International Journal of Scientific & Engineering Research Volume 13, Issue 5, May-2022 44
ISSN 2229-5518
IJSER © 2022
http://www.ijser.org
Invoice Processing using Robotic Process
Automation (UiPath tool) and Artificial
Intelligence
Aishwarya Bhargava
ABSTRACT
This paper describes our recent effort to develop an automative, efficient, accurate, and secure application to transform invoice
processing in Finance operations. As a prime example of the technology’s potential for driving efficiency, Robotic Process
Automation (RPA) can be applied to several finance and accounting operations, invoice processing. RPA bot can automate data
input, error reconciliation, and some of the decision-making required by finance staff when processing invoices. At the same time,
automation can limit errors in such processes and reduce the need for manual exception handling. Most of the invoices are
delivered via emails these days, and with the increase in this digital world, the world of scammers has also increased. The finance
world is most prone to scams. So, in the paper we will be introducing a front-line screening of mails before they get downloaded
for invoice processing, the screening will be done with the help of Artificial Intelligence.
KEYWORDS: Artificial Intelligence, Invoice processing, Machine Learning, Robotic Process Automation, UiPath tool
I. INTRODUCTION
According to various research papers published in the past decades, it can be said that manual invoice processing takes a number
of days for complete execution and even after spending a lot of time, the outcome is not an ideal output as it might be prone to
human errors. There are multiple drawbacks related to manual invoice processing like missing invoices, confusing invoices,
missing data, errors in the invoice data extraction, wrong or missing contact information, and excessive time consumption. In order
to eliminate all these errors to some extent, a few years ago the methodology of automated invoice processing was introduced.
Many technologies have accomplished a great deal of efficiency and accuracy in this field. In this research paper, we will be
discussing the implementation of invoice processing using Robotic Process Automation technology and the implementation will
be done using the UiPath tool with the security of Artificial Intelligence.
Due to the advancement in this era of digitization, the dependency on email has been increasing day by day. The increasing
dependency calls for a way to manage the huge amount of data or emails. The emails conveyed include important as well as
phishing emails. Phishing emails often lead to malicious websites and result in sharing personal details with the attackers. These
phishing emails might look like legitimate invoice emails to the company but on the back-end lead to the downfall of the finances
of the company. For the safety of the company or any other organization, it needs to deploy the program of Artificial Intelligence
which will identify the scams and will not let them move forward for further processing.
II. LITERATURE
[1] Artificial Intelligence
Artificial Intelligence is composed of two words Artificial and Intelligence, where Artificial defines "man-made," and intelligence
defines "thinking power", hence AI means "a man-made thinking power." So, we can define AI as: "It is a branch of computer
science by which we can create intelligent machines which can behave like a human, think like humans, and able to make
decisions." AI is an intelligent entity created by humans. It is capable of performing tasks intelligently without being explicitly
instructed. It is also capable of thinking and acting rationally and humanely.
o Machine Learning: ML teaches a machine how to make inferences and decisions based on past experience. It identifies
patterns, analyses past data to infer the meaning of these data points to reach a possible conclusion without having to
involve human experience. This automation to reach conclusions by evaluating data, saves a human time for businesses
and helps them make a better decision.
o Deep Learning: Deep Learning is an ML technique. It teaches a machine to process inputs through layers in order to
classify, infer and predict the outcome.
o Neural Networks: Neural Networks work on the similar principles as of Human Neural cells. They are a series of
algorithms that captures the relationship between various underlying variables and processes the data as a human brain
does.
o Natural Language Processing: NLP is a science of reading, understanding, interpreting a language by a machine. Once
a machine understands what the user intends to communicate, it responds accordingly.
o Computer Vision: Computer vision algorithms tries to understand an image by breaking down an image and studying
different parts of the objects. This helps the machine classify and learn from a set of images, to make a better output
decision based on previous observations.
IJSER
International Journal of Scientific & Engineering Research Volume 13, Issue 5, May-2022 45
ISSN 2229-5518
IJSER © 2022
http://www.ijser.org
o Cognitive Computing: Cognitive computing algorithms try to mimic a human brain by analyzing
text/speech/images/objects in a manner that a human does and tries to give the desired output.
[2] Advantages of using AI
o AI drives down the time taken to perform a task. It enables multi-tasking and eases the workload for existing resources.
o AI enables the execution of hitherto complex tasks without significant cost outlays.
o AI operates 24x7 without interruption or breaks and has no downtime
o AI augments the capabilities of differently abled individuals
o AI has mass market potential; it can be deployed across industries.
o AI facilitates decision-making by making the process faster and smarter.
