How to Robotize Finance Step-by-Step

Should you start small and scale up? Or leverage the full capability of cognizant, machine learning and artificial intelligence from day one?

SSON Editorial

What's the best approach to adopting IA and RPA?

What is the 'best' approach to apply RPA to your finance services? Do you start with a limited part of a process and then scale up? Or should you leverage the full capability of cognizant, machine learning and artificial intelligence that we are starting to read so much about today? These are the questions that challenge finance leaders who are currently introducing intelligent automation into their operations. With the question of 'weather' to automate no longer an option, CFOs are pushing their finance leaders to design a finance-for-the-future model that will make the most of these new capabilities.

Harnessing disruption

The difficulty is that these options can be risky. Not just are we talking about significant investments, in some cases, but cultural resistance can, and does, derail many a good plan. In addition, finance is a sensitive department, monitored by individuals who are inherently risk-averse.

And yet. Procter & Gamble and Deutsche Bank are not alone in having recognized the significance of "disruptive technologies" in redefining their work processes. Indeed, P&G has set up its own "lab" of sorts, under the leadership of Tony Saldanha, to research ways to harness the power of technological disruption for the benefit of its global business. Similarly, at Deutsche Bank, Roberto Moncone, Global Head of Disruptive Technologies and Solutions for the banking behemoth, has taken on the challenge of developing a new business model through disruptive solutions that include RPA, cognitive, analytics, artificial intelligence, blockchain and the Internet of Things. Roberto’s work entails redesigning end-to-end processes, improving the customer journey, evaluating digital access channels and leveraging data analytics – all brought together under the umbrella of innovation.

"Artificial intelligence is a game changer that extends far beyond the more limited impact of RPA,” Roberto explains. “While RPA replaces mundane work without significantly impacting the underlying process, AI can drive an entire redesign across the process."

The challenge, he says, is that existing legacy systems are complex and unwieldy, and do not easily lend themselves to new-style digital processing. The optimal approach to leveraging the power of new technologies, Roberto explains, is via a complete redesign for the future. "What's needed is to build an entirely new core structure and IT architecture. Nothing short of a fully redesigned foundation will prepare today's organizations for the new digital environment," he warns.

For those organizations willing to take that plunge, there are two strategies: either to migrate everything from the old to the new platform in one swoop – or to run the old platform in parallel to the new, transitioning new clients onto the new platform while maintaining existing clients on the old one. "The latter is the less disruptive approach for those who are willing to resource it," Roberto says. "It also means that the costs of transition can be slowly brought down, as opposed to having to take all of it on in one go."

Deploying new technologies to boost customer experience

Technology gurus like Tony and Roberto know that it's not a matter of "if" but "when" organizations shift into a new digital model. And yet, despite all the advances that technology offers, it’s important to base the enterprise strategy on the customer experience, and not be misled by technology innovations for their own sake. At P&G, the Next Generation Services team Tony leads makes it a point to organize as a project team, not as a technical group looking to solve a problem, as Tony explains. “More often than not, the solution is not to apply RPA or IA or AI – but to eliminate or redesign the process altogether. That's something most companies have not yet understood,” he says.


But, where deployed correctly, artificial intelligence can leverage repetitive trends across any process to free up employees’ time. "Natural language processing is a simple application of artificial intelligence that is already delivering benefits across many call centres by improving the client engagement and experience. Anytime you can classify feedback that is repetitive, AI can be leveraged," Roberto says. "This includes fraud detection, analysis of legal documents, etc." 

Blockchain is the newest kid on the block, whose value, nevertheless, is already being widely recognized. Blockchain will drive significant efficiencies, says Roberto, through its distributed ledger technology which effectively connects all parties across a given, multi-step transaction with what amounts to one version of the truth. "Its application will reduce fraud and speed up transactions without lengthy delays that cost the business," he adds.

Another development that promises a significant break from the past is the emergence of smart contracts. These are digitally enabled agreements based on predefined sets of conditions that, once actioned, will cause the contract to be executed. "Triggering contracts once all the requirements are in place means that we can build significantly more complex contracts and that we can automate their execution. The impact this will have an exponential effect on businesses and will be a game changer," Roberto explains.

Start with the basics and build the toolbox

One common avenue for channelling RPA is via an existing Continuous Improvement or Lean initiative. As part of his responsibilities for process excellence at Raiffeisen Bank’s Russian operations, Dmitry Dmitriev recognized RPA’s emergence onto the scene a few years ago as opportune. "Our challenge was to always look for new ways of promoting excellence across our processes. RPA offered an opportunity to do so without too much involvement or investment in IT – and it delivered good results quickly," he says. "But I would point out that we had already achieved 30 to 60% efficiency gains in processes even without automation, purely as a result of Lean methodologies. The appeal of robotics was to stretch for even better efficiencies.”

One hurdle Dmitry’s team came up against was meeting the hyped-up expectations promoted by the media hysteria in 2016. "One robot really cannot replace 9 people, nor work 24/7," he explains. "Nor is the extent of this technology so extraordinary that you could effectively robotise half of your operations."

At Raiffeisen Bank, Dmitry opted to start simply to better understand the RPA tool and how it could work across the organization. "Because we have deployed Lean for a while already, we did not have that much obviously mindless activity for RPA to take on. Much of that kind of work had already been eliminated over the past 3 to 5 years. Also, you don’t really just ‘replace a person’. You might be replacing 10 or 20% of that person's daily activities. So it's never a simple swap.”

“In truth,” he continues, “RPA tends to work as if a person was still doing the activity.Thus, we still have a limited timeframe for the results of those activities delivered. So we found that we can’t truthfully load the full cost of robotic licenses 100%; truthfully, it's closer to 30 to 40% utilization."

At Raiffeisen, the right approach was to start with the basics, explains Dmitry, and then diversify and enrich the toolbox – “for example, through optic character recognition, which leverages cognitive capability to support the enormous number of scanned documents generated by our processes. It’s provided an opportunity to replace FTEs with cognitive-driven automated solutions.”

The real value-add of cognitive, believes Dmitry, is the ability to test hypotheses. “We are currently all scrambling to be more digital, more agile, more flexible. We are also focusing more on the actual customer and what they want. Understanding our customers’ aspirations rather than the issues experienced by the execution of our services will be key – and this is where I believe cognizant technology will be a significant help,” he ventures. “Customers’ needs today extend far beyond basic banking services. This technology should be useful in helping us understand these.”

A future step will be to add artificial intelligence into the toolbox, which will allow RPA to take on more complex activities. For now, however, some early use of chatbots’ AI elements have not been overly successful, Dmitry says.

No matter how lucrative or enticing the opportunities, keeping an eye on the cost/income balance is a big concern. "I need to make sure we are deploying our RPA licenses appropriately to get the optimal benefit from our investment,” Dmitry says.

Engage the business for a game-changing solution

Most of today's organizations have recognized automation as a valuable aspect of their work streams, deployed either by the business directly, via shared services, through a Lean approach, or an automation Centre of Excellence. And even for those who don't have the stomach for the kind of change described by Roberto, above, the benefits are significant.

Deepak Subbarao, who leads the automation work stream in Finance as part of his area of responsibility at Zurich Insurance, sees robotic process automation, even in its simplest deployment, as a game changer for the business. A big challenge, he says, is selling the concept to the business and the other stakeholders – they want to see immediate benefits before they commit, and this is best done with simple automation solutions, he says.

As for any transformative change, preparation is key. "First you need to understand the landscape; then you need to understand the technology and the provider market, before moving on to proof of concept, pilots, and production," he explains.

In rolling out RPA, Deepak took note of the enterprise’s experience with previous transformation programs. "One of the learnings from previous transformation programs was that, even though the deployment is fast, change management and sustaining the initiative is challenging, if everything goes through one team. We decided to take a different approach for RPA."

Instead, the team opted for a hub and spoke model, whereby automation practitioners were identified within the business, to act as single points of contact and were, effectively, adopted for 3 to 6 months. "These people played a key role in aligning the robotic solution to their business’s needs,” explains Deepak. “And, when we handed them back to the business, they played a vital link in sustaining the solutions.”

This approach has served Zurich well, and although it may take a little longer, as the team approaches business by business, in terms of sustaining the momentum and driving change management “we have found it works much better," Deepak confirms.

Choosing the right process: keep it simple

While senior leadership may push for automating highly complex activities as soon as possible, thereby addressing a burning platform, results tend to be best when starting with simpler activities, most agree. The initial choice of process is particularly important because it is frequently used as a proof of concept and to overcome internal resistance. With Zurich’s RPA initiatives now entering their third year, Deepak concedes that the incremental approach of starting simple and then building up, has indeed worked well. "The more players in the game, the more your initiative will be slowed down," he explains.

At Zurich, the priority was to solve capacity constraints. "Prior to RPA, some teams were feeling the pressure of being under-resourced,” Deepak says, “so we decided it made sense to support them by freeing up time to do more value-adding work. It was a rather high-stakes approach, you might say, but we were confident that if we could relieve the pressure, we could deliver immediate benefits to the business and make an impact. An additional benefit was that the scale of work could increase without the requirement to add headcount."

The actual choice of optimal process will vary across businesses, but at Zurich it separated into high volume operational activities and core mid- or back office processes that were more specific. "We knew that many of the 'know your customer'-type activities [e.g. opening accounts] would provide good opportunities for robotics as they are high-volume, but we also wanted to focus on activities that were not necessarily volume-based, like quarter end closing, to prove that RPA would work even for small volumes,” says Deepak. “Even just a handful of people regularly working overtime were a flag highlighting an RPA opportunity. So we started with some of these core activities in finance.”

Building and scaling

If the initial proof of concept is done right, there should be a buzz around RPA felt across the enterprise, resulting in lots of work coming your way. Deepak’s team was quickly overwhelmed by demand. "The advantage of a pull-based approach is that you have a pipeline from which to source opportunities that fit your key criteria,” he explains. “It's also important to set realistic expectations. The danger of setting high expectations and then not meeting them means you succumb to the hype of RPA, and the business will quickly pull back. Once you've lost that initial enthusiasm, you won’t regain it easily."

Although scaling is a big advantage, and a much-cited incentive for robotic automation, there are certain activities in finance that cannot be scaled as a result of either geographic or business boundaries, or regulations. "The risk-averse nature of finance means there is not much enthusiasm for opening up access to all the systems," explains Deepak. "Your bandwidth is therefore dependent more on the structure of the business than it is on the capability of the automation solutions. This constrains our ability to scale somewhat."

That only applies to some activities, of course. Standard operational activities like opening accounts or managing claims, which are volume-based, lend themselves well to scaling up and out but more specialized areas such as reconciliations, finance accounting, or reporting, Deepak explains, may require a silo-based, focused solution that meets the specific needs of the business.

Adopting cognitive capabilities

While cognitive is creating a lot of interest as the next step in intelligent automation, the reality is that these solutions do entail costs, which are not insignificant, and require a certain knowledge and skill set that is not necessarily readily available within the business. In addition, there is a missing data element that is still not sufficiently recognized, hinging on the lack of common definitions and the challenge of historical data not always being housed in one place or easily accessible.

Cognitive capabilities will certainly be driving more intelligent automation solutions in the future, Deepak concedes. "Certain problems cannot be solved with just a piece of technology, even if it's automation technology. For example, in dealing with loans you come up against the fact that you'll never achieve real-time settlement through process automation alone. That's where cognitive comes in."

Where cognitive solutions play a significant role is in their ability to pull in new data points, for example where a process requires a lot of 'thinking time'. Exceptions processing resulting from automated execution, is a case in point. "If you can link the knowledge inherent in your data to the RPA part of the activity, you will leverage the entire process up a notch," Deepak says.

With more providers offering cognitive- or even AI-driven elements as part of, or bolted onto, their solutions, solution boundaries are blurring. However, the key is to build a strong foundation to leverage for future growth. Artificial intelligence is still predominantly theoretical rather than a practical solution in certain business areas, Deepak concedes, but it is catching up.


Anyone who has embarked on RPA knows it works. Indeed, where projects have ‘failed’, these failures tend to be traced back to incorrect implementation or faulty choices up front. Once RPA has proven itself, to drive the benefits beyond the sample will require either scaling up, rolling out, or applying a different kind of automated enterprise platform, as per Roberto’s suggestion.

Introducing cognitive and AI solutions will allow enterprises to tackle more complex processes. While both are still at the proof of concept stage, the value is certainly there, specifically at the front end of operational activity.