Module 5 - RPA

-Prof. Anamika Dhawan

Cognitive?
Select the right process to automate.
Build a solid business case.
Get buy-in from senior management and IT.
Limit Initial scope of deployment
Select a good set and variety of data for
training.
Scale and expand
OCR
OCR was unable to extract Semistructured
data.
Cognitive Automation

It can unearth data in document , images and mails,

IQ Bots

Improves itself by unsupervised/Supervised

Machine learning.

By observing human set work and how does it

happen?
IQ Bots

AI technology enable to perform cognitive

automation.

IQ Bot combines the power of RPA with AI

technologies such as Computer Vision,
Natural Language Processing (NLP), fuzzy
logic and machine learning (ML) to
automatically classify, extract and validate
information from business documents and
emails.
 Benefits of IQ Bots

[Link] and easy to use interface.

[Link] accuracy during training phase.

Improves accuracy in training phase. Performance report guides the user Correction made by user help the
to get highest accuracy learning system
Benefits of IQ Bots
[Link] domain configurations to save time.

[Link] gives Bird’s Eye view of progress across all

process.

[Link] and categorizes various document

Benefits of IQ Bots

Rates itself on test to major reliability and learns

with every validations
Benefits of IQ Bots
[Link] existing automation platform triggers.

Emails/inboxes, File/folders and workflows.

[Link] more than 25 languages

IQ Bot Mechanism Sensors,Cameras,Radars and AIs

Let’s take an example of self Driven Car.

To build systems that can drive
autonomously
RPA with right set of AI technology
WorkFlow of IQ Bots

[Link] Instance

Digitized Document Uploaded to create learning

instance
[Link] sample Documents
[Link] Training
[Link] Processing
[Link] Validation by human.

Unable due to field extraction and rules

[Link] Monitoring
Cognitive RPA Use Cases

1. One example is to blend RPA and cognitive abilities for chatbots that make a customer
feel like he or she is instant-messaging with a human customer service representative.
Another is to create voice-powered bots for telephonic conversations.
2. Dynamic interactive voice response (IVR) can be used to improve the IVR experience. It
adjusts the phone tree for repeat callers in a way that anticipates where they will need
to go, helping them avoid the usual maze of options. Many email conversations can also
be automated. AI-based automations can watch for the triggers that suggest it’s time to
send an email, then compose and send the correspondence.
3. One of the most exciting ways to put these applications and technologies to work is in
omnichannel communications. Today’s customers interact with your organization across a
range of touch points and channels – chat, interactive IVR, apps, messaging, and more. When
you integrate RPA with these channels, you can enable customers to do more without needing
the help of a live human representative.

[Link] RPA abilities enable the automated system to understand the customer’s intent,
make sense of the unstructured data associated with the customer, predict behavior, and then
execute a request in the backend. AI and cognitive automation can also watch the overall
customer journey and weave themselves ever more effectively into it.

[Link] the example of a banking chatbot that automates most of the process of opening a
new bank account. Your customer could ask the chatbot for an online form, fill it out and
upload Know Your Customer documents. The form could be submitted to a robot for initial
processing, such as running a credit score check and extracting data from the customer’s
driver’s license or ID card using OCR.
Exception Handling with Cognitive RPA

[Link] the system picks up an exception – such as a discrepancy between the customer’s name on
the form and on the ID document, it can pass it to a human employee for further processing.
The system uses machine learning to monitor and learn how the human employee validates
the customer’s identity. Next time, it will be able process the same scenario itself without
human input.
[Link] instance might be an accounts payable process. Your automation could use OCR
technology and machine learning to process handling of invoices that used to take a long time
to deal with manually. Machine learning helps the robot become more accurate and learn
from exceptions and mistakes, until only a tiny fraction require human intervention.
IPA
Intelligent Process Automation (IPA) refers to the application of Artificial Intelligence and
related new technologies, including Computer Vision, Cognitive automation and Machine
Learning to Robotic Process Automation.

This convergence of technologies produces automation capabilities that dramatically elevate

business value and competitive advantages for our customers. UiPath creates IPA solutions by
using artificial intelligence (AI) to amplify the power of our RPA technology in three
[Link]’s IPA is based on advanced computer vision, unattended robotics, and integration
with third party cognitive services from Google, IBM, Microsoft and ABBYY.
IPA

With a continuous release cycle, UiPath is committed to its AI/Cognitive journey and will be
offering more Natural Language Processing and Machine Learning services on Cloud and
on-premise in upcoming releases.
AI+RPA

Software robots are great at completing routine and repetitive work. But with artificial
intelligence (AI), they can do a lot more.