[3] Disadvantages of using AI
o The implementation cost of AI is very high.
o The difficulties with software development for AI implementation are that the development of software is slow and
expensive. Few efficient programmers are available to develop software to implement artificial intelligence.
o A robot is one of the implementations of Artificial intelligence with them replacing jobs and lead to serve unemployment.
o Machines can easily lead to destruction if the implementation of machine put in the wrong hands the results are
hazardous for human beings.
[4] Robotic Process Automation
Robotic process automation (RPA) is a software technology that makes it easy to build, deploy, and manage software robots
that emulate human’s actions interacting with digital systems and software. The technology is used for software tools that
automate human tasks, which are manual, rule-based, or repetitive. Typically, it is like a bot that performs such tasks at a
much higher speed than a human alone. These RPA software bots never sleep make zero mistakes and can interact with in-
house applications, websites, user portals, etc. They can log into applications, enter data, open emails and attachments,
calculate and complete tasks, and then log out. An RPA workforce is precise, accurate and immune to boredom. It can also
be scaled more easily than your human workforce. RPA can perform just about any rule-based work and can do so through
interaction with any software application or website. It’s a robotic connection to the human world of the computer user interface.
If a human can do it, a robot can do it in virtually the same way.
[5] UiPath tool
UiPath provides an RPA platform to automate digital business processes across front-end and back-end office tasks. It
includes products such as studio, software robots, and orchestrator. It provides solutions in various sectors such as Banking,
Finance, BPO (Business Process Outsourcing), Insurance, Retail, Telecom, Manufacturing, Healthcare, Public Sector, etc. It
allows users to perform Web Automation, Desktop Automation, GUI (Graphical User Interface) Automation, SAP Automation,
Mainframe Automation, Citrix Automation, Excel Automation, Screen Scraping, and Screen Recorder, etc.
[6] Advantage of using RPA (UiPath)
o Accessibility: The extensible platform of UiPath provides hundreds of built-in, customizable, shareable activities and
deep integrations with the help of various technologies that are already in use. The UiPath has mobile and browser
accessibility.
o Rapid: The ecosystem of the UiPath tool is optimized for faster development and designed in such a way to deliver the
Quick return on investments.
o Artificial Intelligence: The Artificial Intelligence robotic manager reduces the automation costs and meets the service
level with the help of synchronized queue work and robot deployments with scheduled workflows and events.
o Scalability: The RPA (Robotic Process Automation) at the enterprise level is expected to deploy and manage various
variety and number of processes from front-end-office to back-end-office with regardless of complexity. The user can train
tens, hundreds, or thousands of Robots at the same time by using the UiPath tool. This tool also has absolute consistency
in job performance.
o Quality of the Agile process: The Agile process is the technique that supports continuous iteration of development and
tests the developing module throughout the whole software development life cycle of the project. The UiPath tool consists
of an agile technique, which is very useful according to both client and organization.
o Flexibility: Flexibility is the key advantage of the UiPath tool to build an effective digital workforce. The UiPath tool
provides flexibility to both the user and the organizations.
[7] Disadvantage of using RPA (UiPath)
o The UiPath tool of Robotic Process Automation improves the efficiency of organizations by reducing the repetitive human
efforts, but there are some limitations of automated work. When automation is applied, then, it requires judgment related
to work.
o The RPA tool UiPath is not a cognitive computing solution.
o This tool cannot read any data which is non-electronic with unstructured input.
o When an enterprise uses the UiPath tool to automate the task, then the enterprise needs to be aware of several inputs
that are coming by multiple sources.
o The local hosting of the UiPath Orchestrator server is not available in the community edition of UiPath.
o The main disadvantage in the UiPath tool is its auto-start feature of UIRobot.exe.
o The Number of Robots is limited in the Orchestrator community edition.
o The UiPath tool asks the user to activate the libraries from the nugget package manager, which gets deleted every time.
IJSER
International Journal of Scientific & Engineering Research Volume 13, Issue 5, May-2022 46
ISSN 2229-5518
IJSER © 2022
http://www.ijser.org
[8] What is invoice email scam?
Accounts payable (AP) departments at large enterprises pay thousands of invoices each month. It’s difficult for AP teams to
keep track of every invoice that comes in, making it tough to detect upward trends in pricing or the addition of surcharges over
time. Suppliers can raise prices without warning or send multiple invoices for smaller amounts that are less likely to be
scrutinized. With such a large volume of invoices, companies leave themselves at risk for overpayment and potential invoice
fraud.
According to the “2020 Report to the Nations,” published by the Association of Certified Fraud Examiners (ACFE),
organizations lose an estimated 5% of annual revenue due to fraud. For organized scammers, accounts payable (AP)
departments are often perceived as a poorly guarded cashbox they can target for scams or other malicious attacks. In most
cases, payments are processed through these departments, which makes this function vulnerable and a prime target for theft.
