• Kustaa Kivelä

Five pitfalls of scaling up your RPA pipeline

Updated: Sep 24

An old saying says that "everything that can be automated will be automated" - but there are some real bottlenecks preventing it:

  1. Traditional door-to-door automation assessment

  2. Lack of business ownership and time

  3. Low-hanging fruits delusion

  4. Uneven quality of assessment

  5. Technology-first approach


Traditional door-to-door automation assessment


Finding automation potential, business cases and creating automation backlog is one of the slowest things to do - if it is done by humans. Large enterprises have hundreds of teams who would need an RPA assessment to be done. A solid assessment typically takes at least a couple of weeks so your company-wide exercise takes several months to several years to get through with a door-to-door method. When your assessment workforce is done with their job, you will end up with outdated information about the situation. One option is to level this up to continuous improvement and train your people’s automation mindset, but this also creates issues with uneven assessment quality not even talking about how slow this way is. The first step would be to create your assessment processes so that you do not need door-to-door visits at all. Your assessment design should be data-oriented and remote first in order to be successful and scalable and Covid-19 sufficient.



Lack of business ownership and time


Quite often RPA programs start within a single team, function or unit. They might implement robots for years without expanding to other units. One function is performing really well but the company is losing productivity as other available potential is not addressed. There are multiple things which might cause problems while expanding your RPA program to other functional areas:


1. CxO level sponsorship


Enterprise level automation programs do not succeed without top level sponsorship. Automation and AI does not know process, unit or team boundaries so your program needs support from above. It must be a cross-functional exercise.


2. Business owner’s time


One of the biggest challenges is to get time - especially from business owners, who are already busy. They would need to sponsor, communicate and participate in finding use cases, designing and implementing the new way of working. Whichever assessment model you will choose, please make sure that design does require only a minimal amount of business owner’s time until you can showcase the business case regarding automation to the business owner.


3. Cost sharing and centralization


When automation and AI does not know company silos and boundaries, typical problems include figuring out who does benefit from automation and who will pay the bill. This is a real common problem with RPA programs as visibility of actuals is not available and benefits are typically shared. Companies should place data-driven measurement for automation, business case and cost allocation based on benefits.



Low-hanging fruits delusion


Companies are focusing on low-hanging fruits, big well-known chunks of manual tasks. Who wouldn’t know some real manual burden in the company that needs to be done! But no one likes to do a really cumbersome bulk process like claim handling, mass data transfers etc. Finding these with traditional automation assessment and use case heatmaps is not hard, but it’s still laborious i.e., costly, at least on a bigger scale. Delusion comes as people happen to think that these low-hanging fruits are the biggest gains that can be automated and bypass alternative approaches thus opportunities.


The question is - are you picking low-hanging fruits or are you actually missing a huge amount of strawberries?


If companies limit their thinking on unit, country, function or team level view, they might not have any idea that similar activities are actually done in a modular structure elsewhere. Finding the same automatable activity from hundreds of employees in a modular structure is more valuable than a single low-hanging fruit because you can solve it once with one solution that fits all. It's a question of if your company has executed both horizontal and vertical automation assessments. Without company wide automation potential mining platform, you end up with a handful of fruits and look at the world only from a process angle.



Uneven quality of assessment


One of the hardest things to succeed with in RPA automation pipeline generation is managing the quality of potential cases. From what I have learned during my years in consulting, it’s really hard for people to understand what is automatable and what is not. And, still most automation programs depend on training employees to identify automation potential for technologies such as RPA, ML and OCR. It’s not a surprise that quality varies already starting from the selected coach and a broken phone phenomena does not help it.



Also, it’s about the quality of data used for business case creation. If you ask during the assessment questions about volumes, transactions, time spent etc. getting even a ballpark estimate is really hard. Typically estimates are really off from the actuals. Quality comes from a defined assessment process and actual data.



Technology-first approach


Technology industry is well known for creating acronyms for all the possible things, e.g. AI, IA, IPA, RPA, ML and OCR. For all of these you can find assessment frameworks on how to spot possibilities - and typically assessments are done over and over again with a new technology.


We are also extremely enthusiastic about new possibilities and the way those work. The hard part is not finding a technology to solve the problem. The hard part is to find and identify a problem big or modular enough that it justifies the use of those technologies. Another good saying is that back in the old days problems existed before solutions. So if you start with defining problems and hypotheses such as manual and repetitive work is killing your productivity, then you can focus on finding all manual activities and work in your company, and eventually, address suitable technology to solve that problem.



Time for data-driven case and technology selection

In the end, it’s about the maturity stage of technologies. Right now, everything seems expensive and complex - even technologies to solve minor problems - but tomorrow we might be ready. If companies want to automate everything that is automatable, it would require them to understand all manual activities, flows and patterns done within the company from both process and work angles. The question is how do you identify your organization's all automatable tasks immediately and automatically? For that we are going to respond with Workfellow.ai.



Read more how Workfellow can help automation and RPA initiatives: https://www.workfellow.ai/automation-rpa


Pictures by Icons 8

Follow us for the latest insights:

©2020 by Workfellow.ai