9 Criteria for Processes You Need to Automate (with RPA)
Robotic Process Automation (RPA) is here to fundamentally change the way we work. With RPA, you can improve your business, run fast, efficiently, scale, while ensuring compliance requirements, and enhancing the overall quality of deliverables. It provides immediately tangible benefits that leave you to focus on higher value business goals.
RPA brings a major shift in perspective. But each shift of perspective comes with a set of new questions about making it work. For most organisations, work consists of hundreds, even thousands of processes run daily by teams which are often distributed in many locations worldwide. On top of that, work tasks are incredibly diverse, so much as to cause a true challenge for managers who need to choose the work processes ripe with potential for RPA.
How can you tell which processes you need to automate? There is a method to select appropriate processes in order to work smarter, not harder without the associated madness (read: headache).
What Processes to Automate with RPA?
As a general rule, three types of processes fall under the umbrella of appropriate choices for RPA:
- They run according to the same, repetitive patterns.
- Processing them manually take a long time to complete (individually or in aggregate).
- They are template-based
Under each of these general categories, processes that you need to automate must also suffice multiple criteria (if they optimally suffice all of them, even better). Consequently, when you discuss how to spurt new growth rates for your business by picking up RPA-ready processes, here are the criteria you need to consider:
1. Low added-value
When the human(s) is (are) not adding an inherent value to the task, company or to the customers while performing the task.
Where there is no need for a broader understanding in order to complete the task successfully.
Where creativity, empathy, extensive domain knowledge and problem-solving are not required.
2. Manual & repetitive
According to a 2017 McKinsey report on the future of work, “Activities most susceptible to automation involve […] the collection and processing of data.”
As our business world revolves more and more around data, processes around it are prime candidates for automation.
While a lot of data-related processes are already automated when the data is database-native, countless tasks and processes in businesses, still rely on humans performing “robot work”. This happens especially when the data comes in the form of PDFs (eg invoices, purchase orders, etc..), Excel spreadsheets (eg pricing updates, report generation, etc..) or CSV files.
Five percent of jobs are 100-percent automatable, according to the same resource.
It is a low number, as most jobs do not rely on doing the same thing(s) over and over again. Rather, automation is about identifying the processes that are consuming a lot of human labour time, so humans can be focused on where they can add most value, and so the company can scale.
A 15mns daily task costs 3% of the employee’s salary.
That’s 1k/year for an employee with an annual salary of 30k, not counting the indirect costs of frustration, errors, and context-switching!
Repetitive tasks may suck out the life of your dedicated workforce, but they are the essential nutrients for RPA tools!
Well-structured processes completed in a predictable order, following an agreed set of rules, or step-based processes are among the best suited to automation.
Since RPA works by giving a predefined set of orders to a machine – your computer – it must follow a logical set of steps to effectively and affordably replace the waste created by repetitive tasks. Algorithms at the core of RPA tools are precisely that – a well-thought-out, predefined, clever set of rules for automatic tasks execution.
It must be noted that the advances of AI (Artificial Intelligence) have started to blur the line on this criteria. Defining a set of rules for the robot to perform the task or process isn’t always required. RPA and AI working together is a machine that can learn (or be taught) what to do, based on past examples, rather than having to define a stiff set of rules.
4. Low exception rate
Logically, when there are rules, there should be the fewest possible exceptions for them to work out.
If you run mostly uniform, consolidated, and streamlined processes then a low-exception rate should not be a great concern. But exceptions do occur since no process is perfectly consistent. So, under such circumstances, you need to find a way, not around exceptions, but rather with them or through them.
RPA tools can be incredibly smart. For example, a Bot can be taught to manage exceptions. Yet, as each exception needs to be coded (at least when AI is not at play, and even so, AI still requires maturation), the time to solve errors that start to appear when the Bot gets to work, increases the overall time to get to a process managed accurately by a Bot.
Every time an error occurs, due to an unforeseen exception, the Bot’s code can be improved (by yourself if you are using a RPA tool or by your RPA provider) to deal with that exception appropriately in the future.
Therefore, a Bot’s inability to process an exception happens only once. All future occurrences are eliminated thanks to the Bot’s capability to learn from mistakes.
This results in the following timeline when implementing automation versus continuing to do it manually (scale will vary depending on complexity of the task/process):
In a nutshell:
if a process or a task has more exceptions than rules, automating that task or process might not be the best option.