[9] What are the frequent invoice frauds?
o Fake Vendors and False Billing: Most companies have long-term business contracts with vendors and suppliers. These
are routine purchases; payments are often pre-approved without much inspection and scrutiny. Fake companies can
mimic the actual vendors and send fake invoices. According to the ACFE report, fake billing is hard to detect and poses
a significant risk, with a median loss of $100,000. An employee can also engage in malicious activities by creating fake
vendors and sending fake invoices. An example of this is an Oregon woman who embezzled millions of dollars from her
company over nearly 15 years. Scammers can also work with internal employees to create payment requests for
goods/services that the company didn’t receive and then keep the processed payment for themselves. This is what
happened with a former Honda employee in Ohio, who defrauded the company out of $750,000.
o Overbilling/Overpayment: There are instances when a vendor deliberately introduces new line items to inflate the
amount owed. When an AP process is manually driven and experiences a surge in invoice volume, these wrongly added
line items can go unnoticed. And that results in the company getting overbilled.
o Expense Account Malfeasance: Expense account padding works a bit like fake billing. When employees submit expense
reports for reimbursement on things like mileage, meals on the company dime, and office supplies, they’re generally
supposed to include receipts. But these documents aren’t always closely checked and companies can easily wind-up
rubber-stamping receipts that have no business getting approved. And around a quarter of businesses don’t even require
receipts for reimbursement.
[10] How to prevent invoice frauds using AI?
o Automated Invoice Review: Developing essential protection against accounts payable fraud involves matching key
documents generated across the procurement stages. It includes matching invoice line-item information with purchase
order details and reviewing receipts to ensure goods and services have been delivered. It often involves reviewing other
communications with the vendor. When done manually, this can take days. But with AI-powered automation, the entire
reconciliation cycle can happen in just minutes.
o Employee Malintent Monitoring and Tracking: Employees engaged with accounts payable fraud often operate as
individuals or small groups, having gained strategic information about operations. Technology systems, like ML-based
anomaly detection, can track and monitor employee engagements in the AP process in real time to thwart malicious
behavior. Training employees on such procedures can increase awareness. Building a loyalty program can also help
charter a course toward ethical actions.
o Building Vendor Match Capability: Since AP departments sometimes encounter invoices generated by fake companies,
it’s important to detect such activity at its initial stages and before damage is done. Detection can be done through fuzzy
matching technology, in which the system scans multiple repositories, looks through the existing vendor database and
checks various documents to detect fakes.
o Automated Approval Workflow: Automated approvals reduce the time-consuming manual approval processes.
Automation of approvals ensures compliance by leaving a digital trail of the actions performed across each stage. Training
and generating awareness among employees can help mitigate the risk of malicious intent.
o Digital Solutions to Enforce Compliance: Digitizing the entire invoice processing workflow can reduce AP fraud.
Digitization calibrates actions with time stamps and personas, which can later be used to conduct system audits. Such
systems can be designed to generate automated reports, where employee activities can be monitored and subsequent
areas of improvement can be identified. Also, real-time alerts can give administrators holistic insights into process
operations.
[11] What is invoice processing?
Invoice processing by definition is a business function performed by the accounts payable department which consists of a
series of steps for managing vendor or supplier invoices from receipt to payment, and recorded in the general ledger. Invoice
processing is often performed with software and it is commonly referred to as automated invoice processing or invoice
automation for short. An invoice processing flowchart is a structured guide detailing the steps for how accounts payable is to
process vendor invoices.
The following steps need to be performed for the invoice processing:
o Step 1: The traditional invoicing process starts with receiving paper invoices in the mail or PDF invoices via email or other
electronic means from your supplier.
o Step 2: Next, that invoice is internally assigned for processing, which involves either scanning or manually entering the
invoice data into your ERP system or accounting software.
IJSER
International Journal of Scientific & Engineering Research Volume 13, Issue 5, May-2022 47
ISSN 2229-5518
IJSER © 2022
http://www.ijser.org
o Step 3: From there, the invoice amounts must be checked and approved for payment, which can involve being coded for
the right account, project, or cost center. If your business uses purchase orders, this may also involve PO matching.
o Step 4: Last but not least, the invoice is sent to the appropriate person, often a department head, for review and final
approval before being processed for payment.
[12] Drawbacks of manual invoice processing
o Affects Cash Flows: Manual invoicing is time-consuming and boring, and that can lead to mistakes and delays in
sending invoices to your customers at the right time which can affect getting paid quickly, thus affecting your cash flows.
