eB O O K

Intelligent Document Processing:

A guide to using AI for document review, information
extraction and data analysis
Table of Contents

1. Executive Summary

2. What is Intelligent Document

Processing, and how can it
help your business?

3. AI and human synergy:

The value of combining Intelligent Document
Processing with subject matter experts

4. Industry Adoption
Who is using Intelligent Document Processing
and how

5. Who to Trust
How to choose an Intelligent Document
Processing provider that’s right for your business?

6. Straight-through Processing:
Integrating Intelligent Document Processing for
maximum efficiency

7. How to Scale:
Leveraging Intelligent Document Processing
technology across your enterprise

8. Conclusion

2
EX EC U T IV E S UMMA RY

As the world becomes increasingly digital and as businesses have had to assess emerging risks,
reliant on technology, there has never been more identify new opportunities and find alternate ways of
data at our disposal or challenges around analyzing working.
it. Data is a precious commodity, and companies
worldwide in every industry are grappling with More than ever, businesses need an in-depth
how to unlock the power of their data in ways that understanding of critical information trapped
enable their human workforce to reach new heights in their documents to effectively manage risk,
of success. Artificial intelligence (AI) plays a crucial seize opportunities, automate processes and scale
role in meeting this challenge providing businesses operations. Using machine learning techniques, IDP
with a tool to help them locate and analyze the learns from experience, adjusts to new inputs and
right information and use the resulting insights to performs the document analysis faster, and cheaper
improve decision-making. than humans can and with fewer errors. It provides
a repeatable and scalable solution to a myriad of
Reviewing, processing and understanding unstructured data and document-related problems,
the information within your documents is an making it an essential component in the tech stack
incredibly challenging endeavor that AI can for companies implementing digital transformation
solve. Many organizations need a better way initiatives across their enterprise.
to extract value from their documents. Experts
estimate that approximately 80% of a company’s In the following pages, we demystify what IDP is,
data is unstructured: written text locked inside explain how it can increase efficiencies across
unstandardized files. By extracting, analyzing, your business. We share examples and use cases
organizing and interpreting data buried in digital and to demonstrate how the technology works and the
paper documents, Intelligent Document Processing results clients have seen across various projects and
(IDP) can transform how we review, process, and processes. We provide a guide to selecting an IDP
ultimately gain valuable insights from documents. provider, including the questions to ask and what
This technology is being adopted by organizations to look out for in terms of red flags. We also cover
across sectors who need to understand and analyze the subject of straight-through processing before
diverse types of documents at scale. And the wrapping-up with how IDP can accelerate digital
events of 2020 have accelerated the adoption rate transformation initiatives across your organization.

3
2. What is Intelligent Document
Processing, and how can it
help your business?

In every legal TV drama, there’s a scene where a crack

team of lawyers arm themselves with yellow highlighters
and scour boxes of documents around the clock in
search of a specific piece of information. After spending
several sleepless nights at this task, someone finds
the proverbial needle in the document haystack that
eventually wins the day. While this tension might be
entertaining to watch, the reality for those that have
experienced it is much less fun. It’s also highly likely
that the crack team of lawyers didn’t find all the salient
information in just a few days/nights. Sorry if we just
ruined the illusion, but the process of manually reviewing
reams of documents is incredibly slow and inefficient. It’s
also expensive and prone to error.

In the real world, most documents now exist

electronically and, more importantly, technology has
evolved to eliminate the need for manually reviewing
documents altogether. This technology is called
Intelligent Document Processing (IDP), and it can be
used in any industry that needs to transform data that
is siloed and unstandardized into a structured, usable
format at scale. IDP enables organizations to quickly
and precisely extract and classify data whether its legal
paperwork or financial reports as well as a myriad of
different document types so they can make smarter,
faster decisions.

But what exactly is IDP technology?

Let’s start with the ‘intelligent’ component of IDP Machine learning (ML) is the process of using
which comes from Artificial intelligence (AI). patterns in data to ‘teach’ the machine, so its
AI is the use of any computer method to ‘train’ performance and predictions become more effective
machines to mimic human intelligence so they and accurate over time. Natural language processing
can complete repetitive or complex tasks for us or (NLP) is the branch of AI focused on leveraging ML
predict outcomes. AI is already present in our day- techniques so the machine can understand and
to-day lives powering things like auto-correct on our interpret human language. The diagram below
phones, ride-hailing apps like Uber and Lyft, online shows the relationship between AI, machine
purchase recommendations and facial recognition learning and NLP.
on Facebook.

