RPA in Finance
Having been involved within RPA for approximately 6 months as of writing, the potential of this technology to improve almost all facets of businesses and even industries alike, either directly or indirectly is hugely evident to myself. And it certainly wouldn't be an overstatement to suggest that RPA could potentially induce a revolution within the workforce akin to that of the industrial revolution in the 19th century. Freeing man and woman of monotonous tasks to concentrate and fully focus on creativity, innovation and true development all whilst reducing human error and improving profits. Simply put, why spend countless hours filing excel documents, cross-referencing websites and databases and reproducing similar reports, when these tasks can be easily and effective automated by RPA software?
O FOR THE UNINITIATED, WHAT EXACTLY IS ROBOTIC PROCESS AUTOMATION?
EY gave a brief description as follows:
Robotic process automation is a software, that operates as a virtual workforce controlled by the business operations teams. A software or robot emulates human execution of tasks via existing user interfaces: it captures and interprets existing applications, manipulates data, triggers responses and communicates with other systems, it can be applied to existing applications (without changing the current IT landscape)
Information regarding RPA is freely available and plentiful, with numerous links and resources provided below; but the best way to develop an understanding is to simply play around with existing RPA software directly, with companies such as UiPath offering free community versions.
THE POTENTIAL OF RPA IN FINANCE
My formal education is predominately in physics with a more recent background in software development before switching into the frontier of RPA.
As of now I am currently employed by an exciting start-up that specialises in providing professionally trained RPA developers and consultants worldwide to interested businesses (link to our website provided below). Even In my relatively short time so far, I have contributed to huge and passionate projects stretching as far afield as Krakow and Tokyo, giving a basis to some level of international demand. Already from experience I have witnessed first hand the impact that RPA has had, especially in the financial industry and the impressive rate of growth and development of RPA; which for such a short time-span, must induce real exciting at the prospect of whats to come. Obviously, this is just my word – but with a little research, I can also arrive at the same conclusion, for example in the following report it is stated that RPA will rise from $271 million to $1.2 billion (2016-2021) –
And again another report predicts a 20% growth forecast from late 2017 to 2024, exceeding $5 billion in market share – here – hence its hard to argue against the general consensus that RPA will experience vast growth into the future.
Considering the financial sector, in which 34% of organisations still rely on manual processes, where approximately 2-5 of every 100 human performed task results in error, it still seems like a hugely untapped technology, which can be used to provide 35-50% in productivity gains for financial industries; a huge increase in profit (See). The application of RPA to finance is far-reaching and plentiful, with numerous tasks that enable a reworking of the workforce to more stimulating and rewarding tasks, such examples are provided neatly in this article; of which I hope illustrates at least some ideas.
Implementing RPA into such an institution that mostly relies on manual tasks, would reposition the workforce to concentrate and focus onto other more exciting and prosperous development matters, eliminating laborious procedures and a world of monotony – which can only be a positive in a world that rewards technological innovation and forward thinking
Furthermore RPA would allow for the reduction of costly human error, which in the financial industry can and will lead to legal implications; all whilst calculating asset deprivation, validating transactions and maintaining accounting databases – thus allowing for a signification contribution to risk management. And even more so when considering that RPA allows for automated error handling, and the ability to build and provide a detail and conclusive logging system for the clientele.
The potential of RPA for financial businesses is evident, with the writing essentially on the wall. Big players in both the financial realm are now either assessing the possibility of the implementation of a RPA solution or having passed that stage are now implementing either themselves or with the help with multinational or small companies alike. One such case study is that of SMBC in conjunction with the likes of EY, Deloitte and Accenture to reduce costs of JPY 50 Million by the year 2020, with the intention of also improving overall efficiency and the utilisation of technology (See).
This is further discussed again by EY, in the article mentioned previously who state "Robotic process automation is the Future, and Finance function needs to prepare for it" – this report also gives a small articulation about how a financial institution whether small or large would prepare for an overhaul of their own business and practices in order to incorporate RPA software; of which idealizes its maximum potential and ease of transition.
As some may already know and as I've indicated to beforehand, RPA solutions have particularly boomed in Japan; a country of a declining and ageing population, with many enterprises looking to adopt business practices such that not to overly rely on human labour. It therefore should be abundantly clear to understand why Japan has taken to RPA quicker than other places across the globe. A country where its become increasingly essential to squeeze the most out of a shrinking workforce and thus this is the place that RPA has come into it's own (See). This gives a brief view into whats to come for the rest of the world as RPA starts to enter the mainstream, and is a good starting to point to look at existing study cases if interested.
This leads directly into my own experience of collaborating with UiPath, more specifically with their office in Tokyo. I assume some have prior knowledge of RPA software, such as Blue Prism, Automation Anywhere and others – but since I don't hav e experience in them it would be unwise to comment further and hence the discussion involving UiPath.