It is also difficult to track missed or delayed payments as you don’t get to see the pending invoices at a glance.
o It’s costly: On average, it costs about more to manually process just one invoice. If your organization manually
processes hundreds and thousands of invoices annually, you can just do the calculations! Manual operation is not a
good idea anymore.
o Cannot use latest payment modes: With manual invoicing you cannot use the latest mode of payments, ex. Online
credit/debit card, UPI, QR Codes, PayPal, etc., which can significantly reduce payment receivable life cycles for your
company and makes life easy for your customers too.
o It’s time-consuming: With a manual system, it takes much longer to prepare and send an invoice. Your team spends
more time to get less significant work done. This further contributes to higher cost of operation, as you’d have to pay
more people for longer working hours.
o It causes you to lose money: Issues of missed invoices, delayed payments, wrong values and missed entries are all
very common with manual invoice management. Whenever this happens, you lose money and your company’s bottom
line is impacted.
o It’s boring: And often very annoying. This leads to ineffectiveness for your team. Nobody wants to sit at the same old
desk and enter mind-numbing data for hours, every business day, but unfortunately, that’s what manual invoicing is all
about.
o It’s prone to human error: Since processes are manned by error-prone humans in a manual system, mistakes are
inevitable. They can even lead to bigger issues, which only means wasted time and money for your organization.
o It’s a lot of hassle: Manual invoice processing is often complex and inconvenient. To prepare invoice one needs to
extract information, match details, send it back and forth for different workflows like approvals, prepare checks, and do a
lot of other complicated stuff.
o It requires more hands: A manual invoice processing procedure needs more human workers to function, given that the
related processes are hand-operated. And you would have to pay all of those humans.
o It delays progress: In a manual invoice processing system, since your workers are always sitting on the same desk
entering data or performing some other related tasks, their time is not freed up to work on more value-adding or growth-
oriented action items. This slows down advancement and development, making it difficult to implement new strategies
and make progress on other fronts.
o You can’t easily track your cash flow: Without clear and quick information of pending invoice, you cannot ascertain a
proper cash flow for your organization. A manual system doesn’t let you track this very easily.
o It’s vulnerable to security risk: A manual system is easily exposed to security problems like data theft, data loss, and
the like. When stored locally, anybody can illegally access and use your data as they want. Data security is important
because it represents the financial operation of your organization.
[13] Advantages of automated invoice processing
o Time saving – Reducing the amount of manual input required from the accounts payable team to be able to properly
process incoming invoices means there’s more time for them to spend on other activities, like building strong
relationships with suppliers and working on further improving AP efficiency.
o Cost saving – Studies show that it costs up to $11.57 to process a single invoice, on average. Streamlining the process
and reducing manual requirements can reduce this cost, generating savings in the AP department. AP automation can
IJSER
International Journal of Scientific & Engineering Research Volume 13, Issue 5, May-2022 48
ISSN 2229-5518
IJSER © 2022
http://www.ijser.org
also generate savings in the form of making it easier to spot opportunities to access early payment discounts and
reducing the chance of invoices going missing, therefore avoiding late payment fees.
o Reduced errors – The accuracy with which invoice processing is carried out is massively important. Errors caused by
manual input can include overpayments and duplicate payments, both of which are harmful to the bottom line.
Automating invoice processing can significantly minimize the chance of these errors occurring.
o Fraud protection – Payment’s fraud is a constant risk with traditional methods of invoice processing, but automated
invoice processing removes a lot of the danger. Access to certain abilities, like approving invoices or making payments,
can be limited to just the right people through access controls, meaning fraud is significantly more complex to perform.
o Easier auditing – Finally, automating the invoice processing cycle means that there’s a secure, backed-up trail of all
AP activities that can be relied on for future audits and process reviews. Documents, invoices, receipts, and messages
that are involved with the same transaction can generally be linked together, creating an easily trackable audit trail to
follow.
[14] Automation of invoice processing using RPA (different tools and methodology)
o Document understanding in UiPath: In this methodology, we follow the Optical Character Recognition (OCR)
technique and the machine learning algorithm in the back-end. This methodology has a number of steps that need to
be followed. First, we create a taxonomy for each document type if required, after this pre-processing step we move first
processing step i.e., the digitization step, in this step we digitize the document using the different OCR engines available.
Next, we classify the document using some of the key phrases so that we don’t accidently process the wrong document.
After classification, we move to the step of data extraction where we use ML algorithm for extracting data from invoices.
