Jonathan Ebsworth, partner of the disruptive technologies practice at Infosys Consulting

Earlier this year, Samsung Securities gave 2,000 of its employees an early bonus which must have come as a welcome surprise – especially as this unexpected present was in the form of shares to the value of $105 billion. While no-one doubts that Samsung Group’s employees work hard, an individual award of some $50m per employee seems rather excessive – not to say unfair on the other workers who received nothing.

The problem, of course, is that the employee responsible for issuing the shares had made a catastrophic mistake: in fact, each employee was due to receive a dividend of two billion won, or just under a dollar for each share they owned. But what actually happened was that the luckless administrator accidentally issued two billion of shares.

While the Samsung story is just the latest in a long line of ‘fat finger errors’, human mistakes can cost a business billions of pounds – quite literally. Such mistakes can also lead to terrible reputational damage and compliance liabilities, as happened here. In the 37 minutes it took to correct the error, sixteen Samsung Securities sold their shares for almost $10m each, in spite of having been warned not to by their managers. The lure of a payday beyond the realms of an ordinary worker’s imagination clearly outweighed the consequences of getting fired.

Eliminating risk 

We all make mistakes, but few of us have to face the same consequences as this poor Samsung Securities’ employee, who miss-type wiped 12 per cent of the company’s stock price. Businesses in the financial services industry are uniquely vulnerable to these sorts of mistakes, and not just because of the volumes of high value transactions they make every day.

The answer is to explore where robotic process automation (RPA) can be employed to eliminate human error (or, indeed, deliberate fraud). There is a lot of fear around automation ‘stealing’ humans’ jobs, but for repetitive, rules-based processes involving high-stakes transactions, it’s more likely to safeguard our jobs by preventing cataclysmic mistakes that can cost a company hundreds of millions of dollars at the stroke of a key.

Automation isn’t just about protecting against errors, important as that is. There is enormous potential to bring RPA into other areas of finance; for example, in areas such as compliance, anti-money laundering activity, and Know Your Customer (KYC) initiatives. What’s more, RPA can bring huge cost- and time-savings by automating many of the tedious, process-heavy transactions such as account opening or customer service.

  Building a vision for RPA

With so many clear benefits, one would expect financial services businesses to have embraced RPA; however, recent research by Infosys Consulting revealed that only 10 per cent of organisations currently using RPA or AI believe they are maximising their full benefits and capabilities. For example, a single RPA agent involves a one-off cost of between $5,000 and $15,000 – far cheaper than even the most junior employee. Failing to take advantage of bots represents a major missed opportunity, and the reason is more often than not a lack of clear strategic vision for RPA, and poor understanding of the requirements for effective implementation.

We should all be tremendously excited about RPA’s potential; to achieve this, however, businesses must take a pragmatic and strategic approach to bots in the enterprise. Here are our five steps to RPA success:

• Bots are no panacea

For repetitive, transactional tasks, bots are around three times more efficient at certain processes than an equivalent human worker, but that doesn’t mean that a business should conduct a wholesale replacement of their existing employees. Bots are good at routine processes, but finance is about much more than hitting buttons at the right time – it requires workers with intelligence, intuition, and problem-solving abilities. Organisations shouldn’t calculate their RPA strategy on like-for-like replacement, but must give careful thought to how bots and humans complement each other in various roles such as compliance, invoicing or customer service – as well as considering the costs of retraining, redeployment and sometimes organisational adjustment required.

• Planning for disaster

It’s not technology itself that tends to cause us problems, but rather our over-reliance on new tools and apps. This is especially true when a business faces a business continuity problem such as a power or network outage, or a cybersecurity breach. If the bots go offline for any reason, will your business be able to keep functioning – and will your company retain skilled employees who can take over these processes when disaster strikes? Business continuity questions like these are often overlooked, but should be at the heart of any RPA strategy.

• Updating your security strategy 

Cybersecurity threats are more prevalent than ever. As we pass increasing responsibility onto non-human actors such as RPA software, we need to consider how we update our security strategy to accommodate this changing dynamic. For example, bots aren’t well-suited to authentication methods such as biometric or two-factor authentication. Failing to integrate bots fully into your organisational security plans risks creating new vulnerabilities that could ultimately be just as costly as any ‘fat finger error’.

• Staying agile 

A tactical or poorly-planned RPA deployment can significantly reduce the agility an organisation has, tightly coupling automated processes to the underlying platforms. Automated processes can be quite fragile and particularly sensitive to even minor updates to the core systems they drive. Bots and AI solutions at scale should be governed within the overall architecture framework that underpins the business and not as a stand-alone solution sitting outside of the enterprise architecture. 

• It’s a steep learning curve for everyone 

Automation is still a very new technology, and there are many lessons that we need to learn before we can deploy it uniformly across a business and iron out potential risks and errors. That doesn’t mean that we can’t wring enormous value from RPA today. However, we should at least be alive to the potential risks and consequences of introducing these new capabilities. The secret to success is knowing how to complement human activity with RPA, rather than have them compete. That’s why it’s so important to develop a strong business case for every RPA implementation, based on an awareness of where human error can and should be eliminated – but also considering where humans, in spite of their fat fingers, can continue to add value to the organisation.