7 success factors for Robotic Process Automation (RPA) implementation

01 October 2018 Consultancy.eu

Robotic Process Automation is touted as a digital disruptor in the operations realm. In the constant race to outperform competition, companies are looking for ways to streamline processes, reduce costs and focus on value-add activities. RPA is a promising solution for these problems, and so it is no wonder that large and mid-sized organisations across industries are currently piloting, or implementing Robotic Process Automation.

As with all new technologies that enter the stage, adopting and embracing Robotic Process Automation is a daunting task, with reaping the benefits posing an even larger challenge. Several analyses of early use cases have meanwhile shown that rushing into RPA without a well-thought out approach can lead to costly mistakes. If RPA implementation projects are not managed properly, they will either fail, produce undesirable results, or cost much more than planned. 

Esen Orhan and Marc Geleijn, Director and Associate Director respectively at Protiviti, have been involved with RPA implementations since the dawn of the technology. Building on their insights, as well as leveraging Protiviti’s global track record and knowledge in the field, the two advisors have crafted seven success factors that they believe are key to a successful RPA implementation.

RPA strategy

Before starting any RPA implementation, it is imperative to have the basics in place – the right plan. A company’s RPA strategy should be directly linked to the IT strategic roadmap and business function operational plans. This ensures the RPA Programme’s goals and objectives are aligned to the organisation’s goals and objectives.7 success factors for Robotic Process Automation (RPA) implementation


Common to any change transition, people hold the key, as people are ultimately responsible for driving, and accepting, the change throughout the organisation. This ranges from leadership having to be change leaders, committed and aligned in their views, to managers and professionals on the floor having to be open to change and new ways of working. 

Once the RPA strategy and core implementation team have been identified, five subsequent factors become crucial for successful RPA implementation:

RPA Implementation Partner

Most organisations do not currently have a team of RPA professionals on staff to help deliver a RPA Programme, and as such companies are looking to RPA Implementation Partners to deliver the RPA Programme. The key to selecting the right RPA Implementation Partner is to analyse the fit with the organisation. 

The partner should understand your industry and process areas being considered for RPA. It sounds straightforward, but most implementation partners will state they can implement RPA across a spectrum of industries and processes – which is not necessarily the case. Organisations should challenge potential implementation partners and ask for RPA specific client references similar to your organisation. Ideally the partner has already worked with your organisation and understands your business, processes and information systems.

Depending on the size of the RPA implementation programme, it is also important for clients to select a partner who can meet staffing needs. RPA is growing at a rapid pace and implementation partners are struggling to keep up with demand. Organisations should ensure that potential implementation partners can truly meet staffing requirements within the required timeline. The expectations and final deliverables should be clear before selecting a RPA Implementation Partner.

Protiviti has seen clients select and rely upon implementation partners that were inexperienced or understaffed, resulting in a range of problems including poorly selected processes for RPA implementation, undefined return on investment (ROI) targets, project delays, budget blowouts, and RPA technology that was not optimal for the client’s needs. The RPA Implementation Partners either did not perform a proper RPA readiness assessment, did not have appropriate project governance, or did not manage organisational change appropriately.

Process Maturity

Not all processes are suitable for RPA. RPA generates the best ROI when implemented on processes that are labour intensive, repetitive, rule based, use structured data, and have a limited number of process exceptions. Essentially, RPA is suitable for mature, defined, repetitive, data heavy processes. In order to assess if processes are potential candidates for RPA, a process maturity and RPA readiness assessment needs to be completed prior to RPA execution. 

Ensure the RPA business case and target ROIs have been set before performing the process maturity and RPA readiness assessment, to ensure only high ROI candidates are selected. From our experience, the processes which are the best candidates to complete a proof of concept that generates good ROI are accounts payable, accounts receivable, general ledger reconciliations, employee onboarding and customer onboarding.

"RPA enables companies to employ automation for routine tasks, which in turn frees up talent to add value by taking on more qualitative and strategic level initiatives."

The results of the process maturity and RPA readiness assessment will create a shortlist of RPA process candidates. The assessment can also identify a second tier of potential RPA candidates, where minor process redesign or standardisation is required prior to being automated. This allows organisations to remove as many redundancies as possible before implementing RPA, and ensures greater efficiency of implemented robots and reduces the number of exceptions and potential errors.

If a large RPA programme is being implemented, start automating the easiest processes first. The acquired knowledge and experience from implementing the easy processes can then be used to automate the more complex processes. This may sound simple, but remember, RPA is new to your organisation and potentially the people implementing the programme. 

Project Management

RPA implementation can be a large scale programme or a much smaller project, either way, project governance and project management practices need to be applied for successful implementation. The first step is to build a project management team and define the governance structure for RPA. Ensure the team consists of key personnel from the relevant business unit(s), IT and project management professionals. The next step is to secure executive and process owner buy-in. Together the project management team, executives and process owners will determine the goals of the RPA programme, such as reduction of man hours, errors, costs, or improving the efficiency or quality of work performed. The goals of the RPA programme will help define the RPA business case, ROI and establish rating criteria to be used in the process maturity and RPA readiness assessment.

Once the project management team and RPA programme goals are established, an RPA implementation plan can be developed, including key items such as change and communication management, and the establishment of an RPA Center of Excellence (COE). 

Change and communication management is crucial for all departments and employees impacted by the introduction of RPA technology. Employees are nervous when they hear their job may be replaced by a robot, however this may not be the case (depending on the organisation’s RPA goals). Some roles may be replaced by robots, and therefore change management should include retraining or redeploying of personnel, and the restructuring of departments. For other employees, RPA is a new business tool which eliminates monotonous tasks so they can focus on value add activities, which is exciting.

RPA Technology

When selecting an RPA Technology Provider it is important for organisations to check the software’s compatibility with your organisations systems. Just last year, one of the leading RPA technology providers was having trouble with utilising Google Chrome, rather than Microsoft Internet Explorer for web based processes. So it is imperative your procurement process includes a comprehensive IT requirements review. The IT requirements review should consider items such as technical support, maintenance, security and data standards, hardware and software requirements, licensing fees and implementation costs.

RPA software is fairly intuitive and easy to use, however organisations considering RPA should still review the RPA process design and configuration interface. In general, non-IT professionals should be able to configure and monitor the robots. All advanced RPA technology providers have a control room feature, which allows an organisation to schedule and monitor robot activities. The control room also highlights robots which did not execute properly and provides an error message noting the point of failure. RPA software should also provide user access restriction functionality and comprehensive audit logs for each robot, to ensure RPA is appropriately controlled. 

IT Involvement

RPA is sold to business units as a low IT impact solution to business problems. Although this is somewhat true, RPA is a software and therefore its implementation is a software implementation that requires IT’s involvement. It is critical that the IT department be involved in the RPA implementation early. Business units need to obtain IT executive buy-in early to ensure RPA aligns to the organisations’ IT strategic roadmap. RPA may not be needed if full system functionality can be utilised, or if system updates or upgrades are expected in the short to medium term.

IT is also ultimately responsible for ensuring the RPA software complies with the organisation’s technical specifications and security and data standards, therefore IT needs to be included in the RPA technology procurement process. 

Furthermore, IT will provide practical support throughout the RPA implementation, for example, access to hardware and IT infrastructure, system access, new user set up for the robots, and technical support as issues or access problems arise. IT will also play an important role in user acceptance testing (UAT), including follow up testing and performing impact analysis.

Once the software is implemented, IT’s role will change to include notifying your RPA center of excellence of impending IT changes which may impact operational robots, can assist with technical support, manage change requests and execute maintenance requests.

Closing thoughts

RPA will continue to take hold in organisations as RPA matures and practices become proven. Companies will continue to develop methods to employ greater precision and efficiency in a variety of processes to further utilise RPA across business units and processes. RPA enables companies to employ automation for routine tasks, which in turn frees up talent to add value by taking on more qualitative and strategic level initiatives.


More news on

AI can improve operational risk management in banking

17 April 2019 Consultancy.eu

Risk management is an integral part of banking. By taking financial risks, banks are able to generate the profits that are necessary to survive. Risk management aims to control this process by making potential losses more predictable. This makes the bank more robust to external fluctuations.

