Big data, data analytics, machine automation, AI - why should these concern the auditor of tomorrow

For decades, the audit process has remained fairly similar – the client’s initial results are extracted from their records and briefly analysed, with unusual and/or risky areas identified to be a key focus during the audit itself.

When auditing large companies, auditors are trained to pick a small, relevant sample and this is tested in detail.  Files are pulled from shelves, photocopiers overheating as copies of records are made, with the trainees following the ‘audit trail’ from cash receipts to invoices to sales orders and back again.  Each audit may take weeks and people work on individual areas, with discrepancies found in different places potentially not spotted for some time.  The audit management and partners then use their professional judgement and expertise to decide whether a company’s stated results show a ‘true and fair view’ of what has occurred in the year.

Data analysis is changing

Accountants have always analysed data – introductory courses on finance generally include financial statement analysis and discussions around what this implies about the state of the company.  However, the type of analysis performed is changing.  Audit firms have now developed data analysis and visualisation programmes that can test entire populations –thousands of transactions rather than the small sample they managed historically.  This can also be done in a fraction of the time – meaning much of the audit work can be done in almost real time, rather than months after the event.

The programmes automatically flag ‘exceptions’, allowing auditors to identify unusual results or potential fraud areas with ease.  And the results can be seen in a variety of easy-to-understand dashboards, increasing the chance that anomalies may be spotted.

Risk and evidence

One question currently being addressed is how this process marries with the current auditing approach based on risk and evidence.  Fundamentally, the quality of a piece of audit evidence is assessed as being ‘sufficient’ (enough of it) and ‘appropriate’ (of the right quality).  Clearly, if auditors will now be analysing every transaction rather than just a sample, it should be sufficient, but is it appropriate?  Is the audit trail clear enough?  Is this data ‘evidence’ itself – and, if not, what evidence should be collected given that very soon there may be no ‘paper trail’ at all?  How is ‘completeness’ of the transactions tested?  If we analyse 100% of the data then there will clearly be more unusual results than in a small sample – what level of error is acceptable?  Perhaps the importance or number of anomalies will be the benchmark in future.

A planning activity

Under current auditing standards, analysing data to identify risky areas is a planning activity – to include it as evidence, the calculations must be performed in order to verify a pre-held expectation, which is potentially unlikely here.  Currently, the main emphasis is that auditors must know why they’re using data analytics and what they’re looking for when using them – they are just a tool to help do the job, like many other audit methods.

Will human auditors become obsolete? Luckily for those starting out, the answer is a resounding ‘no’.  In lots of ways, the journey will be more interesting – machine automation will almost certainly mean the end of traditional book-keeping and manual interrogation of client information.  The audit process itself will likely change too – instead of starting with a trial balance after the year end and drilling down into it, as is the current method, algorithms will scan the data continuously – bringing to the fore outliers for further analysis in almost real time; thus the audit may be almost complete by the end of the period.

In its place, even those new to the profession will be tasked with more difficult challenges – understanding the data set given, identifying which anomalies may be important and interpreting the results.

Humans vs machines

In order for this process to work, it’s clear that data scientists are going to play a key role in preparing the data and initial exploratory analysis.  But it’s the ability to interpret and use judgement that accountants will bring to the party.  Despite reputation, accountants have always needed creativity (alongside logical thinking and problem-solving) to succeed – and these ‘human’ traits (similar to many other industries) will be valued more than ever in future.  Accountants are business experts – and this will differentiate them, with automation allowing them to focus on problem-driven analysis to solve their clients’ business problems and drive their businesses forward.  Successful auditing is an art, not a science – there are difficult judgements to be made: What level of error is acceptable?  Are management’s explanations acceptable, or not?  No computer can conclude on these things on its own.

So, what traits should the auditors of the future be fostering?  Well, to bridge the gap between data scientists and clients, analytical and soft skills will come to the fore – understanding business models and accounting principles, using them to extract what would be most relevant to the client, and being able to explain what this means for them in a clear, constructive way will be essential – without good communication, what differentiates us from the machines is lost.

Jennifer South FCA MA (Oxon) is an academic tutor (Accounting and Finance) at Pearson Business School, part of Pearson College London, the first higher education institution in the UK to be founded by a FTSE 100 company – Pearson Plc.