After this step we have an option of validating the extracted output manually, this can be done using human validation
station, this station helps us alter any data if required, this helps in achieving 100% accuracy. After all the data is validated,
the output so extracted is stored in excel and can be used for further processing. The drawback of this methodology is
that due to the addition of human validation station, it takes a little bit longer but this step helps guarantee that the extracted
output is 100% accurate.
o Screen Scraping in UiPath: In this methodology, the pdf is opened via application (Acrobat Reader) and then anchors
are used to recognize information. The drawback of this methodology is that, the automation might give error when unable
to find the selector showcasing an anchor or any element. This methodology uses Optical Character Recognition (OCR)
technique to find the information and elements.
o Regex in UiPath: In this methodology, the entire pdf is read via Optical Character Recognition (OCR) technique and
is stored as a string, from the output string the necessary details are extracted via regex technique and then can be used
for further processing by adding the extracted value to excel or any other file type as per requirement. The drawback of
this methodology is that it is dependent on the regex expression, which might not be correct with different type of invoices.
o IQ Bot in Automation Anywhere: In this methodology, the invoice is processed using the tool called IQ Bot. IQ Bot is an
intelligent document processing platform from Automation Anywhere that can be used to classify, extract, and validate
content from documents. IQ Bot allows us to process unstructured data using AI technologies such as computer
vision, Natural Language Processing (NLP), Machine Learning (ML), and text classification. IQ Bot converts
unstructured data from invoices to structured data.
III. METHODOLOGY
The process diagram/methodology followed by the paper is as follows:
IJSER
International Journal of Scientific & Engineering Research Volume 13, Issue 5, May-2022 49
ISSN 2229-5518
IJSER © 2022
http://www.ijser.org
Figure 1 Process Diagram of the entire automation
In this automation, we have attempted to provide a secure and an accurate automation. The spam detector has an accuracy of
98.6% and the invoice data extraction part has the accuracy of more than 99.8%.
1) Spam Detection Process:
In this process we input the body of the mail and do analysis on the string and predict the type of the mail (spam/ham).
The first step of this process is to create a model which helps predicting the status of the mail. We start by importing a
number of libraries (Pandas, RE (regular expressions), nltk (count vectorizer, train test split, multinomial naïve bayes,
confusion matrix, and accuracy score), sys (system), and sklearn (stopwords and porter stemmer). We then read the
dataset that is used for training and testing the model, this is done using Pandas library. Next, we go to the step of data
cleansing. In data cleansing first we remove all the characters other than lower- and upper-case alphabets, second, we
convert all the alphabets to lower-case alphabets, third we split each sentence into words for easy removal of stopwords
(stopwords are the words which do not help in the prediction), forth we find the root words for each word left and then
we join each of the word into a string and append it to an array. Next, we go to the step of creating bag of words. We do
that by initializing the countvectorizer with the max limit as 2500 and then fit-transform the new list. This step helps filter
out the most frequently used words in the messages, this also helps in easy data cleansing. Next, we convert category
column data to dummy data making ham as 1 and spam as 0. Next, we split the dataset into training and testing
dataset. We have used naïve bayes classifier as it works really good with NLP (natural language processing). Next, we
fit multinomial naïve bayes with X and Y training dataset in order to create the model. We can use confusion matrix and
accuracy score to check the no. of values correctly predicted and also to get the accuracy percentage. After the model
is developed, the main task starts. We get the input from user via command line argument (the input is the mail body
that needs to be processed), after getting the input we perform the data cleansing steps and also the bag of words step.
After that we apply model to the list and get the prediction stored in a text file for further use.
IJSER
International Journal of Scientific & Engineering Research Volume 13, Issue 5, May-2022 51
ISSN 2229-5518
IJSER © 2022
http://www.ijser.org
Figure 2 Process Diagram of Spam-mail detection process
2) Invoice Data Extraction Process:
In this process we extract all the relevant data there is from the invoice and consolidate it in an excel for the particular
date. The first step is to create a taxonomy, specified according to the data that needs to be extracted. Next, we load
the taxonomy and digitize the document using OCR engine. We then classify the document, so that for the further steps
only invoices are sent. Post this step we perform data extraction using ML extractors, they have built-in machine
learning algorithms. This step extracts all the data from the invoices which can be cross-checked in the next step of
validation station, where the person can check and validate each extracted value, so that no value is incorrect in the
final output. The approved data is then exported into datasets which then can be written in excels and mailed to the
respective user.
Figure 3 Process Diagram for Invoice Data extraction process
IV. LITERATURE REVIEW
The invoicing process is an important part of a wider set of business processes including the placing and acceptance of an
order, delivery and payment. Some research papers regarding invoice processing system were studied by us. These included
various techniques of invoice processing using cognitive approach, generic system which uses OCR engine, using CBR (Case-
based reasoning), optimization approach which makes use of SVM, Entropy, using IQ Bot of Automation Anywhere, using regex
technique of UiPath, etc. Some of these approaches for invoice processing system are mentioned below.
Sagar Sahu, Sania Salwekar, Atharva Pandit, Manoj Patil in their paper proposed the methodology of using UiPath’s inbuilt OCR
engines for example Microsoft OCR, UiPath OCR, Tesseract OCR, etc. They extracted the data from the invoices using the
IJSER
International Journal of Scientific & Engineering Research Volume 13, Issue 5, May-2022 52
ISSN 2229-5518
IJSER © 2022
http://www.ijser.org
OCR engine and then used the regex expressions to extract relevant information from the string and fed it to excel for further
processing.
