How to verify vendor claims and manage RPA deployment
As more finance functions aim to automate repetitive tasks and free up time to help drive business growth, they now need to navigate through a growing number of robotic process automation (RPA) vendors competing for customers.
Speaking from his experience deploying RPA in the finance team, Eric Cheung, group transformation director at Tricor Group, said finance heads need to be sceptical of vendor claims in the selection of a solution provider. This step is critical to begin RPA deployment on the right foot.
Also, instead of looking only at short-term benefits, finance should focus on long-term improvements because RPA, like other technologies, requires finance to go through specific deployment procedures, which also take time to complete, he said. These procedures include identifying the necessary process to be automated, choosing a solution, and moving into the test phase, he added.
A Hong Kong-based business expansion service company, Tricor has 56 bots working alongside its human workforce in Hong Kong, Japan, and Malaysia. Using machine learning technology, the hybrid team of bots and humans extracts and processes data from bank statements and invoices for both its internal finance department and external customers. The company is also exploring the possibility of using the same technology for accounts receivable to improve its credit control.
1. Verifying vendor claims
Below are some questions that finance heads should ask about vendors’ sales pitches and actions they should take, according to Cheung.
Does the vendor have the claimed expertise? Finance functions are no strangers to vendors that claim they are experts in a certain technology.
A way for finance to verify any such claims is to conduct its own proof-of-concept, Cheung said.
By conducting a proof-of-concept, finance tests the capabilities of a vendor to see if a proposed solution can do what one intends the technology to do effectively, he said, adding that the process can also help engage stakeholders and manage users’ expectations about a technology.
In its Successful Implementation of RPA Takes Time report, PwC highlighted that companies should choose a small part of a simple process for a proof-of-concept to assess the potential of RPA. The report added that companies should treat a proof-of-concept as an opportunity to learn about a technology and a product, and it should not take many months.
In addition, vendors often claim that they have a team of experts that can help during implementation. However, that is not the same as promising to send enough of those people to implement a solution. Finance should get vendors to confirm the number of experts they will send for implementation before striking a deal, Cheung advised.
Does a vendor overpromise how fast you can achieve the gains? Finance functions should not expect to see benefits immediately if they expect to use RPA together with machine learning to fully automate a process, he noted.
In the invoice-handling scenario, it might not take much time for an RPA tool to learn to identify invoices sent via email by looking at the subject lines (which usually contain the word “invoice”), download attached invoices into a predefined folder, and create bills in an accounting software used by a finance function.
However, as companies try to automate more of the accounts payable process such as extracting information from invoices — supplier names, products or services purchased, invoice due dates, amounts due, invoice numbers, and more — they will need to deploy machine learning together with RPA.
In this case, it could take a machine learning-enabled bot six months to learn from a sufficient number of invoice samples before it can extract data correctly.
“If this is the first machine learning-enabled bot that you implement, you should expect a longer time frame [to learn how to do something],” Cheung advised. “It might take at least 50% more time than your initial estimate, as your staffers need to spend time learning how to work with it as well.”
STP? What does the jargon mean in context? RPA vendors meeting with potential customers may use jargon such as straight-through processing (STP) percentage and accuracy.
According to Cheung, STP refers to the percentage of documents that a bot — equipped with optical character recognition and machine learning — can confidently and accurately read.
For documents that a bot is confident in reading, accuracy refers to the percentage of fields that the bot can read correctly, he explained.
While it’s crucial to know what these terms mean, it’s equally important to know what it takes — time and necessary human resources — to achieve the desired STP percentage and accuracy, Cheung pointed out.
2. Managing deployment after selecting a vendor
After getting vendors to clarify their claims and deciding on a vendor and its solution, finance still has the following issues to deal with to ensure deployment success, Cheung noted.
Define deployment success. CFOs need to think about how they define success and have a clear set of key performance indicators (KPIs) to justify the investment.
The KPIs should include both short-term and long-term goals, Cheung said.
According to him, a short-term goal could be the number of steps in a process that one can reduce while a long-term goal could be determining whether a bot deployment can contribute to an existing continuous improvement programme.
Prepare sufficient high-quality data. Finance needs quality data to correctly train its bots.
To ensure data quality, Cheung said finance will have to seek tools or help from IT to remove inaccurate data or duplicates.
If any images are needed, the finance team will have to find relevant high-resolution images for bot training, he added.
Manage change. Companies need to engage both senior and junior users during pre-implementation and throughout the project.
Change management tasks include clearly communicating the vision and objectives of RPA, involving core users in designing the bots, and getting an early start in familiarising users with changes in processes, according to Cheung.
Engage management. Finance also needs to work closely with the company’s management team and communicate effectively to ensure RPA success, said Wendy Wang, group CFO and COO at Tricor Group, who works with Cheung in finance’s technology projects.
According to Wang, the RPA deployment is part of the Tricor digital transformation blueprint crafted by the management team. “We don’t make assumptions. As a team, we believe effective communication helps us survive ups and downs and work together in overcoming challenges during deployment,” she said.
Set up an automation centre of excellence (ACOE). After the initial stage of deployment, finance teams can build on RPA’s successes for the long term. An ACOE — a team or a shared facility that provides leadership, best practice, research, support, and training for a focus area — is a value-add that all business functions can benefit from.
The scope of an ACOE may differ across companies, but it can include infrastructure, a command centre of the bots, bot development and testing, process improvement, business analysis, and project management, Cheung said.
While an ACOE helps functions and units use RPA effectively, the centre can also enable other teams to contribute to the entire organisation at a later stage.
“When the use of RPA becomes mature, a company might distribute some of the ACOE activities across business units and functions including finance,” he said. “Staffers get trained to become citizen developers who develop bots and drive continuous RPA improvement for the entire organisation.”
Remember your goals
As RPA technology evolves and vendors continue to vie to provide finance teams with the latest tools, finance leaders can stay focused and succeed by keeping in mind the goals of RPA deployment.
“Finance deploys RPA because of the goals it has to achieve,” said Cheung. “Remember the goals, and they will guide you throughout the RPA journey.”
— Teresa Leung is a freelance writer based in Hong Kong. To comment on this article or to suggest an idea for another article, contact Alexis See Tho, an FM magazine associate editor, at Alexis.SeeTho@aicpa-cima.com.