BUSINESS PROCESS AUTOMATION: A GLOSSARY OF TERMS

When it comes to automating key business processes, there

are acronyms to describe practically every aspect: AI, ML,

RPA, OCR, IDP, and the list goes on. Other aspects of these

terms - accuracy, throughput - are understood and used

differently by stakeholders across an organization.

As part of a new wave of automation solutions that are

challenging legacy technology and some of the old definitions

the industry has used for decades, Hyperscience approaches

business process automation with a different lens. A focus

on meaningful automation is what sets the Hyperscience

Intelligent Document Processing [IDP] solution and platform

apart from other vendors. How can a fresh perspective on

automation change the way we think about traditional legacy

solutions and the common terms they employ?

This glossary takes some common industry terms and

updates them with Hyperscience definitions and perspectives

from team experts in order to help you more meaningfully

introduce automation into your organization.

2

Accuracy[Noun] "How close a measurement is to a known or accepted value."1

One large financial services company is hitting 99.5% accuracy

targets across its data extraction processes.

By leveraging the latest in Machine Learning, Hyperscience

allows for accuracy to be prioritized; clients set the software to

their desired target accuracy [based on internal SLAs or other

compliance requirements] and the tech will automate against

it. In addition, the software learns as it goes, looping in humans

when necessary to review and resolve edge cases, which are

used to finetune the underlying models. This decreases manual

oversight and review while continuously refining and improving

our automation levels over time.

What Hyperscience achieves in terms of accuracy goes

far beyond legacy technology like Optical Character

Recognition [OCR]. While OCR typically transcribes text

character-by-character - and measures accuracy accordingly

- Hyperscience measures accuracy at the field level and reads

a page with context, the way humans would. For example,

while a phone number with 9 out of 10 digits correctly

extracted may be considered 90% accurate according to OCR

standards, it's considered 0% accurate by Hyperscience.

3 2ThoughtCo. 2019.

Artificial Intelligence [AI][Noun] A wide ranging branch of computer science concerned with

building intelligent machines capable of performing tasks

that have traditionally required humans

Companies are able to extract information from W-9 forms

automatically using artificial intelligence.

AI is an interdisciplinary science with multiple approaches, but

advancements in Machine and Deep Learning are creating a

paradigm shift. Automation and Artificial Intelligence [AI] are

intertwined.

More practically, as Bernard Marr, a strategic technology advisor

to governments and businesses says, “Generally, people invest in

AI development for one of these three objectives:

1. Build systems that think exactly like humans do [“strong AI”]

2. Just get systems to work without figuring out how human

reasoning works [“weak AI”]

3. Use human reasoning as a model but not necessarily the

end goal.”2

As AI continues to advance, and computers and software

become more intelligent, they're able to execute and imitate

more and more complex - and human-like - functions, delivering

greater automation benefits.

4 2Forbes. 2018.

Automation[Noun] The technique or labor-saving technologies that reduce

human intervention.

TD Ameritrade achieved 98% automation for its institutional

and retail inbound mail processes.

Hyperscience considers an output 100% automated if there is

zero human involvement in processing the input.

By automating tasks and also automating the identification of

tasks that need human intervention, Hyperscience provides

higher quality data. This prevents the usual automation and

accuracy tradeoff of legacy approaches to automation.

For Hyperscience clients, accuracy makes all the difference,

which is why our solution is designed to interface with

humans in meaningful ways that allow for high levels of

accuracy, automation, and efficiency - we refer to this as

meaningful automation.

5

Blocks[Plural Noun] Core components of the Hyperscience Platform that can be

independently developed & deployed to accomplish a single task.

The healthcare institution is able to implement and customize

the specific blocks they need for their medical claims

processing.

6

Blocks separate business logic from function - they are the

LEGO® pieces we give our customers to automate their

workflows. Right now the blocks consists of extraction,

classification, collation, validation, sorting, API and decision.

Each accomplishes a single business function, and some are as

simple as a yes or no to the question of “Is this a W-9?” or as

complex as reading messy handwriting.

Each block’s business function is best accomplished through

ML & HITL together and is manageable by technical and non-

technical user personas.