5. Data-related and electronic input
Inputs are all ways in which a computer interacts with the environment. For example, a mouse, a keyboard, a camera, a touchscreen, or a microphone are all input methods you use almost daily.
RPA tools are computer-based. They must work by interpreting, processing, and taking action on data and electronic input.
The RPA started to boom in the last decade, thanks to the rise of software testing tools. As developers worked on ways to automate software testing, creating bots that were able to mimic a user by clicking around, scrolling and entering data, some started to think that this approach could actually be used to mimic a human actually working.
So Bots are able to connect to systems, click where needed, select options, download files, submit data and much more.
Not all input is equal though.
The most common pitfall remains for a Bot to understand information that is not provided in a structured manner. For example a PDF file (like an invoice or purchase order) can be either a scanned document or one generated by another computer.
In the latter case, the file includes the information that can be read accurately by another computer. So automating a process around such files will result in 100% accuracy.
A scanned PDF on the contrary, while being a digital file, is basically a picture. In that case, a Bot will use OCR (Optical Character Recognition) to try and decipher the data (what’s written in the PDF). As the name implies, it will look at each character, and assess to the best of its ability what that character is. But the difference between a
0 (zero) and a
o (the letter o) , or
1 (one) and
l (the letter l) can be hard to guess sometimes… and make a world of difference when handling financial data!
While OCR technologies have and still are improving (especially with machine learning), working with scanned documents will not allow a 100% accuracy. At OfficeBots, we are implementing in our PDFee Bots with the latest and greatest OCR technologies, leveraging machine learning, yet we still see some errors in how data is read from scanned PDFs.
Not all input is equal!
6. High volumes
Scalability and maintenance are two important concerns when choosing processes to be automated. RPA tools must be scalable and require very little maintenance long-term if you want them to be money-savers. If a task shows once or twice, intermittently, or doesn’t significantly impact your organisation’s work processes, it is most likely not the best contender for automation.
Automating high-volume tasks should deliver clear benefits, accelerate business processes, improve output quality, increase productivity, or augment compliance.
You shouldn’t forget about the cost-effectiveness of RPA. Truth is, there is no simple answer to the question of how to evaluate the ROI of RPA tools before you gather more details or implement them. There are simple metrics you can use to calculate ROI rates, including:
- The time/hours saved in an overall process
- Number of eliminated errors
- Hourly costs of staff resources
- Final output, for example, answered phone calls or generated leads.
But these need to be put against the costs of the automation tool, its development and implementation, the infrastructure needed to set the new process, as well as support and maintenance. Evidently, there are lots of factors to consider. However, since certain RPA tools promise up to 40 percent productivity increase in the first year only, it is worth investigating the ROI potential for your business.
8. Systemic stability
The uniformity required for task automation must spread not only in space and volume across the organisation but also in time. Processes must be mature and remain constant if you want to reap tangible benefits from automation.
Each change that happens in a system affects the laws between the factors in play. Since Bots are trained to follow and act upon a set of rules, a change instigates a new variable which will need setting up new rules to account for the new variable. If systemic changes happen often, the process might not be ripe for automation.
9. Process maturity
RPA tools work best when there is some maturity and stability in an already set up process. The process in place needs to have proved its efficiency.
“The first rule of any technology used in a business is that automation applied to an efficient operation will magnify the efficiency. The second is that automation applied to an inefficient operation will magnify the inefficiency.”Bill Gates
Completely new tasks or processes that have just started may not be the best candidates to automate.
For example, if you work with emerging technologies and perpetually changing team objectives, automation tools just may not be able to completely replace human effort. But they can optimise repetitive, carefully chosen tasks, that add up (“every little helps”, to quote a well-known UK supermarket chain).
If you want to read more about RPA and the “how-to” of implementing it in your business, I highly recommend all the literature from Dr. Mary C. Lacity & Professor Leslie Willcocks who co-authored a couple of books and have become authoritative thought leaders in the space, based on extensive research and focus on that field.
Is it time for you to invest in bots and take that nagging, productivity-reducing work off your back?
If you are interested in identifying the tasks and processes that are ready to be automated in your business, I am available with the OfficeBots team to answer your RPA questions, discuss your current workflows and potential for automation.
You can book a time here: https://officebots.io/booktime