They can read documents and emails, analyze language and images, and understand speech.
AI helps robots perform cognitive tasks, navigate uncertainty, and resolve inconsistencies.

And the more that robots can think and understand on their own, the more they can do. The
faster they can do it. And the bigger the impact they can make.
UiPath vision for RPA

For UiPath, AI isn’t an add-on or complementary technology.

We believe in the power of AI to transform, so we’ve built AI into every part of the UiPath
Platform.

AI isn’t an automation disruptor. AI is propelling automation forward, driving it to do more

while continuously improving by learning from humans.
With AI, RPA robots can
emulate a wide range of the
decision making and
problem solving that people
do—which expands the range
of work that you can
automate.
AI empower’s to work with
Unstructured data:
Let a robot work with ‘messy’ data in any form to perform tasks like invoice extraction, email
routing, and speech to text.

Variable processes:
Put robots to work on processes that can vary based on input and context, including
CV/résumé matching, purchase decisions, and language translation.

Predictive Outcomes:
Use AI-equipped robots to apply machine learning (ML) models to help with property
valuation, loan default prediction, and inventory forecasting.
Lifecycle
(AI+RPA)
Lifecycle (AI+RPA)

Discover—What are your best automation opportunities? AI with RPA can help you make data-driven decisions
on automation targets with Task Mining and Process Mining.

Build—With AI and RPA, you can build robots with the skills to process unstructured data, handle dynamic
interfaces, recognize speech, and apply ML models, and natural language processing (NLP). With over 25
prebuilt models including Document Understanding and AI Computer Vision we’ve got you covered.

Manage—Losing the “last mile” battle in scaling AI? Use automation to go the distance. With AI Center, you can
easily drag and drop ML models into your workflows. Drive compliance and governance with centralized
monitoring and management.

Engage—As your AI-powered robots get feedback on their performance, they get better at their work. Their
accuracy increases. Bring people into processes when human intervention is required with Action Center.
AI+RPA Use Cases:

Artificial Intellgence applications on which (UiPath) RPA can take advantage of to create
"Cognitive RPA" include:

● StanfordNLP Text Analysis,

● Google Text Analysis,
● IBM Watson NLU Text Analysis,
● Microsoft Text Analysis.
Use Cases

Use Cases where the integration of the above technologies have been proven are:
Google Cloud Contact Centre - Robots can retrieve and manipulate data from multiple systems
and legacy applications in a time-constrained contact center setup, resulting in faster response
time to customer query

Processing Unstructured Incoming Emails - The solution involved machine learning, natural
language processing,intelligent OCR and analytics capabilities, blending into a single user
interface the accuracy, speed and scale of UiPath’s Enterprise RPA Platform with the expert
capabilities from Microsoft Cognitive Toolkit, IBM Watson and ABBYY.
Stanford NLP - Using Stanford's natural language processing (NLP) capability, UiPath robots
contribute to the automation of financial reports by standardizing the business process logic
and validating data in tables and in the narrative

ABBYY FlexiCapture - In providing for the transformation of unstructured data inputs into
structured data that can then be further processed and communicated, such integration of
OCR technologies allows for a perfectly aligned interplay with RPA.
As a consequence, the content of an invoice that arrives via post, for example, does
not have to be manually entered into the system.
Instead, the document can be scanned at arrival, whereupon OCR can extract the relevant
information and the RPA software robots are able to send the data to the company’s database.
Types of Bots
Attended Bots

Attended bots typically work alongside humans to deliver specific parts of the process when it
can’t be automated from end to end. They can be thought of as a human ‘virtual assistant,’ and
exist to boost productivity with front-office tasks via attended automation. In a simple process, a
team member prompts or triggers a bot to respond, and it automatically begins to assist while
the employee interacts with it as it helps.
These attended bots can run on workstations, private servers, or in the cloud, and are useful in
reducing average call handle time, enhancing customer experience, and increasing compliance.