How the Hyperscience Platform uses blocks to automate a

process input-to-outcome:

INPUT OUTCOMEOut-of-the-box

connectors for data ingestion points [email, shared drive, msg queue,

fax, etc]

Decision kicks off next

claims payment + denial

workflows managed by Hyperscience

Decision

Claims adjudicator reviews context + recommendation

to makes final decision

API

Claims adjudicator

receives notification of claims ready

for review + decisioning

Sorting

Based on extracted

data points, route claims to

appropriate claims adjudicator

Validation

Ensure claim is

consistent, matches internal records and in good order

Collation

Remove

previously received docs to avoid

costly rework + overpayments

Classification

Automatically

classify structured claim forms + supporting documents

Extraction

Extract data

from structured + semi-structured +

unstructured documents

CollationThe grouping and ordering of documents

to facilitate further processing or storage.

Collation keeps shared context across

separate documents within a transaction.

Intelligent Document Processing [IDP]

tools extract relevant information and

then collate that information in order to

process it effectively.

ClassificationThe categorization of documents into

taxonomies. In other words, the software

asks, “What is this?” and then files it correctly.

Sorting automatically grouping documents

or data by similarities, rather than putting

information into predetermined categories.

Hyperscience can sort, classify, and involve

humans in a novel way instead of automating

errors if the machine is not sure.

APICustom block behavior configured

through code. Design and build custom

blocks with maximum flexibility to

support the way you run your business.

Existing Blocks

DecisionBuild decision points into your workflows

and manage the flow of information

based on already processed data and

inputs from external systems.

ExtractionQuick extraction of data regardless of

inbound document type: structured,

semi-structured, or unstructured.

ValidationData fields, context, and entities are

identified, validated against third-

party data or external sources, and

automatically processed according

to your logic.

Fine tuning works like a French press. The more data running

through a solution, the better. Each new piece of data received

allows the platform to become more efficient at extracting

accurate data, while discarding inaccurate data.

Along with accuracy and automation, a third part of

Hyperscience’s philosophy is confidence. A solution's

confidence can be measured by determining how the

percentage of accuracy desired by the customer compares to

the percentage of accuracy from extracted fields.

As it learns, the solution's confidence is improved and calibrated

through Machine Learning. This means that accuracy and

automation are also increased, while the final product is fine

tuned and improved upon.

Fine Tuning[Gerund of the Verb, Fine Tune] The process of improving a software’s performance.

The government agency’s machine learning solution is

constantly fine tuning itself.

9

One large financial services

company reached 99.5% accuracy

after fine-tuning kicked in

Human-in- the-Loop [HITL] [Noun] A description of software processing that involves humans in

automation processes.

When extracting the name field, the software was unsure if it

was reading Max or Nax so it sent the field to a human keyer

for review.

Human-in-the-loop works just like it sounds - humans

are brought into automation processes in order to act as

checkpoints for accuracy or automation. Most processes

that we think of as “automated” have human-in-the-loop

components. How and when humans are brought into the

loop depends on an organization’s automation ideology

and processes.

Hyperscience's approach to meaningful automation brings

humans into the loop in a way that maximizes efficiency and

resources. In the Hyperscience view, the cutting-edge of

automation is reimagining where humans and machines can

work alongside each other more seamlessly.

What does this look like in practical terms? Hyperscience’s

processes involve humans in a novel way that doesn’t straight

through process documents with errors. If the machine is unsure

of its confidence in it is extraction or classification, it will flag a

human to decide what to do; unlocking operational efficiencies

and reducing error rates in the process.

10Increase accuracy with efficient human + machine collaboration.

Get Started With a Demo

Hyperautomation[Noun] The augmentation of automation technologies to increase

the end-to-end output of a process.

By leveraging hyperautomation technologies, one Fortune

500 insurer was able to identify and extract data from multi-

page, hand-annotated invoices with variables such as fax marks

obstructing the document clarity.

Hyperautomation is the application of disruptive technologies

such as Robotic Process Automation [RPA], Artificial

Intelligence [AI], Machine Learning [ML], and Intelligent

Document Processing [IDP], process mining and more, that

when combined, transform business processes and improve

overall quality of work.

By combining hyperautomation technologies with redesigned

operational processes, companies will be poised to reach the

benefits of time and cost that automation was originally set out

to do.

With Hyperscience, companies can implement a platform that

fits into their tech stack and works with a variety of tools that

may already be in place.

11 Learn More About the Hyperscience Platform

Intelligent Automation [IA][Noun] Merging Artificial Intelligence [AI] and Machine Learning [ML]

automation technologies to create rapid end-to-end business

process automation; accelerating digital transformation and

increasing speed and accurary.