Whereas profits can be made by accepting certain financial risks, operational risk is intrinsically different. Operational risks only cause losses – financially in terms of bottom-line impact and non-financially in the form of for instance reputational damage. The consequences of operational risk events can have a large impact on an organisation and the financial system as a whole as experienced during the last financial crisis. It is not surprising therefore that operational risk is receiving more attention within the financial sector, with banks trying to minimise the operational risks they take, given the resources available while keeping in mind the strategic goals of the organisation.

In the past decade, there has been major progress in the development of artificial intelligence (AI). AI algorithms excel at data analysis and have evolved to the point where they surpass human performance for a wide variety of tasks. More and more businesses exploit these technological advances to optimise different kinds of processes such as marketing, sales and e-commerce, manufacturing and logistics. In today’s growing data-driven world, this trend is expected to continue on the back of widening opportunities for use cases, including in the area of operational risk management.AI can improve operational risk management in banking

Challenges in operational risk management

In June 2011 the Basel committee published the Principles for the Sound Management of Operational Risk (BCBS), which provides a framework for the development of proper operational risk management. Three years later, a survey was conducted to measure to what extent banks complied with these principles. One of these principles states that banks should write a risk appetite and tolerance statement. Banks reported that this is more challenging for operational risk than for other risk categories and attributed this to the nature and pervasiveness of operational risk. The banks that did comply with this principle often reported the use of backward-looking metrics of operational risk, such as operational losses as a percentage of gross revenue.

The above example underlines the challenges which banks face in the management and measurement of operational risk. Compared to financial risk, operational risk is a more qualitative field of study. Whereas financial risk management has been the main priority of banking for a longer time, operational risk management is much younger resulting in less extensive historical data. Predictive modeling becomes more of a challenge in this situation.

On top of that, the events in operational risk are much more diverse in scope. The Basel committee defines operational risk as the “risk of loss resulting from inadequate or failed internal processes, people and systems or from external events”. Internal fraud, data leakage and reputational damage are very different problems, yet can be very closely related as well. A lot of the data in operational risk consists of textual input which contain qualitative information. The qualitative nature of operational risk is reflected in the Basel framework, which encompasses guidelines for organisational structures, culture and awareness, and qualitative reporting.

The computer as a reader

Artificial intelligence could play a valuable role in upgrading operational risk practices, with in particular machine learning – the field of self-learning computer algorithms – showing promise. Machine learning algorithms can make predictions based on data fed to the algorithm. Recently, major progress was made in the field of natural language processing (NLP). NLP focuses on using textual data for predictions. As an example, an algorithm can learn to rate hotel reviews. By processing large amounts of reviews together with their given ratings, the algorithm can learn to give a rating to a new review it has never seen before.

“There are many opportunities for operational risk management to exploit AI and other related technological advances.”
– Lars de Ruiter and Matthias Geerse, Solid Professionals

In operational risk, many textual reports are written regarding specific risks or risk appetite of the organisation as a whole. These reports are generally written by risk managers – experts in their field. Assuming that risk reports hold information which is absent in historical loss data, this information could be extracted and used for predictive purposes. So why should computers perform this task instead of risk experts? The answer lies within the physical limitations of human being – as well as the natural biases they have to cope with. While humans may struggle to remember a piece of text, a computer algorithm easily processes thousands of books and finds structure within using statistical methods with near-perfect precision.

Towards prediction of operational risk

The performance of NLP algorithms using word embeddings has increased tremendously of late. Using the embedding as a starting point, these algorithms learn the meaning of full sentences and use this information for further predictions. This is more generally called sentiment analysis. A popular application in the financial setting is predicting stock price movement from news articles, or twitter feeds. A recent paper by Denmark’s national bank predicts corporate distress of firms by analysing their annual reports. Taking textual data under scrutiny can make a contribution to quantitative predictive modelling, even in the financial sector where the use of words is very different to day-to-day language.

In summary, there are many opportunities for operational risk management to exploit AI and other related technological advances. One possibility is the classification of risk events. As an example, a computer algorithm can read risk descriptions written by risk managers and classify them according to their impact and frequency. Combining loss data with risk reports, improved prediction of risk events might lead to more accurate prediction of future losses. A different application is the measurement of more abstract concepts such as the financial health or maybe the cultural aspects of a company.

An article by Lars de Ruiter and Matthias Geerse. Both are finance & risk consultants at Solid Professionals, a consultancy from the Netherlands.