Devanshi Desai, Ansh Jain, Dhaivat Naik, Nishita Panchal, Dattatray Sawant in their paper proposed the methodology of using
Automation Anywhere’s IQ Bot. IQ bot is used to extract structured or unstructured data as it uses AI, ML, NLP to learn and
extract information from the documents. The reason that they gave for choosing this is that it has AI-based learning algorithms to
recognize and classify the content and the fact that it doesn’t depend on OCR is a really good plus point. Also, because it’s an
IQ bot, it continuously learns from user feedback and validation therefore increases accuracy. They used ML based IQ Bot. The
IQ bot contains a ML tool using Naïve Bayes Algorithm. Bayesian Naïve Bayes algorithm is based on Bayesian decision theory
and is widely used. The algorithm is built on features that it employs to identify the text in question.
Enes Aslan, Ethem Unver, Tugrul Karakaya, Yusuf Sinan Akgul in their paper proposed a new invoice parsing method which
consists of two-phase optimization structure and eliminates invoice classes. The first phase uses individual invoice part
detectors such as SVM, maximum entropy and HOG to produce candidates for the various parts of different types of invoices. At
the second phase, the basic idea is to divide an invoice into different parts and then arrange it together. As PBM is an
optimization-based method, it can handle any type of invoices. The proposed system is tested with real invoices and are found
to be promising for the real-world experiences.
V. REQUIREMENT ANALYSIS
Requirement’s analysis involves all the tasks that are conducted to identify the needs of different stakeholders. Therefore,
requirements analysis means to analyze, document, validate and manage software or system requirements. High-quality
requirements are documented, actionable, measurable, testable, traceable, helps to identify business opportunities, and are
defined to a facilitate system design. After the extensive analysis of the problems in the system, we are familiar with the
requirement that the current system needs. The system requirements are categorized into the functional and non-functional
requirements. These requirements are listed below:
A. Hardware requirements
a. Minimum 4GB RAM.
b. 200 MB of free Hard Disk space.
c. Browser: Chrome(v49), Internet Explorer (v10) or higher.
d. Processor: 3GHz or higher
B. Software requirements
a. Operating System: Windows 7 and above.
b. UiPath Studio.
c. UiPath Robots.
d. UiPath Orchestrator.
e. Python.
C. Functional requirements
Functional requirement are the functions or features that must be included in any system to satisfy the business needs and
must be acceptable to the users. Based on this, the functional requirements that the system must require are as follows:
a. Customer Name: The Name of customer is required to identify the person who has initiated the transaction. The name
should be spelled correctly in order to avoid confusion.
b. Total Amount: This field gives us the final amount of the product.
c. Invoice Number: This field shows us the Receipt (Invoice) number of the transaction. This number is useful in servicing
centers to identify the product warranty or guarantee period.
d. Invoice Date: Date at which the invoice was printed.
e. Product Description: This field gives us the information about either products or services including prices and
quantities. Often includes standard product description and inventory number.
f. Currency: Currency of the total amount to avoid any confusion during the time of payment or calculations.
D. Non-functional requirements
Non-functional requirement is a description of features, characteristics and attribute of the system as well as any constraints
that may limit the boundaries of the proposed system. The non-functional requirements are essentially based on the
performance, information, economy, control and security efficiency and services. Based on these the non-functional
requirements are as follows:
a. Security: - Security requirements ensure that the software is protected from unauthorized access to the system and its
stored data. It considers different levels of authorization and authentication across different user’s roles. For instance,
data privacy is a security characteristic that describes who can create, see, copy, change, or delete information.
Security also includes protection against viruses and malware attacks.
b. Reliability: - Reliability defines how likely it is for the software to work without failure for a given period of time.
Reliability decreases because of bugs in the code, hardware failures, or problems with other system components. To
measure software reliability, you can count the percentage of operations that are completed correctly or track the
average period of time the system runs before failing.
IJSER
International Journal of Scientific & Engineering Research Volume 13, Issue 5, May-2022 53
ISSN 2229-5518
IJSER © 2022
http://www.ijser.org
c. Performance: - Performance is a quality attribute that describes the responsiveness of the system to various user
interactions with it. Poor performance leads to negative user experience. It also jeopardizes system safety when it’s is
overloaded.
d. Availability: - Availability is gauged by the period of time that the system’s functionality and services are available for
use with all operations. So, scheduled maintenance periods directly influence this parameter. And it’s important to
define how the impact of maintenance can be minimized. When writing the availability requirements, the team has to
define the most critical components of the system that must be available at all time.
e. Scalability: - Scalability requirements describe how the system must grow without negative influence on its
performance. This means serving more users, processing more data, and doing more transactions. Scalability has both
hardware and software implications. For instance, you can increase scalability by adding memory, servers, or disk
space. On the other hand, you can compress data, use optimizing algorithms, etc.