By leveraging intelligent automation, a leading financial advisor

group was able to reduce its document processing time 60%.

Intelligent automation [IA] is all about synthesizing various

automation technologies and software in order to streamline

traditional business processes end to end, improve customer

experience and boost employee productivity and satisfaction.

A Deloitte Insights article says intelligent automation,

“Applications range from the routine to the revolutionary:

from collecting, analyzing, and making decisions about textual

information . . . to guiding advanced robots. It is already helping

companies transcend conventional performance tradeoffs to

achieve unprecedented levels of efficiency and quality.”6

Intelligent automation is often used by those in the industry to

distinguish from traditional automation, which is the automation

of any type of repetitive task.

12 6Deloitte. 2014.

Intelligent DocumentProcessing [IDP][Noun] An automation solution that extracts, organizes, and

processes data and other information.

Using IDP, one investment management company was able

to reduce document processing time by 60%.

Everest Group defines Intelligent Document Processing

[IDP] as "any software product or solution capturing data

from documents (including email, text, PDF and scanned

documents), categorizing it and then extracting the relevant

data for further processing using AI technologies such as

computer vision, OCR, NLP and machine/deep learning."7

The Hyperscience IDP offering automates up to 95% of data

capture, classification and extraction with >99% accuracy,

turning low-resolution images, PDFs and messy handwritten

forms into machine-readable data that can be used for faster,

more reliable downstream processing and decision-making.

The Hyperscience difference is to reach high levels of accuracy

and automation because of how the solution improves over

time.

13

Establish a framework to understand Intelligent Document

Processing, evaluate vendors, and select the right tech - and

solution - for your business.

Download the E-Book Today

Machine Learning [ML][Noun] The algorithmic process that allows automated systems

to learn and improve from experience.

Thanks to Machine Learning, the solution continues to learn

on an organization's data to drive lower error rates and higher

automation.

Whereas older technologies rely on explicit rules, the beauty

of Machine Learning is that it trains on real-world data and

continues to learn and adjust itself in response to the data

it’s exposed to. Machine Learning has unlocked capabilities

that weren’t possible before, taking us from a place where we

couldn’t possibly write software to read and account for all

the imperfection and variations that accompany real-world

documents – including the diverse document types and text

inputs it contains – to one that can train and teach itself.

There are two steps to Machine Learning: Collecting and

training the data, and then “operationalizing the Machine

Learning, such as by using it to help provide insights or as part

of a product.”8

Hyperscience takes a novel, ML-driven approach to business

process automation, enabling customers to leverage the

continuous improvement power of ML without data scientists

or engineers. This cutting-edge approach enables greater

automation and accuracy with less human intervention.

14 8Forbes. 2020.

[Noun] Automation software that extracts an image and converts it to

machine-readable text.

With OCR tech, companies can convert computer typed W-2

forms into editable and searchable data.

Optical Character Recognition [OCR] recognizes the text

inside images and converts it into readable data, character-by-

character. There are two types of algorithms that OCR software

can use to recognize text within an image: pattern recognition

software looks for patterns based on examples of text it has

already been given; feature detection software relies on a given

set of rules for each character that enables it to recognize those

characters in a document.

While OCR solutions require certain conditions in order to

deliver accurate results, an IDP solution can handle even

the worst document imperfections - from poorly scanned

documents to hard-to-read handwriting to half machine typed

and half handwritten entries. OCR software only reads what’s

within the box, whereas the Hyperscience IDP offering uses

proprietary ML techniques to know the difference between

what is a response and what should be “dropped out” [like a

field name] and transcribe low-quality documents.

15

Optical Character Recognition [OCR]

Learn what makes Hyperscience a superior IDP solution.

Schedule a Demo

[Noun] Ensuring the accuracy of a software’s automation capabilities

either by humans or automated reporting, or both.

Through constant quality assurance, the software is able to

continue refining the accuracy of its output.

Humans aren’t perfect, and neither are machines, so ensuring

the two parties agree is vital to making high-quality ongoing

improvements and for accuracy reporting. Quality Assurance

[QA] drives the Machine Learning improvements that make

Hyperscience best-in-class and helps the solution learn your

data and maximize automation levels.

