The financial crisis of 2007 – 2009 has spurred many companies across the world to adopt lean management to reap the benefits in terms of simultaneously improved cost efficiencies, customer satisfaction, and employee engagement on all fronts, and it is also no secret that many such initiatives have succeeded in delivering the promise in all dimensions. However, the progress on the digital front has been irregular and uneven. To take an example, let’s consider the insurance sector. A FIS study reports that 99.6 percent of insurers faced major hurdles in the front of digital innovation. This problem is compounded by the fact that at the same time, 80 percent of insurers surveyed realized that their need to meet business challenges necessitated digital products and services. This led to a boom in the “insurtech” industry with a funding of around 3.5 Billion USD in 2015 only to grow to 7.5 Billion USD in 2020 during the Covid-19 pandemic. Recent macroeconomic conditions have put pressure on companies across all sectors to improve their cost productivity and unlock fresh value from their investments on the digital front. This has necessitated the need for refinement and end to end automation of their processes and operating models. This is precisely what Intelligent Process Automation (IPA) does. This suite of technologies will be an integral part of the digital infrastructure of companies across the world. And you need not wait for the future! This technology is a long-overdue upgrade to robotic process automation (RPA) with Artificial intelligence (AI) and Machine Learning (ML). This blog covers the limitations of the current state of RPA, the components of IPA, the benefits of IPA, and two use cases of IPA in action. Read on to find more about IPA.

Robotic Process Automation

Robotic Process Automation (RPA) is a software technology that enables the development, deployment, and management of software robots or bots that can emulate human actions involving digital systems and software. RPA has been one of the most disruptive technologies in recent years with enterprises incorporating bots that have taken over and streamlined repetitive and rule-based tasks such as data entry and populating digital forms, saving billions of dollars in terms of labor hours and value all over the world.

Problems with the current state of RPA

Most RPA projects fail to meet business expectations despite some providing high returns on investments. Many companies initially achieve some success with RPA projects but over time the bot maintenance and overhead will lead to an increase in costs. Also despite the prerequisite “citizen developer” technical capabilities, 99 percent of bots require custom scripting most of which lack resiliency and require highly technical skilled programmers for maintenance and management.

Limitations of RPA bots

● RPA bots which work on their own are called unattended bots. While unattended bots are supposed to work without human supervision and interference, the reality is different from the theory. Any change in the process for which the bot is employed implies a restart or a change in the bot’s programming. Unattended bots are mostly used for back-office processes where the rules do not frequently change with time. However, if any change happens in the workflow or any new tool is added to the process, the bot has to be restarted with a change in its processing logic.

● Changes in the UI due to updates in software tools render the bot incapable of operation. To take an example, suppose that there is a change in the User interface (UI) of a web application tool in the workflow. Due to a change in the placement of text boxes and buttons in the UI, the bot can no longer identify what the new UI means and its whole programming logic breaks down. This is the most common situation in several companies which use RPA.

● RPA bots’ custom scripts are so complex that they end up having their own software development lifecycle. This makes agile development time-consuming as any tiny change in the programming logic is a testable event requiring it to undergo unit testing, integration, Quality Assurance, and User Acceptance Testing before it can be deployed. Any change in the whole digital infrastructure requires the RPA bots to undergo the entire SDLC process which can have ripple effects on other tools used in the workflow.

What is Intelligent Process Automation (IPA)?

Intelligent process automation is an emerging suite of technologies that include fundamental workflow redesign, RPA, and AI. It takes the robot out of the human knowledge worker by removing repetitive, replicable, and rule-based processes from his work. Thanks to advances in deep learning and cognitive science, IPA can mimic human actions and activities working in context with the capability to orchestrate thousands of events across the entire workflow simultaneously, and also, suggest improvements to the process.

Components of IPA

IPA consists of four other technologies than RPA at its core:

Smart Workflow

A smart workflow tool is a process-management software which tracks all the details of a process like the initiation of a process, the handoffs between machines and humans, and collects statistical data about the bottlenecks. In an IPA suite, it sits on top of the RPA layer and manages the entire end-to-end process.

Artificial Intelligence/Machine Learning and Advanced Analytics

These algorithms involve supervised and unsupervised algorithms that learn patterns in a structured dataset before making predictions on their own. Supervised algorithms detect patterns in data presented to them and consequently make predictions on new data. Unsupervised algorithms are a bit different. They do not require a dataset to learn patterns before making predictions. They detect patterns in the dataset presented to them. Advanced analytics has been used in HR groups to identify key attributes and qualities in leaders and managers to predict their behaviors, determine career paths and identify top performers for leadership succession.

Natural Language Generation

Natural language generation or NLG is a group of software engines that translate observations and data into prose by following predefined rules and patterns in language. You might have seen stories about sports games as they are happening in real-time. This is an illustration of NLG. One interesting application of NLG is piping performance data into such software engines which would then compose external and internal management reports automatically.

Cognitive Agents

Cognitive agents are a combination of ML and NLG that are a completely virtual workforce or agents. These agents can execute tasks, learn from datasets, communicate with other employees, and even make decisions based on sentiment and emotion detection. They can be used to support other employees and customers over a call or a chat such as in customer care centers.

IPA in Action

To illustrate what IPA looks like in action, take a look at the two examples below:

● Consider an employee in a bank where he has to pull data from seven disparate systems to present a report by interacting with a customer. An IPA system would employ:

○ RPA to take care of the manual clicks

○ NLG to take of text-heavy communication

○ ML and AI where there are decisions that are not strictly rule-based

○ Cognitive agents interact with the customer over a call or a chat

○ Smart Workflows to track the end-end process including handoffs between systems and humans

● To illustrate the use case of automated weekly reports, consider what a conventional RPA and what an IPA suite might do:

○ A conventional RPA bot would log into an ERP system, gather inventory data from different columns by pre-programmed rules, add the relevant columns and save the resulting report into a file which is stored in the database and emailed as an attachment to the human employee.

○ An IPA suite would first identify the purpose of the report – which would be to take an action based on inventory levels. It would then log into the ERP system, and gather data about the inventory levels (RPA), determine if those inventory levels are normally based on patterns in seasonal and historical data (ML), compose and present a report to the decision-maker (NLG), then replenish the inventory levels if they are below normal by interacting with the suppliers (Cognitive Agents) and track the details of the entire process from start to the end (smart workflow).

Benefits of Adopting an IPA Solution

The benefits of Intelligent Process Automation are perceived differently by different people depending on which industry and the role they work in. For a healthcare employee in a hospital, automation might mean managing data in a single centralized system which connects to different subsystems with no need to manually enter data into multiple systems. For a customer service managerial employee, automation might mean using technology to sort and categorize support tickets instead of manually doing so. But now, automation has come to mean using technology to make processes and life easier, faster, and more efficient. The high-value benefits of adopting an IPA solution in your digital infrastructure are mentioned below:

Make sense of unstructured data

Implementing the RPA component of IPA can certainly bring some level of automation and returns associated with it. But the fact is that RPA can only make sense of rule-based processes working with structured data. When hit with an unknown situation, it doesn’t know what to do and the process falls apart. But IPA solutions use AI and Optical Character Recognition (OCR) to make sense of unstructured data and take the appropriate action based on the context.

Learn and Improve with Experience over Time

IPA solutions use machine learning and artificial intelligence algorithms that use data to learn and improve as they get new data over time. As the IPA solution encounters new situations it adjusts the internal workings of its algorithms to get better and the efficiency improves. In contrast, most automation solutions follow a one-size-fits-all approach and do what they are told as they rely upon hard-coded rules in the programming logic. For example, a conventional appointment scheduling tool is likely to double-book meetings in some situations unless a human spots the anomaly and intervenes. An intelligent appointment scheduling tool “knows” not to do that based on existing data it has encountered before during its training.

Improve customer and employee experience

IPA solutions improve customer and employee experience simultaneously. Take the example of multi-branch financial institutions like banks and credit unions. With an IPA solution that can make sense of unstructured data like images, videos, and audio samples, employees can push data into the centralized server faster and more accurately resulting in customers getting a hassle-free and easier onboarding process. At the same time, employees working on tedious manual processes can be freed to cultivate new high-value skills like problem-solving that can’t yet be performed by machines, resulting in a satisfied and happier workforce.

● Reduces Operational Costs and Drives Revenue

Consider the situation where there are 50 employees employed for a specific process earning 50K USD per year. Assume that they can take care of 250K instances of the processes in a year. This costs the company $10 per instance. Now, suppose that the company deployed an IPA solution to perform the same process saving it USD 2.5 million in a year. The revenue also goes up as these employees can now be transferred to perform high-value activities.


Companies are now using IPA to quickly and dramatically lower costs and derive value by automating processes end-to-end. In this blog, we have seen the limitations of employing RPA alone and what IPA can do once deployed. Even this scratches the surface alone as IPA solutions get better over time and the current capabilities only hint at what is to be coming in the future.