Firstly, I was immediately impressed at how intuitive UiPath Studio was to use from the get go, with great maintainability even for people without a technical background. UiPath also provide an academy of sorts to explain how to use and the best practices within the way it was designed (https://www.uipath.com/rpa-academy). Scalability is also a non-issue when using the purpose-built orchestrator by UiPath, which simply put uses queue and threading mechanics in the background as ‘robots' to organise and implement tasks developed in the UiPath Studio environment; of which can be human or machine operated, depending on the situation at hand. Obviously, this is a very basic description of what UiPath does and I implore to you to read both their website and white papers to develop and more thorough understanding (https://www.uipath.com/whitepapers).
But the power of UiPath cannot be understated, with mechanics such as image & text recognition and all programming basics (since it was built on the VB .Net framework) all neatly packaged into a user-friendly interface on the studio environment; allowing for UiPath to be used to automate countless procedures and applied to numerous applications.
For a technical company to breed real excitement and potential, then in my opinion an active and intelligent development team is fundamental and having worked with UiPath directly I can conclusively state that this is in fact the case. With the small amount of bugs present being fixed around the clock and a team, of which contains some of the smartest and most forward thinking people I've met, of whom constantly strive to innovate and produce real substance.
AI & RPA
A real excitement for me personally is artificial intelligence, especially machine learning and the use of neural networks; and essentially any process that requires little or no human invention after setting up, to deliver an intelligence and purposeful solution. Something that RPA has to some extent (such as OCR), even though it's still in its infancy stage, but does start to struggle in this aspect when there is a sense of suggestiveness – and a human decision is required.
BUT WHAT IF THIS HUMAN INTERVENTION WAS SKIPPED AND THE RPA PACKAGE IS ALLOWED TO “DECIDE” FOR ITSELF?
A neural network, of which is vaguely inspired by the vast swathe of neurons work found within the brains of most animals, are generally designed without task-specific programming and with the main goal to improve performance of a given task with respect to time – how the network learns is specific to the network itself, but the majority can be put into categories; such as Radial Basis Functions (conceptually similar to the k-Nearest Neighbour algorithms), Recurrent Network (which have the ability to propagate data not only forward but backwards) and Modular Neural Networks. The improvement of network's ability to perform a given task is usually implemented by reinforced feedback, not too dissimilar to how one would train a dog a new trick.
A quick example I can give would be designing a network to play, perform and master Super Mario Bros published for the NES. The idea of doing well in this game is to firstly stay alive by avoiding all the dangers whilst simultaneously aiming to reach the end of the level – typically left to right.
And to control Mario you have a few controls; left, right, up, down and to button to jump. The fundamental idea of the network is then to see the environment around Mario, identify possible dangers and move towards the end of the level and determine best control button to “press” to give the highest probability to achieve this. Therefore over numerous iterations of performing this, coupled with reinforced learning – EG “telling” Mario he did a good or bad job depending on the situation. Hence with time, one would expect the AI controlled Mario should improve to the point he can complete a level without losing a life. YouTuber SethBling gives a demonstration of how this possible:
However, please take into consideration this above description is extremely simplified as the topic of neural networks is incredibly complex but incredibly interesting – I will be doing many more posts regarding AI & Neural networks, especially in conjunction with Python 3.4. – but in the meantime I have gathered a few YouTube videos that explain the topic very well and go further in-depth for those with a peaked interest.
This one gives a brief introduction into Recurrent Neural Networks
A more general description and a very enjoyable (for myself anyway) series
and finally a more in-depth series about data propagation within neural networks
The potential of coupling RPA with neural networks and AI APIs such as TensorFlow by Google (Here) and NPL (Natural Process Language) could be profound and extraordinary and turn RPA into a real juggernaut, and in a perfect world allow a business to essentially automate its entire operation end-to-end. Which of the likes is already being developed and being called Intelligent Process Design (IPA) and for anyone interested I implore you to read the blog post by my colleague (https://pathiyathrpa.wordpress.com/) and the UiPath Page (See). It seems like IPA could be the holy grail and with UiPath for sure will be at the forefront, with many google technologies already implemented and more planned.
In relation to financial institutions, of whom many use smart algorithms and artificial intelligence applications such as neural networks, combining with existing RPA software for an even more generalised solution could accelerate their business to new levels, by further automating trades and risk management and even allowing for better decision-making going into the future, along with the previously mentioned reduction in monotony and revolution of the workforce.
In conclusion its hard to argue against the revolution that is RPA and IPA, especially in the future. It's also hard to predict how far this technology will go and what future developments holds. But what it's hard to state, is that the excitement is real and right now in the present and in the world of RPA; exciting times are ahead.
I would gladly take feedback on this post and any discussion points you may have at my email: Alex.Perkins@accelerateRPA.com