Quality Assurance [QA] involves optimizing the solution's

confidence, or how accurate it is according to the user’s

preferences. With Hyperscience’s models, QA would flag

a human to come into the loop and check its accuracy as

depicted below:

16

Quality Assurance [QA]

Quality Assurance[Reporting & ML improvements]

Supervision[Fields the machine wasn't sure about]

Company Firewall

JSONIngestion via API

Downstream System[s]

Asynchronous Synchronous

[Noun] A digital enablement technology that predominantly creates

scripts that automate routine, predictable data transcription

work based on structured inputs.

RPA works well for repetitive, rules-based tasks, such as

copying data from a spreadsheet and pasting it into the correct

database or dragging and dropping files into a folder.

Robotic Process Automation [RPA] works well for rules-based

processes with structured data inputs.

Structured data, however, is often a miniscule portion of an

enterprise’s data stack; the vast majority of information that

enterprises need for business operations resides in documents

with unstructured data that are inaccessible to RPA solutions,

such as images and PDFs. That is where models-based techniques,

like machine learning and other forms of AI, which train on real-

world data and continue to learn and adjust in response to ongoing

information, add tremendous value in document processing. They

accomplish that by unlocking and lifting data contained in various files

and document types with a high degree of accuracy and scalability.

17

Robotic Process Automation [RPA]

[Noun] An approach to automation that enables enterprises to build, run

and manage business processes like software; A new category of

human and machine co-experience.

Through Software-Defined Management, decisions are made

quickly, based on the most accurate data available, and by the

most suitable and knowledgeable resource.

Hyperscience concepted and is a champion of Software-

Defined Management [SDM]. We believe that data should

flow freely between organizations. By incorporating software

development best practices, business processes will become

measurable, scalable and automated by humans and ML

working in tandem. Automation is reinventing processes -

revisiting, rather than replicating.

There are some design principles that guide an SDM approach:

1. It all starts with data entry, step zero of the business process,

when unstructured content is turned into structured data.

2. It's crucial that automation is followed up by tight

integration with human-in-the-loop supervison.

3. Business processes improve continuously and

organizations need to evolve and adapt to different

market dynamics quickly.

Software-DefinedManagement [SDM]

18

[Noun] The compatibility of extracted data in order to use it in

automated processes.

The global insurance enterprise receives a mix of document

structures in its plan request packets.

Structuring: The extraction of data into a normalized, structured

format. Hyperscience is concerned with structuring as it relates

to transcription abilities, regardless of the inbound document

type. There are typically three types of documents that are

fed into an IDP solution: structured, semi-structured, and

unstructured.

Structure

19

Semi-structured:

You know what data you want off the page, but not where it’s going to be. These entities need to be key value pairs like invoice number, pay date, and check amount.

Unstructured:

You don’t know what data you want or where it’s going to be; in other words, everything else.

Structured:

You know what data you want off the page and where it is.

Form W-9(Rev. December 2014)Department of the Treasury Internal Revenue Service

Request for Taxpayer Identification Number and Certification

Give Form to the requester. Do not send to the IRS.

Pri

nt o

r ty

pe

See

Sp

ecifi

c In

stru

ctio

ns o

n p

age

2.

1 Name (as shown on your income tax return). Name is required on this line; do not leave this line blank.

2 Business name/disregarded entity name, if different from above

3 Check appropriate box for federal tax classification; check only one of the following seven boxes:

Individual/sole proprietor or single-member LLC

C Corporation S Corporation Partnership Trust/estate

Limited liability company. Enter the tax classification (C=C corporation, S=S corporation, P=partnership) ▶

Note. For a single-member LLC that is disregarded, do not check LLC; check the appropriate box in the line above for the tax classification of the single-member owner.

Other (see instructions) ▶

4 Exemptions (codes apply only to certain entities, not individuals; see instructions on page 3):Exempt payee code (if any)

Exemption from FATCA reporting

code (if any)(Applies to accounts maintained outside the U.S.)

5 Address (number, street, and apt. or suite no.)

6 City, state, and ZIP code

Requester’s name and address (optional)

7 List account number(s) here (optional)

Part I Taxpayer Identification Number (TIN)Enter your TIN in the appropriate box. The TIN provided must match the name given on line 1 to avoid backup withholding. For individuals, this is generally your social security number (SSN). However, for a resident alien, sole proprietor, or disregarded entity, see the Part I instructions on page 3. For other entities, it is your employer identification number (EIN). If you do not have a number, see How to get a TIN on page 3.