There are three types of attended bots: Task Bots, Meta Bots, and IQ Bots. The kind of bot that
will work best for you and your business depends on what you need assistance with: structured
or semi-structured data.
Attended bots

Structured data:
Includes data sources with clearly defined fields such as:
● Databases
● Spreadsheets
If you need help with tasks involving structured data, task bots and meta bots will best assist
you.
Semi-structured data:
● Includes any documents where the location of a field or line items may vary like:
● Mortgage applications
● Invoices
● Contracts
Attended bots

Task Bots:
Task bots consist of a collection of actions that have been programmed into them. They
replicate rule-based tasks and process structured data. Best used for repetitive tasks that may
be time-consuming or tedious for humans, the Task bots, can execute with extreme accuracy
and speed. These are the starting points for all RPA tools. As the most basic of attended bots,
they are required before you can launch a more complex type of bot. Task Bots are often seen
in HR administration, procure-to-pay, and quote-to-cash, just to name a few.
Attended bots

IQ Bots:
IQ bots are highly advanced (even at the most basic level) cognitive or intelligent automation
bots. They can learn on their own, in real-time, extract languages from the given data, and
perform the tasks according to their newly acquired knowledge. Using a combination of
computer vision, machine learning, and natural language processing, the IQ bot processes and
manages fuzzy rules, and can handle semi-structured and unstructured data with its cognitive
intelligence capabilities. Developers can add additional cognitive abilities to extract
information from semi-structured or even completely unstructured data.
Attended Bots

IQ bots require more human interaction than the other bots because they have to be
programmed, set up, and shown exactly how to behave (especially at first). These bots will
process the information presented to them and ask for feedback. It will automatically adjust
itself based on the answers you provide and keep trying until you say it is correct. The IQ bots
will store and learn from this information and keep trying until it reaches 100% STP. STP (or
Straight Through Processing) is where there is no visibility to human eyes of the steps/action
taken by an automated process. This makes the IQ bot perfect for processing semi-structured
data. When the format is generally the same across documents, the IQ bot can quickly pick up
on the repetition and recognize when a field is different.
Attended Bots

Meta Bots:
Meta bots are designed to be a reusable blueprint for an application and can quickly adapt to
changes. They can perform visual captures that are used for front-end automation. They are
best used for complex and scalable processes and in places of scripting. They can act as a script
or a map for inputted data and are usually paired with Task bots. While the Task bots take care
of the automatic, rule-based tasks on the front end, the Meta Bots can scale and process the
data even more efficiently when paired together. Meta bots are often used in enterprise-wide
automation because minimal maintenance is needed.
Unattended Bots
An Unattended bot is one that can complete tasks without human involvement. This is the aim
of many business processes, to rely on end-to-end automation, where a bot can execute an
entire process independently. These bots are often prompted by a specific preset schedule, or
logic in the process flow, such as an email. Then, they follow a rules-based process to
completion. Both attended and unattended bots are able (and usually do) to work together
within a greater RPA solution.
Unattended Bots

An example of unattended bots at work may be when there is a batch of new client
information, received in a spreadsheet, that needs to be entered into multiple applications. An
unattended bot would automatically take care of this data. Unattended bots are able to work
around the clock, so they are ideal for industries that also operate 24/7, such as hospitality,
healthcare, or telecommunications, among other applications. This automation means that the
bot has no need for human assistance or participation on the particular tasks the bot is
sequenced to perform. Simply put, they are set up to run entirely independently by using the
rules they have been taught.
Unattended bots

Like attended bots, unattended bots can run on workstations or private servers or in the cloud.
The most popular use of unattended RPA bots is in the back office, where a tremendous
amount of data is collected, sorted and analyzed. Unlike attended bots typically used for
short-term efficiency, unattended bots can be used for more long-term automation strategies
to better assist your workforce, colleagues and your business. The outcome being that a
business can reduce its operating costs, increase its productivity, re-allocate its employees’ skill
sets away from repetitive tasks, improve its compliance, and eliminate errors.
Web Scraping

Web scraping with Robotic Process Automation (RPA) utilizes bots to automate the process of
web data extraction from selected websites and store it for use.

