Robo Treasury: Positives and Pitfalls

Robotic process automation has the potential to be a powerful tool for treasurers – especially those looking for new ways to garner efficiency gains. What’s more, as a relatively easy-to-deploy technology, RPA can deliver significant value in a short timeframe. But is RPA really a silver bullet for treasury? We ask four industry experts for their opinions.

Whether you’re an R2-D2, RoboCop, or WALL-E fan, stereotypical movie views of robots aren’t helpful when it comes to discussion around robotic process automation. Unlike the all-singing, all-dancing robots of science fiction, RPA is simply an automated rules-based software that executes pre-programmed tasks.

Nevertheless, RPA is making waves among the treasury community. According to PwC’s 2019 Treasury Benchmarking Survey, 47% of treasurers believe RPA will be either relevant or highly relevant over next two-to-three years. The survey also identifies interesting use cases for RPA – ranging from payment execution to accounting and management reporting (see fig.1).

Fig 1: RPA Use Cases Source: PwC Treasury Benchmarking Study 2019

While the opportunities around RPA are significant, there are limitations to consider as well. To get a balanced view on all of the RPA arguments, TMI gathered together the following experts for a roundtable discussion:

Eleanor Hill, TMI (Hill): Let’s start with the basics. How would you define RPA?

Steven Lenaerts Head of Product Management - Global Channels, BNP Paribas

Lenaerts: Basic RPA is essentially using software to create a certain level of automation for a repetitive process that is currently not digitalised or automated. That piece of software is referred to as a ‘robot’ or ‘bot’, but, of course, it is not an actual robot – that’s a common misconception.

The robot captures and automates these repetitive actions, typically to unlock legacy infrastructure and to connect the dots between different systems. It’s important to remember, though, that RPA can be combined with other technology to increase its level of intelligence.

Comins: The important difference between RPA and traditional automation tools is that it operates extremely well across the multiple systems that a human employee would use in their day-to-day. For example, RPA offers integration with email, Excel and PowerPoint, as well as browser automation.

This means that RPA can string together all of the information and processes that a human employee would need to perform a defined task – but in a much quicker timeframe. If you get RPA right, it can be like having a digital employee.

Hill: Could you outline some of the potential use cases for RPA in treasury?

Le Blévennec: Treasurers can use RPA in many different ways – at Honeywell, we currently have 48 different processes in mind to apply RPA to. One of the most successful RPA projects we have live optimises the intra-day cash forecast of the in-house bank. And I believe that all types of cash forecasting could potentially benefit from RPA, provided that structured data is available.

To explain more about what we have done here, Honeywell has 162 bank accounts that directly impact the EUR and USD balances of the in-house bank. By leveraging RPA, we can see during the day which cash has reached a source account in a particular country and which funds will reach the in-house bank later in the day.

Séverine Le Blévennec Senior Director, EMEA Treasury, Honeywell

Because we have this early visibility over funds, we can actually invest cash before it hits the in-house bank. This means that when performing our investments, we take into account every flow that hits any of the 162 bank accounts until our investment cut-off times. In the past, as we were manually tracking the data, we could only consider 40 out of the 162 bank accounts because we physically didn’t have enough time between the transactions occurring on a source account and our investment cut-off times. We also use the bot to perform reconciliations for the in-house bank, looking at the funds that were expected to come in and the amounts that actually arrived, as well as any breaches.

We also plan to use RPA to help with compliance monitoring and ensuring data integrity in our systems. For example, we have been doing a clean-up of our databases as part of a TMS re-implementation and as part of that process we have introduced some reports that detect any anomalies in our database. If an anomaly is detected, defined rules will allow the bot to clean up the anomalies where possible. Of course, there are exceptions that require human intervention, but a significant proportion of the workload will be taken on by the bots. Automation will also enable us to increase the frequency of the data audit.

So, when deployed properly, RPA can significantly raise treasury’s productivity. Using a bot takes away the boring, repetitive work from the treasury team – and bots will work 24/7 without getting tired. As such, we schedule many of our bot processes to run throughout the night, so that when we come into the office in the morning, all the information we need is already prepared.

Ng: In the past, we’ve seen treasurers using RPA for recurring, scheduled activities, such as management reporting or cash flow forecasting. Essentially, treasurers have been using RPA to extract information from different reports or different systems – whether revenue or expense-related – and to aggregate that data. 

More and more, however, treasurers are also using RPA for transactional-related activities, such as invoice processing, where they need to extract information from invoices to facilitate filling of templates, or reporting. RPA is also being used to complete application forms that they submit to the bank.

