that the auditor will need to focus on those more functional skills of actually interpreting the information received.
For example, if you receive a dashboard from your system with the margin per customer, then you might want to just copy-paste it to your audit report, to show that all is good because the margins are all positive.
But if you are a good auditor, you might think, that all looks fine, but are there any products that have a negative margin? Maybe, if you ask the questions like that, the machine will always come up with the answer, “no”. What if you asked the question “are there any customers+products that have a negative margin?”. Then the machine might say “yes, customer ‘Mercedes, Germany is getting product ‘leather upholstery’ on a negative margin basis for the last 3 years in a row.’”
The machine has given a very interesting answer. Now, the auditor could think, “great I can put that in my report.”
But if you are a good auditor, you might then think.. “ok, but why?” Then you could start asking other questions, such as: