[Robotic Process Automation Design & Development] [21CS744

]
`

ROBOTIC PROCESS
AUTOMATION
DESIGN &
DEVELOPMENT
(21CS744)

MODULE 5
Exception Handling, Debugging, and
Logging
Referred Textbook: Alok Mani Tripathi, Learning Robotic Process Automation,
Publisher: Packt Publishing Release Date: March 2018 ISBN: 9787788470940
&
Textbook 1: Chapter 13 (Future of RPA)

[Dept. of CSE, AJIET] Page 1

 [Robotic Process Automation Design & Development] [21CS744]
`

MODULE 5
Exception Handling, Debugging, and Logging
Exception handling
Exception handling is a way to handle exceptions for a process that the program or the
procedure has failed to execute.
For handling exceptions in a program, the best practice considered is to use the Try catch
activity.
The Try catch activity can be found in the Activities panel.
By dragging and dropping the Try catch activity into the workspace, we can handle
exceptions. For handling errors in the Try catch block, we can divide the whole process into
four parts just to make it simpler:
 Drag and drop the Try catch activity
 Try block
 Catch block
 Finally block
1. Drag and drop the Try catch activity: Create a blank project. Drag and drop the
Flowchart activity into the Designer panel. Search for the Try catch activity in the
Activities panel and drag it into the Flowchart. Set it as the Start node:

[Dept. of CSE, AJIET] Page 2

 [Robotic Process Automation Design & Development] [21CS744]
`

2. Try: When we double-click on the Try catch activity, dragged and dropped inside the
workspace, space for the Try activity appears, as shown in the following screenshot:

Inside the Try block, we have to drop the activity we want to perform. Drop a Write
line activity to test the working of Try Catch block, as shown in the following
screenshot:

[Dept. of CSE, AJIET] Page 3

 [Robotic Process Automation Design & Development] [21CS744]
`

3. Catches: Inside the Catches activity, first we have to click on Add new Catch and
then click on Add Exception option, from which we have to select the type of
exception. In most cases, [Link] is preferred. The following screenshot
shows the types of exception. There are many more exceptions which can be viewed
by clicking on the Browse for Types option:

Say the execution fails: for example, the Click activity is unable to be executed
because of the unavailability of a UI element. In such a case, we can use the Catches
block in order to either view the error that has occurred or for an alternative method to
be used if that particular error occurs. As shown in the following screenshot, we will
drop the activity in the Catches block. To print a message, we use a Message box:
When we click on Add new catch, we are asked to select the type of exception. We
have selected [Link]. Now inside the exception block, we have dropped a
Message box activity. Entering [Link] will display the error that
occurred during execution.
4. Finally: When we have defined the exception for our sequence, the Finally block will
always work, regardless of whether the execution was successful or not. Suppose we want
to display a message to the user notifying that the process is complete. To make sure that
the whole Try catch activity is executed, we will just drop a Message box activity in the
area provided in the Finally block, as shown in the following screenshot:

[Dept. of CSE, AJIET] Page 4

 [Robotic Process Automation Design & Development] [21CS744]
`

Common exceptions and ways to handle them

Unavailability of UI element:

When working on UiPath, especially on the web, we may encounter this type of error.
This is because the UI element was not found due to the dynamic behavior of the web page.
To handle this exception, we have to make changes in the selector attributes or we have to
add new attributes to the selector so that the UI element can be easily found.

For example, if we have a variable which is dynamically changing, we can use a

wildcard so that it can be easily found by the robot. We can edit the dynamic attribute of the
selector using wildcard characters. Another way is to attach it to the live element.

Handling runtime exceptions:

We may encounter runtime errors while working in UiPath. To rectify these errors,
one of the best practices is to use the Try catch activity, which can be used to handle

[Dept. of CSE, AJIET] Page 5

 [Robotic Process Automation Design & Development] [21CS744]
`

exceptions at runtime. By keeping an alternative inside the catch block, we can also
overcome the error which we encountered before.
So placing your sequence or workflow inside the Try catch activity will help you
handle runtime exceptions.

Orbit reference not set to the instant of an object:

This type of error usually occurs when the default value required for some variable is
not provided. In that case, we are required to give a default value to the required variable, as
shown in the following screenshot. In the empty area indicated in the following screenshot,
just type the default value of the variable in order to overcome this error:

Index was outside the bounds of an array. Index out of the range

This error occurs when we try to iterate array elements by an index which is out of
range. This happens when we are not aware of the size of the array and we just randomly type
the index to access the element. To resolve this, we must check the size of the indexes of the
array or the collective list.
Image not found in the provided timeout:

This type of exception is thrown because the image was not found. This may be due
to a change of environment, such as resolution or theme settings. In this case, using some a
selector attribute or indicating an anchor will work well. Indicate Anchor will help us
indicate the UI element nearby so that the recorder can identify the correct image.

[Dept. of CSE, AJIET] Page 6

 [Robotic Process Automation Design & Development] [21CS744]
`

Click Generic error - cannot use UI CONTROL API on this UI node please
use UI Hardware ELEMENTS method

This type of error occurs when the environment in which we are trying to use the Click
activity does not support Simulate or Send message activity (used by us to click the UI
element). Sometimes, either SimulateClick or the SendWindowMessages may be checked.
In both cases, when an exception is thrown we just have to uncheck the appropriate box.

Logging and taking screenshots:

UiPath has a multi-process architecture that offers to execute each workflow

separately in the executor. Executors are managed by UI robots. So, if any executor stops
working, then the entire process will not be affected.

Client logging

Client logs basically enable a server to record connections. These logs can be used by
content providers in various scenarios, such as to generate billing, to trace media server
usage, or to deliver suitable quality content depending on the speed of the client's server.

For client logging in UiPath, we have an NLog configuration file which makes it easy
and flexible to integrate with databases, servers, or any other NLog targets. Logging can be
configured with this [Link] file.

UiPath Studio, Robot, and workflow execution generate log messages on the client side:

 Messages which are produced by the workflow execution are logged with the
execution logging source. Messages produced by UiPath Studio are logged as Studio
Source and those produced by UiPath Robot are logged as Robot logging Source.
 We can also access these logs from UiPath Studio.

We can access the stored logs by clicking on Open Logs in the EXECUTE option.

By default, these Logs are saved in %Local App %\Uipath\Logs :

 The automatic logging mechanism for all errors generated, including values of variables
and arguments, may be enabled in the [Link] config file, which is present in C:

[Dept. of CSE, AJIET] Page 7

 [Robotic Process Automation Design & Development] [21CS744]
`

\Users\Username\AppData\Local\UiPath\app-17.1.6435, by setting the Log

parameter from 0 to 1 inside the <Switches> section.
 We have two activities that can be used for logging and these are the Log message
and WriteLine activities.

Server logging

If you have configured the UiPath server, then all logs generated by the execution are also
sent to the server. You can take a screenshot anytime by pressing Ctrl + PrtScrn.

Debugging techniques:
There are various techniques provided by UiPath Studio for debugging in order to check
whether the workflow is running successfully or to find out errors in order to rectify them. At
the top of the UiPath window, we can see various available methods of debugging inside the
EXECUTE block, as shown in the following screenshot:

[Dept. of CSE, AJIET] Page 8

 [Robotic Process Automation Design & Development] [21CS744]
`

As shown in the preceding screenshot, there are various techniques for debugging. They are:

 Setting breakpoints
 Slow step
 Highlighting
 Break

Setting breakpoints:

While debugging a workflow, we can set breakpoints in between if we want to run the
program up to a specific location. This is useful when we have to stop before an activity ends
completely. In such a case, we should use a breakpoint on the previous activity, as shown in
the following screenshot:

The highlighted region indicates the breakpoint since the execution stops just after the
breakpoint. In order to continue any further, we have to click on the Continue button on the
top corner indicated by the arrow:

[Dept. of CSE, AJIET] Page 9

When we click on Step into, the relevant part will start to execute; after we click on Step
over, execution will jump to the next part, and so on

Slow step:

This is an activity in the EXECUTE block through which we can reduce the execution speed
of a particular process or activity. This way, we can identify each and every process and keep
an eye on where to find the error. In the Output panel, all activities or steps can be viewed.
The following screenshot shows how to use the Slow step activity:
As indicated by the arrow in the preceding screenshot, when we click on Slow step the
execution time for this particular step increases.

Highlighting:

Highlighting is used to highlight the steps we have taken during automation and to identify
each and every step in the workflow. It is very useful while debugging and its panel can be
found in the Options menu of the Execute section in the Ribbon:

Break:

The Break activity is used to break a process at a certain point. Suppose we have a sequence
performing seven activities together and we want to break the execution at a certain activity.
For this, we can use the Break activity, as shown in the following screenshot:
 11

While debugging, an option for Break is available (indicated by the arrow in the preceding
screenshot). We can break at any point we want to. If we want to continue any further, we
just have to click on Continue, as indicated by the arrow in the following screenshot:

Or we can stop the execution at that point by clicking on the Stop option.
Collecting crash dumps
Collecting crash dumps basically refers to collecting information when your UiPath Studio
crashes. We can enable and disable crash dumps. These dumps provide us with information
regarding the UiPath crash.

Memory dumps are of two types:- full dumps and minidumps. Full dumps provide us with
complete information about the encountered crash while minidumps provide us with just the
main information regarding the crash.

When a crash is encountered, we first have to identify the process which has crashed.
Usually, a dialog will appear on the screen indicating the nature of the crash and the
application involved. A UiPath process could crash, such as [Link], [Link], or
[Link], or the target application you want to automate may crash.

Enabling crash dumps:

The following are the steps to enable crash dumps:

1. To enable crash dumps, we first have to download the [Link] file for full
dumps from [Link]
t=151332 6308120

or the [Link] file from

[Link]
26308120

2. Double-click the file and click Yes. Administrator rights are needed to access the
registry settings

3. The dumps folder is %TEMP% whose complete path is like

C:\\users2;username\AppData\Local\TEMP

4. When the application crashes, you will find the .dmp file in the TEMP folder. For
example, if UiExplorer crashes then a file such as [Link] will be
found in the TEMP folder
Disabling crash dumps:

To disable crash dumps, perform the following steps:

1. Download the [Link] file from

[Link]
120

2. Double-click the file and click Yes to disable crash dumps, Administration rights are
needed for this action.

Error reporting:

A user may encounter an error in UiPath and want to report it.

Enterprise Edition customers
If you are an Enterprise customer, then you can report the error to the UiPath community in a
very simple way:
1. Just go to this link: [Link]
2. You will be re-directed to a page where you have to fill out a simple form containing some
basic details and then upload the file in which the error is encountered, as shown in the
following screenshot.
3. After uploading, just click on the Submit button. UiPath will respond to you with the
proper solution:
Community Edition users
Since the Community Edition is free, UiPath does not provide support to Community Edition
users. However, all solutions to errors encountered can be found by going to the UiPath
forum. All types of errors and their solutions are properly discussed in the forum. You can
also go to the resources page and find the solution to your problem. For this, visit
[Link]

FUTURE OF RPA

1. A Look at the Next Few Years

Back in 1965, when computers were still in the formative stages of development, Gordon
Moore came up with “Moore’s Law” that postulated that the number of transistors on a chip
would double about every two years. While it was somewhat off – the timing is usually about
18 months or so – it was still revolutionary. Moore’s Law showed the incredible
transformative power of computers. Of course, Moore would go on to start one of the world’s
most iconic companies, Intel. But there was something else about his law – that is, it is one of
the few things in the technology world that has been highly predictable. The pace of change
and innovation can be mind-boggling. Keep in mind that many of the top tech founders and
visionaries have been wide off the mark on their predictions. For example, when Apple
launched the iPhone in 2007, Microsoft CEO Steve Ballmer had this to say: “There’s no
chance that the iPhone is going to get any significant market share.” Or how about the time
when Digital Equipment Corp. president Ken Olsen said in 1977: “There is no reason anyone
would want a computer in their home.” So yes, making predictions about the future is a
hazardous activity! But then again, it is important to try. There needs to be ongoing debate to
help fuel innovation. In this chapter, we’ll take out the crystal ball and make some predictions
about RPA.