Lenaerts: Data collection is the obvious treasury use case for RPA, especially when it comes to gaining visibility over cash and liquidity by pulling data from multiple systems. As well as collecting this data, RPA can also cleanse, structure and consolidate the data, so that it is in an easily consumable format – thereby enabling treasurers to perform analysis on that data much more swiftly.

As such, RPA is not only an efficiency and productivity tool, but also a driver of quality. Removing human interaction from the early stages of data collection and cleansing significantly improves the accuracy of the data.

Comins: One of the main aims of digital transformation within treasury is to bring together data from disparate sources. It can be challenging to automate data extraction from certain systems, especially online banking and liquidity portals. RPA is a great helper tool to get to some of those hard-to-reach data sources and deliver a full, robust picture of cash and liquidity positions to the treasurer – in real-time. Moreover, RPA enables the extracted data to be presented in a consumable format, in a spreadsheet or straight into a TMS, for instance.

Also, RPA tools are very accessible to business users – very rarely is much coding required. Many bots can follow a process a user performs and record it, making it very simple to program the bot.

Hill: And how are transaction banks and TMS vendors embracing RPA to improve their corporate offerings?

Jasmin Ng Head, Group Cash Product Management, Global Transaction Services, DBS

Ng: Within banks, the use cases for RPA are no longer limited to core transaction systems. DBS, for example, is using RPA and intelligent processing automation (IPA) in the loan servicing space when there is a re-fixing of interest rates. Typically, this used to involve a large flow of emails between operations and relationship managers (RMs). But now we have automated that flow so that a system pushes a benchmark rate to the relationship manager through a portal, which allows the RM to view the loans which need to be re-fixed and select the rate they have agreed. This agreement is then sent straight back to operations and the information automatically updates into the back-end system.

Lenaerts: At BNP Paribas, we are increasingly using RPA in various parts of the bank. For example, in cash management we deploy RPA in our daily contract and service management, thereby increasing the quality and timeliness of our client service. RPA enables us to capture client information and reflect it across all of the relevant systems in the bank – avoiding silos and re-keying.

Comins: At GTreasury, we use RPA – in combination with machine learning – to bolster our fraud scanning processes. The bot never needs to take a break, so it is monitoring flows for potentially fraudulent transactions 24/7/365. The bot flags any anomalous activity and reports on that in real-time so that a human can come in and make a decision on whether or not a transaction looks suspicious.

Interestingly, we also use RPA to assist with certain tasks of our TMS implementations. Often there is data that needs to be transferred between systems and it is a far better use of time to programme a bot to take care of that than to have our implementation people focused on such low-value tasks. Using RPA in this way means that implementation timescales can be significantly reduced and the implementation experts can focus their skills on tasks which will add much greater value to the treasurer.

Hill: Should treasurers be worried about losing their jobs to robots?

Roger Comins Director of Product Management, GTreasury

Le Blévennec: RPA is actually a great tool for raising the profile of the treasury team. Bots enable us to focus on tasks which are more valuable to the business, such as optimising our investments and our relationships with different business partners. So, for those people worried about robots taking their jobs, I would reassure them that, in practice, this is not the case at all. The bot takes away a lot of the mundane tasks and gives you a new perspective on treasury – allowing you time to step back and think more strategically.

Using robotics for the intra-day cash forecast of the in-house bank, for example, has saved about two hours’ work each day for our senior treasury analyst. But he is certainly not bored or worried about losing his job. The human element is required in many parts of what we do, and a bot cannot replace a human!

Comins: RPA is not always about removing humans from workflows but enhancing how man and machine work together. It’s a great productivity tool that can enable tasks to be performed much more quickly and painlessly, and is perfect for high-volume, repetitive processes.

Ng: Within DBS, we have embraced a philosophy called ‘Design for No Ops’. The teams working on the project were initially concerned that once the automation was in place, people would lose their jobs. In reality, however, the bots have freed up more time for humans to be able to perform more value-added tasks. Jobs are being redefined and through that process, people are learning new skills and adding more value to the business – which ultimately makes their roles more fulfilling.

Treasurers are also able to spend more time performing data analysis and pinpoint further areas for optimisation within treasury operations. There could also be added efficiencies - by looking at ways to improve cash and liquidity management, for example, or intelligent ways to digitise payments and/or collections, which could add significant value.

Hill: How can treasurers go about implementing and fully leveraging RPA?

Le Blévennec: While RPA requires extra controls, one of the main benefits of RPA is that is not invasive, and therefore quick to implement. The upfront investment is also relatively small.