RPA delivers faster results by eliminating the need for manual data entry and reducing human
errors.
Examples
❖ Extracting product details for competitor analysis.
❖ Extracting product data for comparison shopping on Amazon, eBay, or other similar
sites.
❖ Pulling stock prices for a more data-informed investment decision.
❖ Extracting information for brand and company monitoring.
❖ Utilizing an RPA scraping tool for website data before a site migration.
❖ Leveraging an RPA scraping to fetch financial information for in-depth market
research.
❖ Programmed bots can gather some social media data to understand customers’
views and mindsets easily and faster.
Web Scraping
Once data is pulled, it gets stored in databases or transferred to another system or
application. Automation not only allows to acquire data quickly but can pull actionable
information and automatically store it where required, whether a database or another
system.

The opportunities to successfully implement a website scraper in your daily work are almost
endless. Web scraping tools help you and your team dig deeper into particular topics to gain
insightful information for strategic planning and short-term assignments.
Web Scraping

Web scraping data from numerous sources involves a simple algorithm tailored to each
company’s goals and business requirements.
The web scraping steps involve:
1. Identifying relevant URLs.
2. Locating necessary data for extraction.
3. Writing and running the code.
4. Storing data in required format.
There are many types of web scraping solutions: from manual copy-paste operations to
browser extensions and robotic process automation. However, RPA remains the most popular
and effective solution when it comes to scraping large amounts of web data.
Parameter Manual Copy & Paste Browser Extension RPA

Scraping Can scrape any Browser extensions such as RPA tools such as ElectroNeek’s web
Capacity quantity that is ChromeScraper can scrape only scraping bot can scrape any quantity
humanly possible. one page at a time. of data.

Cost Inexpensive in Inexpensive with a few paid Involves more expenditure than the
comparison to the options. other two options. However, with
other two options. Popular extensions: ElectroNeek’s subscription model, the
ChromeScraper, Outwit hub initial cost is reduced drastically.
Firefox

Time The higher the It is faster than the manual Much faster than manual and browser
quantity of the data to process but longer than extension solutions. It can scrape a
scrape, the longer the automation as it can scrape only large quantity of data in just a few
process. one page at a time. minutes.

Efficiency Prone to more human A slight chance for errors as it Error-free as it involves almost no
errors than the other involves more manual manual intervention.
two options intervention than compared to
RPA solutions.

Flexibility Easy to do but is Less flexible than RPA in terms Easy to install and highly customizable
limited to human of customization and dealing in comparison to the other two
capacity as everything with UI elements like the ”Load [Link] also allows you to deal with
is manual. more” button. complex UI elements easily.
Benefits
Leveraging an automated data scraping tool for websites provides the following benefits:

● The data collected by an RPA bot is more accurate in comparison to manual data entry.
● Automation makes the web scraping process faster. Tasks that take weeks to complete by
hands get done within hours.
● Automation solutions are easy to implement and integrate within a company’s IT ecosystem.
● Low maintenance is needed as automated web scrapers require almost no upkeep in the long
run.

RPA bots integrated with other technologies and tools such as machine learning makes the bot an
even more powerful tool. For instance, machine learning combined with bots can locate companies’
websites from their logos
Challenges
An RPA site scraper depends heavily on GUI elements to locate relevant data, making it challenging to
automate when web pages don’t display content consistently. Some of the main web scraping
limitations include:
● User Interface (UI) features: Some UI features such as infinite scrolling, popup messages and
changing dynamic web content make it difficult for scrapers to handle data extraction. The
solution would be additional bot programming for specific complex features.
● ”Load more” button: Many pages load data in small parts using the ”Lload more” button. In
such cases, the bot will stop at the end of a visible part end and miss further information
without exploring more data. Creating an if-loop within the program of a bot to click the
“lLoad more” button until no more buttons remain on a site
Challenges

● Pop-up advertisements: These advertisements often hide GUI elements from the RPA
Bots view and impair data extraction. An effective solution would be to add the
ADBlocker extension for your browser.
● Protected systems: Some web pages utilize advanced security technology to prevent
their site from automated scraping, which becomes another roadblock. There are two
possible solutions:; first,ly to either working with a company that offers web data as a
service, or combining RPA with proxy servers to build bots that cannot be
distinguished from humans.
Web Scraping with RPA

Let’s say you want to automate the procedure of collecting news on various topics.

The first step is creating a simple file (such as an excel file) for the bot to store collected data.

Secondly, the bot must be programmed and looped through search queries and load results.

Lastly, the RPA bot can parse through relevant data, extract and store it quickly and accurately.