2. Consolidation and IPOs

“RPA vendors need to diversify beyond their own niche technologies and core products to
grow their position in the industry,” said Dr. Gero Decker, who is the CEO of Signavio.
“Specifically, vendors will need to better examine the use cases and capabilities that
automation can do, therefore expanding the reach and capabilities behind their automation.
They also need to make it easier to implement automation, and automate capabilities beyond
older, redundant tasks.” There will also likely be some IPOs. Currently, the only pure-play
public company is Blue Prism. But in the next few years, we will likely see Automation
Anywhere and UiPath pull off IPOs. But being public provides other benefits, including the
following:
1. Prestige: Not many companies have what it takes to be public. Often, there must be high
levels of revenues, a diverse customer base, an extensive product line, and a path to
profitability. Being public also raises a company’s visibility with the media and analysts.
2. Liquidity: Employees may have taken lower salaries in exchange for equity in a company.
Because of this, they eventually want to see a return on this – which is much easier when a
company is publicly traded.
3. Transparency: With many RPA companies private, it is difficult to gauge the industry.
What really are the revenues? Is there customer concentration? Any issues with the core
infrastructure? For the most part, a public company is legally required to provide full
disclosure of material information. This not only makes it easier to evaluate a company’s
prospects but builds confidence with larger customers.

3. Microsoft
Historically, Microsoft has been a follower when it comes to new technologies. This was the
case, for example, with the GUI-based operating system as well as the Internet browser.
Microsoft has enormous advantages: seemingly unlimited capital, a trusted global brand, a
strong ecosystem of partners, and a massive customer base. So what about RPA? Yes, it’s
now becoming a priority. At the Ignite conference in November 2019, Microsoft announced
its official foray into the market. The company leveraged its Power Platform – which allows
for BI, low code, and workflow management – to create Power Automate. It also uses
Selenium, which is an open source application that allows for the recording and
automating of web applications. It comes with more than 275 prebuilt connectors to apps
and services. There are also AI capabilities and of course native integrations with Office 365,
Dynamics 365, and Azure. Interestingly enough, when it comes to creating alerts and
triggers, you can seamlessly use Microsoft Teams. Power Automate is definitely a solid
offering. However, it still lacks enterprise-grade features, such as identity management. But
such things will likely be added in the next couple years. If anything, Power Automate will
allow Microsoft to learn and iterate on RPA systems. And given the company’s deep
reservoir of talent and financial resources, it seems like a good bet that the progress will be
rapid .

4. Attended Automation
RPA technology has been mostly about unattended automation. And this should not be
surprising. This type of automation is easier to handle. Hey, there is no involvement with
people! But attended automation has much potential for making a huge difference.
Attended automation should lead to much synergy – the best of humans and machines. The
good news is that the technology is starting to get better, especially with improved integration
and AI. Attended bots are programmed to provide employees with the guidance and
assistance they require, in real time directly from their desktops, and automate anything
which is repetitive and doesn’t require their special skills. Enterprises are realizing that
attended automation will have a positive impact on their employee’s adoption of automation
and ultimately on the service they provide to their customers.” To get a sense of this, consider
Automation Anywhere, which has acquired Klevops, a start-up based in Paris. According to
the deal’s press release: “Automation Anywhere fast forwards the RPA category to Attended
Automation 2.0, where managers can easily orchestrate workstreams across a team of
employees and bots, driving a higher level of employee productivity and improved customer
experience. This enables customers to automate more processes than ever before, with the
same level of central governance, security and analytic capability for which Automation
Anywhere has always been known. This approach is likely to see much traction in industries
that rely heavily on contact centers, such as banks and telecom companies.

5. The RPA Career

Even though the emergence of the Internet has resulted in the reduction of certain jobs,
there have been new professions created like digital marketing.
Something similar is happening with RPA. “Hiring for RPA skills will explode across all
industries and job functions,” said Prince Kohli, who is the CTO of Automation Anywhere.
This will not just be for developers, business analysts, and managers. There will also be
growth for those who have experience in verticals, such as IT, BPO, HR, education, and
insurance.
But for those companies that implement RPA, there will be a need on how to transition
existing workers. This is certainly an opportunity – but will not be easy. Companies must
rethink the traditional approaches to talent management.
 “We will see more companies embracing training opportunities to upskill their employees
for RPA jobs, opening the doors for new career opportunities and growth,” said Barry
Cooper, who is the enterprise group president at NICE.
To do this, companies must go beyond just offering training courses. There will need to be
a focus on areas like providing interesting opportunities and incentives.

6. Scaling RPA
While the early stages of an RPA implementation are generally successful, it usually gets
more difficult to scale the technology. To help improve things, RPA vendors are retooling
their systems to help achieve scale, such as by adding AI and process mining.
“RPA began as a stand-alone technology but the future of RPA is clearly integration into
larger end-to-end process automation platforms and programs,” said Michael Beckely, who is
the CTO of Appian. “When unified with human workflow, BPM, API integrations, Artificial
Intelligence/Machine Learning services, and RPA can be efficiently orchestrated with
humans and other systems to maximum benefit. The so-called ‘problems’ with RPA really
become features when viewed within the greater context of an overall Automation project.”
He notes that the financial services industry – which was an early adopter of RPA – is already
shifting toward a holistic approach. “Whereas RPA was sometimes started as a way to bypass
overburdened IT departments, as RPA projects scale, IT is increasingly taking charge,” said
Beckely. His company has built a system, called the Robotic Workforce Manager, that uses
low code to enable the scaling of RPA projects to hundreds and even thousands of bots.

7. Vertical-Specific Companies
In general, most of the RPA vendors have solutions that span multiple industries. Yet the
companies still usually have specialties. Actually, as RPA starts to mature, there will likely
be more vertical-based players. Customers will want a more specialized approach, as there
will usually be unique use cases. Just look at Olive, which is targeting the healthcare
industry.
The name Olive is the “digital person” that provides the automation. She is quick and does
her work in a confidential manner, helping with tasks like collections, coding, and
credentials. There is also a deep understanding of the complex language of healthcare as well
as the EHRs, patient accounting systems, and third-party clearinghouses
What’s most exciting is that every Olive is able to learn collectively, like a network, so that
hospitals never have to solve the same problem twice. The healthcare industry is certainly a
prime target for RPA.

8. Hype Factor
Back in 1995, Gartner analyst Jackie Fenn developed the “Hype Cycle,” which caught a lot
of attention. The irony is that her framework was, well, hyped quite a bit! Her premise: Major
technologies undergo fairly predictable cycles.
In fact, the Hype Cycle has five of them, which include the following:
The Technology Trigger: There emerges a transformative technology that quickly gets the
attention of the media, entrepreneurs, and VCs. There are some prototypes for proof of
concepts. But the commercial viability of the technology has yet to be established. Peak of
Inflated Expectations: At this stage, the technology has shown to be effective and this gins
up lots of excitement. This can happen rapidly, especially with the power of social media and
the availability of huge amounts of cash to fund new technologies. As should be no surprise,
the hype gets to giddy levels. You will hear things like “game changer,” “inflection point,”
and so on.
Trough of Disillusionment: There begin to be signs that the enthusiasm is evaporating.
Some of the companies in the space are failing or not meeting expectations. Stock prices are
getting hit. Customers are realizing that the new technology is not generating the promised
ROI.
Slope of Enlightenment: Despite the issues and problems, the technology nonetheless gets
more refined. The result is that the impact gets stronger. But the media no longer shows much
interest and is focused on other red- hot categories.
Plateau of Productivity: Here the technology becomes mainstream and there are standards
of its use. It is really just a natural part of the business world. This framework is far from
perfect. Consider that Gartner has had some flubs with its own predictions, such as with BPM
and cloud computing. But the 5-step cycle is still useful. It’s a way to allow for a longer-term
perspective and to avoid being too early when it comes to adopting technologies.

9. Software-as-a-Service (SaaS) and Open Source

RPA software is still mostly on-premise. But there will probably be a transition toward a
SaaS model. Part of this will be about having a cloud-native platform, which will make it
easier for upgrades and data access (this will be critical for AI applications). SMBs (small
and medium size businesses) will want to have access to business process automation but
they will not always have the required sophistication to adopt the technology. This gives an
opening to a new type of service provider, a robotics-as-a-service (RaaS) operator, that can
help SMEs by automating their business routines and maintaining the software robots for
them. These RaaS operators can either focus on certain verticals, like car dealers or real estate
agents, or they can work as general automation providers in their local area. But also expect a
change in the business model. As RPA scales, the costs can get prohibitive because of the
per- bot fees. So customers will be looking for alternatives, say, a subscription approach.
Companies like SAP, Intellibot, and Nintex have been offering lower-priced strategies as a
way to get traction in the crowded RPA market, for example. Companies may also look to the
open source approach as a way to deal with this as well. But of course, there are other
benefits, such as a strong ecosystem of developers who keep innovating the platform.
Granted, the landscape for open source software is fragmented. The projects tend to be small
in terms of adoption – and the technology is not extensive. But this is expected to change. As
seen with other industries, open source has become ubiquitous. If anything, open source is
likely to be critical for the RPA category.
With that said, the industry is moving toward a more open and flexible model, and I
anticipate we’ll soon see parity between the two models, giving customers the ultimate in
flexibility and choice.”