VI. TECHNOLOGIES USED
A. UiPath Studio
UiPath Studio is a complete solution for application integration, and automating third-party applications, administrative IT
tasks and business IT processes. One of the most important notions in Studio is the automation application.
An application is a graphical representation of a business process. It enables you to automate rule-based processes, by
giving you full control of the execution order and the relationship between a custom set of steps, also known as activities in
UiPath Studio. Each activity consists of a small action, such as clicking a button, reading a file or writing to a log panel.
The main types of supported workflows are:
a. Sequences - suitable to linear processes, enabling you to smoothly go from one activity to another, without cluttering
your workflow.
b. Flowcharts -suitable to a more complex business logic, enabling you to integrate decisions and connect activities in a
more diverse manner, through multiple branching logic operators.
c. State Machines - suitable for very large workflows; they use a finite number of states in their execution which are
triggered by a condition (transition) or activity.
d. Global Exception Handler - suitable for determining the workflow behavior when encountering an execution error, and
for debugging processes.
B. E-Mail Activities Package
a. The Mail Activities Pack is designed to facilitate the automation of any mail-related tasks, covering various protocols,
such as IMAP, POP3 or SMTP. UiPath also features activities that are specialized for working with Outlook and
Exchange.
b. Activities such as Save Mail Message and Save Attachments are not intended to be used with certain mail protocols.
Instead, they save the MailMessage object variable retrieved from activities such as Get IMAP Mail Message to a
specified folder on the current machine.
C. Document Understanding Package
The activity package that enables Machine Learning Document Understanding features, such as the Machine Learning
Extractor and the Machine Learning Classifier. The UiPath Document Understanding Framework is designed to help users
combine different approaches to extract information from multiple documents, not necessarily with the same structure.
D. Excel Activities Package
a. The Excel activities package aids users to automate all aspects of Microsoft Excel, as we know it is an application
intensely used by many in all types of businesses.
b. It contains activities that enable you to read information from a cell, columns, rows or ranges, write to other
spreadsheets or workbooks, execute macros, and even extract formulas. You can also sort data, color code it or
append additional information.
E. UiPath Robot
a. The Robot is an execution agent, meaning that you have to provide it with the automation applications you want it to
run.
b. After creating an automation application in Studio, it needs to be published locally or to Orchestrator. Once an
application is published, you can send it to the Robot machine and start executing it.
c. This is populated by default as follows:
i. When NOT connected to Orchestrator
ii. When connected to Orchestrator - the default Orchestrator feed
F. Intelligent OCR Activities Package
The package contains core activities that enable the usage of a complete document processing framework, from taxonomy
definition, digitization, document classification, data extraction, data validation and classifier / extractor training. Taxonomy
can also be created specifically for each type of document.
G. Python software
The Python programming language is a popular programming language because of its simplicity, ease of use, open-source
licensing, accessibility, renowned community, great support and help, tons of packages, tutorials, and sample programs –
which make it easy for a beginner to learn and code with Python. Python can be used to develop a wide variety of
IJSER
International Journal of Scientific & Engineering Research Volume 13, Issue 5, May-2022 54
ISSN 2229-5518
IJSER © 2022
http://www.ijser.org
applications – ranging from Web, Desktop GUI based programs/applications to science and mathematics programs, and
Machine learning and other big data computing systems. Machine learning (ML) is a type of artificial intelligence (AI) that
allows software applications to become more accurate at predicting outcomes without being explicitly programmed to do so.
Machine learning algorithms use historical data as input to predict new output values.
H. Pandas Library
Pandas is a Python library used for working with data sets. It has functions for analyzing, cleaning, exploring, and
manipulating data.
I. RE Library
Regular expressions (called REs, or regexes, or regex patterns) are essentially a tiny, highly specialized programming
language embedded inside Python and made available through the re module. Using this little language, you specify the
rules for the set of possible strings that you want to match; this set might contain English sentences, or e-mail addresses, or
TeX commands, or anything you like.
J. NLTK Library
NLTK is a toolkit build for working with NLP in Python. It provides us various text processing libraries with a lot of test
datasets. A variety of tasks can be performed using NLTK such as tokenizing, parse tree visualization, etc.
K. SYS Library
The sys module in Python provides various functions and variables that are used to manipulate different parts of the Python
runtime environment. It allows operating on the interpreter as it provides access to the variables and functions that interact
strongly with the interpreter.
L. SKLEARN Library
Scikit-learn (Sklearn) is the most useful and robust library for machine learning in Python. It provides a selection of efficient
tools for machine learning and statistical modeling including classification, regression, clustering and dimensionality
reduction via a consistence interface in Python.