Note. If the account is in more than one name, see the instructions for line 1 and the chart on page 4 for guidelines on whose number to enter.

Social security number

– –

orEmployer identification number

–

Part II CertificationUnder penalties of perjury, I certify that:

1. The number shown on this form is my correct taxpayer identification number (or I am waiting for a number to be issued to me); and

2. I am not subject to backup withholding because: (a) I am exempt from backup withholding, or (b) I have not been notified by the Internal Revenue Service (IRS) that I am subject to backup withholding as a result of a failure to report all interest or dividends, or (c) the IRS has notified me that I am no longer subject to backup withholding; and

3. I am a U.S. citizen or other U.S. person (defined below); and

4. The FATCA code(s) entered on this form (if any) indicating that I am exempt from FATCA reporting is correct.

Certification instructions. You must cross out item 2 above if you have been notified by the IRS that you are currently subject to backup withholding because you have failed to report all interest and dividends on your tax return. For real estate transactions, item 2 does not apply. For mortgage interest paid, acquisition or abandonment of secured property, cancellation of debt, contributions to an individual retirement arrangement (IRA), and generally, payments other than interest and dividends, you are not required to sign the certification, but you must provide your correct TIN. See the instructions on page 3.

Sign Here

Signature of U.S. person ▶ Date ▶

General InstructionsSection references are to the Internal Revenue Code unless otherwise noted.

Future developments. Information about developments affecting Form W-9 (such as legislation enacted after we release it) is at www.irs.gov/fw9.

Purpose of FormAn individual or entity (Form W-9 requester) who is required to file an information return with the IRS must obtain your correct taxpayer identification number (TIN) which may be your social security number (SSN), individual taxpayer identification number (ITIN), adoption taxpayer identification number (ATIN), or employer identification number (EIN), to report on an information return the amount paid to you, or other amount reportable on an information return. Examples of information returns include, but are not limited to, the following:

• Form 1099-INT (interest earned or paid)

• Form 1099-DIV (dividends, including those from stocks or mutual funds)

• Form 1099-MISC (various types of income, prizes, awards, or gross proceeds)

• Form 1099-B (stock or mutual fund sales and certain other transactions by brokers)

• Form 1099-S (proceeds from real estate transactions)

• Form 1099-K (merchant card and third party network transactions)

• Form 1098 (home mortgage interest), 1098-E (student loan interest), 1098-T (tuition)

• Form 1099-C (canceled debt)

• Form 1099-A (acquisition or abandonment of secured property)

Use Form W-9 only if you are a U.S. person (including a resident alien), to provide your correct TIN.

If you do not return Form W-9 to the requester with a TIN, you might be subject to backup withholding. See What is backup withholding? on page 2.

By signing the filled-out form, you:

1. Certify that the TIN you are giving is correct (or you are waiting for a number to be issued),

2. Certify that you are not subject to backup withholding, or

3. Claim exemption from backup withholding if you are a U.S. exempt payee. If applicable, you are also certifying that as a U.S. person, your allocable share of any partnership income from a U.S. trade or business is not subject to the withholding tax on foreign partners' share of effectively connected income, and

4. Certify that FATCA code(s) entered on this form (if any) indicating that you are exempt from the FATCA reporting, is correct. See What is FATCA reporting? on page 2 for further information.

Cat. No. 10231X Form W-9 (Rev. 12-2014)

[Noun] The measure of a machine’s end-to-end average handling time.

By automating their insurance applications process,

ONE Insurance has attained more than 2X

throughput improvements.

Hyperscience is concerned with throughput as a measure of

average handling time, or the rate at which something goes

through a process. In the case of IDP, throughput can be

measured by how long it takes to process a document. As

accuracy and automation increase, throughput improves.

Throughput

20

[Plural Noun] How blocks are connected together to connect information.

The financial services institution customized a workflow for their

new customer onboarding process.

Workflows are the flow of information - they are what our

customers configure to automate a business process. For

example, IDP is a single workflow. Workflows don’t exist in

isolation, but rather they’re stitched into external and internal

systems. Each has at least one input and one output to an other

systems or another workflow.

Workflows

21For more information, please visit: www.hyperscience.com

Start operating more efficiently and break away from

your competition with Hyperscience's cutting-edge

approach to automation.

Contact Us to Get on the Right Path Today

© Hyperscience, Inc. 202022