10. Chatbots
Sometimes there is confusion between chatbots and RPA. While both involve the use of
software bots, there are clear differences. A chatbot is a system that uses NLP (natural
language processing) to communicate with people. In the consumer world, this would be Siri
or Cortana. But of course, many companies are using chatbots for handling customer service.
The market for this technology is growing quickly. According to research from Reports and
Data, the spending on chatbots is forecasted to go from $1.17 billion in 2018 to $10.8 billion
by 2026, which represents a compound annual growth rate of 30.9%.Some of the biggest
drivers include the pervasive use of social media and smartphones. Chatbots are expected to
have a notable impact on the bottom line too.
Then what does RPA have to do with chatbots? There are several use cases. First of
all, a chatbot can be used internally as an assistant to help employees with gaining access to
information or getting insights. But RPA can also help automate processes that deal with
customer interactions. Look at Aflac. The company, which has been around for more than 55
years, is a top provider of financial protection for health matters. There are over 50 million
customers. No doubt, Aflac has had to find ways to help automate processes. At first, Aflac
leveraged the Facebook Messenger bot service. It was built to address common questions and
would route to call center people when the chatbot did not have an answer.
But in the end, it failed and Aflac took it down. But the company did not give up. One
of the problems was that Facebook Messenger could not identify the user, which made it
impossible to provide personalized answers. Next, the database of Q&A information was
subpar. “They started using a chatbot system where their employees would answer the
questions. By doing this, they were able to create the right kind of data.” This proved to be
spot-on. The new chatbot system had a login to identify the users and a much richer database,
with the number of question types going from 500 to over 20,000. The result has been a
higher level of satisfaction from customers and reduced call center volume during peak times.
 11. Artificial Intelligence
AI is likely to be the most important driver for RPA. Based on research from PwC, this
technology is expected to add a whopping $15.7 trillion to global GDP by 2030, which is
more than the combined output of China and India. According to the authors of the report:
“AI touches almost every aspect of our lives. And it’s only just getting started.” Despite this,
AI is still limited and narrow. Consider your smartphone that has Alexa or Siri installed on it.
While the technology is powerful, the core functionality is really about handling simple
commands. It will not be until many years that we will be able to have a free-form
conversation with an AI device. In other words, when thinking about AI and RPA, look for
clear-cut business use cases. Interestingly enough, according to a report from PwC, the focus
should be on solving “boring” problems. The report highlights that executives are starting to
realize this as they run into challenges with implementing projects. The PwC survey shows
that – for 2020 – only 4% of companies plan to scale this technology across their
organizations. Rather, the priority will be on functions like finance, HR, tax, and compliance.
It may be as easy as something like intelligently extracting key information from a form. The
survey also shows that 44% of executives see AI as a way to operate more efficiently and
42% believe it will lead to better productivity. Even before thinking of putting together an AI
project or buying a new software platform, it makes more sense to look at your existing
systems. Keep in mind that they already likely have AI features built in! For example,
companies like Oracle, [Link], and Microsoft have been investing heavily in
revamping their software offerings. So the first step is to try out the new features. This should
not only lead to improvements in automation but also provide some understanding of how to
work with AI. Many of the RPA vendors are also aggressively adding AI features (this is
often referred to as intelligent automation or IA). But do not rush with these either. Take the
time to operationalize the AI within your workflows. What is relevant for your company?
What types of AI can be effectively managed? Granted, it’s a lot of work and there will be
the need for extensive planning – but this will be well worth it
Contract Management : “Poor contract management can be costly. Businesses might not
realize they are owed credits and they may overlook automatic renewal dates, or even fail to
send invoices. The automation capabilities available in RPA platforms can address many of
these issues, but may also be limited in their effectiveness due to the variability of language.
Invoice Automation: “For invoice processing, RPA can automate data input, reconcile error
correction and make binary decisions. But the real challenge is dealing with the many formats
different vendors use for their invoices. Using NLP and other machine learning techniques,
AI can understand and pull out necessary data from the invoices, normalize it to a structured
format, then send it back to the RPA platform for automated data input, error handling, etc.”
Financial Document Analysis: “Financial firms compile lots of data for monthly and
quarterly reports. RPA assists by automating data collection from various structured sources.
However, once you introduce unstructured PDF documents, RPA is generally ineffective.
With the OCR and NLP capabilities of an AI solution, relevant information can be
automatically pulled out and converted into a structured format so the RPA tool can deal with
it.”
Insurance Claims: “Insurance companies use RPA to automate some aspects of their claims
process, such as inputting data from structured sources and ensuring all required fields are
filled out. But insurance claims are full of unstructured data, such as photos showing auto
damage, PDFs of drivers’ licenses, or CT scans for a healthcare insurance claim. AI solutions
with machine learning can be used to extract relevant information from these sources, once
again adding value to the RPA tool

12. Privacy and Ethics

Governments are getting more aggressive in dealing with regulating the use of data,
especially as AI becomes a more powerful force. Europe has enacted the wide-ranging GDPR
law and California has introduced the CCPA (California Consumer Privacy Act). These will
likely be a framework for new laws in other countries and states. As RPA gets more
sophisticated and goes beyond rules-based approaches, companies will certainly need to be
mindful of the new laws – but also be aware of the ethical minefields. A breach can definitely
have a deleterious impact on a company, in terms of potential fines and reputational damage.
There should also be steps taken to help deal with the inevitable bias in the data. To this end,
models need to make sure the datasets have been cleaned up and are diverse, in terms of
backgrounds and characteristics. Next, companies will need to achieve explainability with the
models. Black boxes will ultimately lead to distrust with the outcomes.

Conclusion
As AI accelerates, the results should be even better for RPA. We will also see more startups
trailbalize new approaches. Conversational AI makes RPA more accessible and extends its
reach to a wider audience, unlocking the power of RPA. In other words, the future certainly
looks promising.
RPA is going to claim its place as a central platform for other enterprise automation
tools. Basically, RPA will become the repository for automation in the same way that
YouTube is a repository for video content. The centrality of RPA will run parallel with the
development of automation code that is more useful and reusable, enabling it to spread even
further than it previously has.”