VII. RESULTS
Following are the results we obtained after performing a series of test with real time invoices. The accuracy we obtained from
spam-detection automation is 98.6% and that of data extraction is 100%.
1) This is the Sample copy of an invoice which we have used for testing.
Figure 4 Sample copy of an invoice
IJSER
International Journal of Scientific & Engineering Research Volume 13, Issue 5, May-2022 55
ISSN 2229-5518
IJSER © 2022
http://www.ijser.org
2) The invoices are received via email in the particular mail folder.
Figure 5 Invoice mails in the mailbox
3) All Invoices sent to UiPath Robot through email is successfully downloaded to a dedicated folder which is assigned during
development. The robot never redownload the invoice which has been already downloaded previously. These invoices were
downloaded only after they got the green flag from the spam detection python script with 98.6% accuracy.
Figure 6 Successful download and storing of invoices in the current date folder
4) After successfully downloading the invoices, the robot correctly fetches invoices from that folder one by one. After this the
robot digitizes the document for OCR engine to work accurately and then checks if the document is invoice or not and if it is
an invoice, it extracts data from the pdf and stores it in an excel with 100% accuracy.
IJSER
International Journal of Scientific & Engineering Research Volume 13, Issue 5, May-2022 56
ISSN 2229-5518
IJSER © 2022
http://www.ijser.org
Figure 7 : Successful data entry of content in the Excel file
5) After successfully registering each invoice, the software robot is then able to send post notifications in the form of email to
the concerned employee or to the vendor in question.
Figure 8 Successfully notifying the concerned employee or vendor
6) The spam mails are transferred to the spam folder of the mail box.
IJSER
International Journal of Scientific & Engineering Research Volume 13, Issue 5, May-2022 57
ISSN 2229-5518
IJSER © 2022
http://www.ijser.org
Figure 9 Spam folder with the spam mail-invoices
7) The processed invoices are transferred to the processed invoices folder of the mail box.
Figure 10 Processed Invoices folder with the mails and invoices that have been processed by the robot
VIII. SWOT ANALYSIS
Strength, Weakness, Opportunities and Threats, SWOT, is a framework used widely to assess the businesses four pillars.
These four pillars help understand about all there is to be known. This 2x2 matrix, created after brainstorming every block in
relation to our product, will tell us where we stand.
IJSER
International Journal of Scientific & Engineering Research Volume 13, Issue 5, May-2022 58
ISSN 2229-5518
IJSER © 2022
http://www.ijser.org
Figure 11 SWOT Analysis Diagram
IX. CONCLUSION
Automated invoice processing can achieve powerful results for accounts payable departments. Thanks to technological
advancements in robotic process automation and computer vision technologies, invoice processing can eliminate bottlenecks
within the AP process and turn the department into the profit center it can be. Automated invoice processing enables touchless
automation across the entire accounts payable process, and can transform the business in just months, creating a powerful
return on investment. Any organization that receives a large number of vendor invoices on paper can benefit from invoice
processing technology. The more data from each invoice that you are hand- keying into your accounting software that more
benefits you can get from each page you automate.
With the help of the solution discussed above the finance department of any company would be effectively able to prevent from
processing any spam invoices and will also be able to save a lot of time as they will not have to do anything manually. The
solution takes care of filtering the spam/ham mails, and also of extracting the required data and also sending the extracted
results excel to the concerned person. This solution is secure, effective, efficient and accurate.
X. BIBLOGRAPHY
[1] Sagar Sahu, Sania Salwekar, Atharva Pandit, Manoj Patil, “Invoice Processing Using Robotic Process Automation”, 2020,
International Journal of Scientific Research in Computer Science, Engineering and Informational Technology.
[2] Desai, Devanshi and Jain, Ansh and Naik, Dhaivat and Panchal, Nishita and Sawant, Dattatray, “Invoice Processing using
RPA & AI”, 2021, International Conference on Smart Data Intelligence
[3] Harikrishnan NB, Vinayakumar R, Soman KP, “A Machine Learning approach towards Phishing EmailDetection”, 2018,
ACM IWSPA
[4] Jorge Ribeiro, Rui Lima, Tiago Eckhardt, Sara Paiva, “Robotic Process Automation and Artificial Intelligence in Industry 4.0
– A Literature review”, 2020, CENTERIS - International Conference on ENTERprise Information Systems / ProjMAN -
International Conference on Project MANagement / HCist - International Conference on Health and Social Care Information
Systems and Technologies
[5] Ravi Teja Yarlagadda, “The RPA and AI Automation”, 2018, International Journal of Creative Research Thoughts
[6] A. SARAVANAN and S. SATHYA BAMA, “A Review on Cyber Security and the Fifth Generation Cyberattacks”, 2019,
Oriental Journal of Computer Science and Technology
IJSER