The starting point is to manage your database and have a TMS in place. Then you can connect your platforms to your TMS, revamp your processes, and set up alerts and reports in the TMS. At that point, RPA can be introduced to run the reports and analyse the basic data or to make inputs from one system to another when data cannot be structured in a way that can be interfaced.

And to get the most out of RPA, treasurers would do well to review and revamp their processes ahead of automating them. Otherwise, you are simply automating an inefficient process. Furthermore, there are many other steps that need to take place to digitise treasury processes before exploring RPA.

Ng: When thinking about implementing RPA, it is important to conduct research in the market to understand the different types of solutions on offer. It is not just about looking at RPA alone, but at intelligent bots as well. In other words, combining RPA with things like artificial intelligence - this can expand the scope of automation enormously beyond structured data to unstructured data.

In addition, to perform RPA, you first need to have the correct information in your system. But what kind of information is not coming into your ERP in the most efficient manner? What about invoice information, for example? How can this be entered into the ERP in the fastest, most accurate and streamlined way? These are questions treasurers need to ask in order to make the most of RPA.

Lenaerts: While RPA is relatively easy to implement compared to many treasury technologies, some IT budget will be required – so building a sound business case will be important. I would therefore urge treasurers to pick their battles carefully. Look for easy wins for your first RPA projects.

Also, tackle processes in bite-sized chunks to enable quick gains to be realised. These functional chunks can be subsequently combined into more complex automation – so you can achieve the end-to-end benefits, but you get incremental wins along the way as well.

Hill: Is the power of RPA being overestimated? Is there too much hype about its potential within treasury?

Le Blévennec: RPA will not replace Excel and, by itself, it will not revolutionise the way treasury works. That said, RPA is one of the technologies you can use to make treasury more efficient, especially when you want to navigate between systems or sources of information.

Over time, some of the work that we currently use RPA for may be replaced by APIs. At the moment, however, APIs are still evolving, not terribly customisable, and there is a lack of standards around them – which means they are not currently ideal for the job.

Lenaerts: We all know that today’s treasurers are facing a growing remit – from investor relations to regulatory compliance. But headcount in treasury departments has rarely grown with this expanding role. RPA is therefore a very welcome, relatively low-cost tool that can enable treasurers to focus on tasks where human intelligence is valuable: analysis, interpretation, and judgement, for example.

In an average treasury function, I would estimate that 20-40% of tasks are sufficiently repetitive to be able to be undertaken by RPA – not necessarily basic RPA on its own, it could require additional intelligence, but certainly the potential is there. However, for me, RPA is a facilitator for digital transformation, not the end game.

Hill: What are common RPA pitfalls to avoid?

Lenaerts: Maintenance of the RPA workflows is critical. What happens if the bot breaks down? When infrastructure changes are made, are RPA processes re-tested and tweaked? Are you measuring the quality of the robot and its work? It is these kinds of questions that treasurers often overlook but need to consider in order to make a success of RPA.

Furthermore, to get the most value, it is important to reconceive existing processes in a digital way, otherwise RPA is just automating and digitalising a legacy process, which doesn’t necessarily make that process optimal for the digital world.

Le Blévennec: Basic RPA, as opposed to intelligent bots, needs to be rules-based and works best with robust, predictable, meticulous workflows that are based on structured data. Otherwise, you risk breaking the RPA process with unstable inputs.

Ng: RPA can only perform certain tasks or parts of tasks, not all. This can actually be a limitation because humans still need to perform the tasks that the bots cannot, or any actions that are not repetitive or structured enough – and this can lead to more exceptions in the workflow, which is inefficient, and does not result in end-to-end automation.

Internally, we engage our IT team to look at how to automate processes end-to-end so that the automation becomes meaningful, controllable, and logical. Rather than deploying RPA for the sake of it, we look at where it fits within a broader picture that will yield greater results.

Also, some areas of RPA require the computer monitor to always be active for the RPA to kick in. If the monitor goes into sleep mode, the RPA stops working. This is one pitfall to watch out for as the process will break if the monitor switches off.

Hill: What is your best advice for treasurers looking to make a success of RPA?

Le Blévennec: Don’t be afraid of RPA! It’s not really that complex to put in place. The critical thing is to have a very strong process in place in the first instance and to build RPA into that already-efficient process. Also plan for potential errors in the RPA workflow and have a well-documented, step-by-step agenda for dealing with any issues.

Comins: Alongside building robust processes before implementing RPA, and around the RPA tool, it is important to appoint someone to manage the bots and all of the automation that has been put in place. Ideally, this management would be handled by a centre of excellence so that no single person is responsible for all of treasury’s use-cases. Other than that, my best advice is honestly just to give it a go!