Back in 2014-2015, I was working on a release management use-case. The project spanned over eight months with 2 weeks sprint cycles, with functionalities like sending emails and extracting information from the service management systems occupying sprint lives of its own.
Those were the days when service industries wanted to keep challenging the possibility and establish that automation can be done without the help of the tailored-made automation tools.
Needless to say, there were hurdles in this approach, specifically for applications where APIs were not exposed or native integrations were not present.
The GUI based automation was dependent on control IDs of the elements which were difficult to find, and even more exhausting to develop when trying to deal with dynamic windows.
The other alternative was to use libraries that aid Image-based automation, which was relatively easy to develop (IMAGINE THAT!!) but would have failed even at the slightest change of resolution, screen, fonts, etc. and moreover the automation might have run at a speed slower than the human counterpart.
While I was happy simply by seeing the robots work on the machines ( Yes, I’m that excited kid of the lot!), flaunting the bots on every client visit- working on the SAPs, interacting with service now, generating reports- no matter how slow the execution was or how long it took to deployment,
I was completely oblivious of the fact that business is demystified and have seen robots work faster, with more precision and with far lesser time to deployment.
With such frustratingly slow progress, stern business feedbacks, and piling up of hurdles in every retrospective meeting the program finally had to give in to the use of the Automation Tools.
With Automation tools like UiPath, came a wider set of possibilities. Activities like sending of mail and information gathering from service management systems took just a day or two to develop instead of sprints – it was a refreshing change.
The image-based automation was more accurate and resilient, the Non-intrusive native GUI automation was an equal treat to develop and watch (it was FAST). The pre-built activity set was huge, there was a central management system (orchestrator) to manage the bots and workload, and with the minimal learning curve, it was easy to start developing workflows in a short span of time.
This was RPA establishing itself as a tool where you can design the repetitive rule-based tasks as workflows and your bots will keep performing the same as per your settings and requirements. While this was good the industry soon realized that it was not enough.
There are a plethora of use cases that do not fit the RPA criteria point by point but is still something that is repetitive and something that the industry would like the bots to take over.
For e.g. the invoice processing, where the inputs consists of invoices from hundreds of vendors with no fixed format, or chatbots interacting with humans to take relevant inputs or actions that may change based on the needs of the end user, and yet a bot can tend to the request and close it.
There was this wonderful insurance-based use case, where UiPath working with Google’s vehicle damage assessment system and Kore.ai chatbot was able to provide end to end claim resolution with the UiPath robot using the chat agent as its input source post conversation with the user, and leveraging Machine Learning to assess the percentage of damage and then releasing the report and the claim amount taking the decision based on that. The RPA became intelligent!
The Anatomy of RPA
While we are still on the curve of inducing intelligence in automation, learning on various aspects (patterns of receipts, invoices, legal documents, objective specific mails, decision making based on values of various input parameters etc.), where does the evolution of RPA stands today? As I see it, the anatomy of the RPA can broadly be categorised in two parts:
1. Action elements: These are the action items of the workflow (connect to SAP, if connection failed send outlook mail, click on a button, type into a textbox etc.)
2. Intelligent elements: These elements help perform activities like understanding the context of a text (whether it is a spam mail), extract data based on patterns (Invoice data extraction, damage assessments etc), decision making (fraud detection based on mail content), etc.
Phases of RPA Evolution
1. The first phase of RPA was the monolithic development of the action elements when complete automation workflows were being developed natively with the same active elements being reintroduced as a separate module for every individual project.
2. The second phase was when these action items got categorized and activity sets like outlook, excel, salesforce, snow, etc were being provided OOTB by the automation tools to reduce the time to develop and deploy the solutions.
3. The third phase-shifted the paradigm from pure action-based RPA to RPA + ML/AI. The intelligent automation. This is where we are. The intelligent components help introduce the understanding, pattern recognition and extractions, and also decision making to the existing action-based automation workflows.
The intelligent elements are right now constricting in nature, they have their own protocols, own way of providing the output. Thus, to use an AI/ML proprietary model, you need to adhere to there requirements and modify your workflow if switching from one model to the other. That’s where UiPath’s AI Fabric comes into the picture.
4. The fourth phase will see the rise in RPA solutions setting up contracts and adapters to make it possible to use any AI/ML (even your in-house models) and place it in your solution without going through the trouble of modifying the solutions to adapt to the Model’s requirement.
This will make the intelligent elements loosely coupled with the solution and hence will provide greater flexibility and choice while developing intelligent solutions.
The peak of the automation first era will see the robots learning from the users, and from their own past executions. The robots will expand their capabilities knowing when to switch among intelligent elements and will be able to run workflows dynamically making the solutions more tailored to each context.
With the open architecture, integrations with industry-leading solutions on various aspects, and the sheer motivation to make robots out of the humans, RPA has evolved tremendously in a very short span of time and has become one of the driving factors of the fourth industrial revolution.
Disclaimer: The views, information, or opinions expressed in the article are solely those of the author and do not necessarily represent those of UiPath and its employees.