Automation – Problem & Solution Continuum: A journey towards Hyperautomation

As COVID and other disturbances in the market are pushing organizations to evolve and adapt, more and more IT and Business leaders are looking to increase automation and agility of their processes. But you face a multitude of technology and tools, which provide some of these capabilities. They range from Robotic Desktop Automation (a spin-off term from RPA) to Robotic Process Automation, to more established technologies like BPM and ERP. The features and capabilities of these tools are evolving, and you see a liberal dose of AI/ML/NLP capabilities being integrated into them. In this scenario can you use a framework\construct to assess and compare these tools? What we see is that Market Analysts, Service Providers, and Tool Vendors take technology and solution-driven view to answer these questions.

The typical answer is that RDA/RPA can be used by organizations at the early stages of adoption and as you mature you graduate to incorporate ML, AI, and analytics capabilities. I agree with this view. But at the same time, this construct misses the nuances that can be understood by taking a business process and problem space viewpoint. You can reframe the question. You may ask, based on the business process that I want to automate or customer's pain point, which of these technologies will provide an optimum solution. This probing will bring the below factors into your view,

1.    Process Variation & Scope: What is the variation in the process that is expected? Are we talking about a stable process or one which has many variations? What is the process scope that I am looking at? This refers to whether you want to automate a few activities and steps in a process, or you are looking at end-to-end process automation, which may even cut across the function. E.g.: Do I want to automate interview scheduling activity for potential candidates, or I want to automate my end-to-end Talent Acquisition process?

2.    Cost Savings: Am I looking for large cost savings to meet my corporate objectives? Or more modest savings within a team or department?

3.    Investment: Is this part of an incremental improvement initiative or does the organization wants to invest and transform a particular function entirely?

4.    Local vs Global: Do I want to encourage grassroots level innovation and empower employees to resolve their immediate pains? Or Do I want to standardize the process globally in a top-down manner?

5.    Regulations & Controls: Do I need to ensure control and compliance requirements in place? If so, do I need to comply with regulations across countries? Or there is minimal regulatory control for the process under scope?

Answers to the above factors will influence which of the technologies will best suit your needs. The answers may not be binary but asking the above questions will provide greater clarity on the organization's objectives and goals. When we look at the technology choices, based on the capabilities and typical width and scope they cover I have organized them along a spectrum (as shown below).

No-Code\Low-Code form image credit: https://extensions.joomla.org/extension/powr-form-builder/

RDA

Coming to the tools themselves, it is essential to have an understanding of their fundamental capabilities and limitations. Robotic Desktop Automation or other Desktop based automation tools, including the versatile Excel Macros, can automate repetitive but small steps in a business process. They are also very difficult to scale and control. But that is changing with Microsoft bringing PowerApps Flow, its own IFTTT tool. With Flow, an Operations Analyst can create an automation routine, which can listen to a mailbox for daily performance metrics for an e-commerce site. Store the log file, invoke Excel to import the data, generate graphs using macros, refresh graphs, and send this Excel report to wider team before a daily review meeting. The Operation analyst will need a few hours of training to go ahead and carry out this automation. More and more your desktop tools can talk to each other, tools like Flow can act as a glue or a desktop level activity orchestrator. The tech-savvy Operation Analyst can also share their automation routine to a wider team through the Office 365 cloud.

RPA:

RPA tools can automate processes with little to no change. They can replicate typical human interactions with software applications. They achieve this by interfacing with the Graphical User Interface and can enter data into application screens or extract data through screen scraping. Thus, there is minimal to no change to the process and systems involved. But this flexibility and agility result in some compromise. The GUI changes can break RPA automation and impact reliability. But when you encounter use cases where the user inputs are required at the start or middle of the process, such disconnected and disjointed scenarios may require splitting the automation solution or other means to provide this input. In this scenario, the No-Code platform can be used to gather the initial input. Whereas any intermediate manual interventions, human judgments, and approvals can be handled by integrating RPA with BPM.

No-Code:

There are also processes in an organization which is largely email-driven. This process may not have been onboarded to a case\service request management tool. Let us say you are automating an expense approval process. Today the employees send an email with expense details, cost center\project, and few other details to a shared mailbox. The shared service monitors this mailbox and routes this Finance and other approvals. This poses a challenge for RPA automation. You can take this process without any change and automate it. But other than mailbox monitoring there are no other means to identify when a request is triggered. Let us say you want to capture a basic set of data for all expense reports. With an email input, this standardization can't be achieved in an elegant manner. A simple web-form submission can help here. A no-code platform can be an ideal fit in this scenario. The No-Code\Low-Code platform can provide a means to collect user inputs and also to trigger a request. An RPA Bot at the backend can fulfill this request by running on a schedule.

BPM:

BPM can also play a role when you want traceability of where a particular transaction is within a process. e.g.: In the above employee expense approval process, the user will want to know where his request is and who has to approve next. BPM also scores by providing a better audit trail and built-in dashboards to keep a tab on the requests.

AI & ML:

In both of the above scenarios, and ML/AI component can be introduced if there is a need to process unstructured data-intake. Let us say you want to extract the expense amount from scanned receipts. It can also be leveraged to learn and automate human judgment and decision-making activities. Provided such activities don't cross the red lines an organization may have w.r.t regulatory, data privacy, and fairness guidelines, etc.

Enterprise Packages:

As you can see the above technologies are quite complementary and can be fitted like Lego blocks as needed. At the same time as you add more blocks the overall system complexity, time to market and cost will increase. You will soon reach a point where specialized packages like ERP, CRM, Supply Chain Management, will make more sense. As the process scope increases ERP and enterprise packages offer many advantages - they come with domain-specific features, incorporate best practices, and can standardize the process.

Niche tools:

In addition to the above, may niche tools that can automate a specific activity are also emerging. Frequently, these tools leverage AI & ML. As an E-commerce Operations Head, you may bump into tools for customer sentiment analysis based on customer feedback in Google Maps and other platforms. If these tools leverage proprietary AI\ML algorithms, they are typically hosted in the cloud and offers in a SaaS model. If you would like to leverage these tools as part of your E-commerce value stream, it is essential to integrate them. This integration can be achieved by an RPA solution and maybe a BPM solution as well, based on API availability and specific use cases.

When you combine the dimension from the problem space, with the capabilities of the tools and technologies, you can slot the tools based on your problem space. A small task that is used by a small team can be handled by a citizen developer using desktop automation tools like Flow along with Excel Macros etc. Whereas a little more complex process which requires a moderate throughput and variation will need an RPA solution. If end-to-end process automation is required for the purpose of tracking each and every transaction or if there is a need for approvals, then BPM is a better fit. Based on whether learning is required, or unstructured input data is involved, then ML\NLP can be incorporated. If the organization is looking for a global solution for an entire functional area like Talent management, Supply Chain Management, etc., then an enterprise package implementation will be required. I will be happy to hear your thoughts, learnings, and questions through comments or notes.