4
 ARTIFICIAL INTELLIGENCE:
Using machines to complete complex tasks. Any technique that enables computers
to mimic human intelligence. Example: self-driving cars, AlphaGo

MACHINE LEARNING:
Statistical techniques that enable machines to improve tasks
with experience by learning from and making predictions based
on “patterns. Get computers to act without being explicitly
programmed. Example: recommendation engines

NATURAL LANGUAGE PROCESSING:

Machine learning techniques to understand and
interpret human language. Example: chat bots,
document review

Figure 1: The relationship between AI, ML and NLP

In the case of IDP, ML and NLP are used to train platform sits on top of the technology enabling users
a computer to simulate a human subject matter with no previous AI, ML or NLP experience to quickly
expert’s review of a set of documents. The result is train the computer to extract the specific data they
a computer capable of ‘understanding’ the contents need from different document types. This is what’s
of documents, including the contextual nuances commonly known as ’no-code’. The platform puts
of the language within them. The technology can the power of machine learning in the hands of non-
then accurately extract information and insights technical teams and users like lawyers, analysts,
contained in the documents as well as categorize operations and accountants.
and organize the documents themselves. An IDP

Why might your business need IDP?

We are six decades into the information age, and, expensive and inefficient and many businesses have
by any measure, data has become a competitive reached a tipping point, where the scale, complexity
currency in business. Yet, experts estimate that and speed of incoming data exceeds the capacity
more than 80% of a company’s data is unstructured. and cost of manual processing.
This data typically takes the form of written text
containing dates and numbers located in reports, At Eigen, we help organizations with many different
contracts, memorandums, agreements, emails etc. document and data-related challenges. For some
Manually collating this information requires people clients, it’s the capability to analyze previously
who are familiar with the data to spend their time inaccessible data so they can optimize their capital
searching for it and preparing it to be analyzed or prevent losses. For others, it’s the ability to
before the actual analysis can be performed. automate existing manual document processes
Companies operating this way are underutilizing relating to client onboarding and operations so they
their most important and expensive resource – their can do more business. Or for some, it’s being able
human talent. to handle one-off regulatory exercises or ongoing
record-keeping and reporting requirements without
IDP provides organizations with a better way having to hire in lawyers or consultants to review all
to extract value from their unstructured data. the relevant paperwork.
Repetitive, manual document processing is

5
How does IDP work

Most businesses are inundated daily with data, Clients can also choose how much support they
with it coming in from a variety of sources, but want and can either have Eigen train their ML
implementing another system may seem daunting. models or do it themselves. Non-technical users
Eigen’s no-code IDP platform is easy to use and with no previous ML experience can easily handle
can be fully integrated with upstream document this process; all they need is an understanding of the
repositories and downstream databases to create documents and data they need to extract. Here’s a
seamless processes. step-by-step guide on how IDP works:

1. Scope and Agree Requirements 2. Upload and Label Documents

Eigen works closely with the client to understand Documents are uploaded to the platform, and the
their document challenges, their data needs, their relevant data for extraction is labeled by either Eigen
internal capacity and overall goals. Based on this (full-service) or the client (self-service) to train the
information, Eigen will recommend a service package machine learning models. Logic can be applied to
with deployment and client support options to meet the output answers based on the information located in the
client’s specific requirements. documents. However, if a client wants to get started
immediately, they can choose to use Eigen’s existing
warmed-up models and readymade question templates.

3. Machine Learning Model Created 4. The Documents are Analyzed

Using the labels, the machine learning models are The models analyze all new documents to retrieve the
automatically built to meet the client’s specific correct data points and provide answers to questions.
requirements. The models are flexible and can be trained The platform guides users to verify any low-confidence
and re-trained to provide additional data points or answers, which is used to further improve accuracy levels.
answers at any time by repeating step 2 noted above.

5. Data Extracted and Exported

The extracted data is exported or sent directly to other
systems via APIs. Legacy or new documents can be
processed through the platform as often as required.

For those new to IDP technology, it’s worth noting that it requires
only a few documents to train machine learning models. This
enables clients to gain value from the platform and be able to
answer questions about their specific data rapidly.

6
3. AI and human synergy:
The value of combining Intelligent Document
Processing with subject matter expertise

Since the Industrial Revolution in the 1700s, machines

have been introduced into the workplace to increase
productivity. Back then, it was the invention of steam
power that was the catalyst for change. Goods previously
made in small batches by hand could be mass-produced
in factories through the implementation of steam-
powered mechanized production. Fast forward to today,
and it’s AI that’s transforming industries around the
world. Solutions powered by AI are changing the way
people work across sectors, making them more efficient,
effective and able to focus on higher value work through
the automation of previously manual tasks.

Say goodbye to costly manual document review

There are many reasons why organizations need experience of a consultant, then those costs may
to gather and analyze data from documents such be excessively high. Thankfully manual document
as contracts, reports, memorandums and policies. review is no longer a necessity, and IDP technology
These include risk analysis, regulatory changes, can do the heavy lifting in terms of getting to the
cash flow or capital optimization, litigation, due facts and answers faster and cheaper. It’s now
diligence, issue handling, financial reporting and possible to extract information and gather insights
industrializing information retrieval for better from documents on an industrial scale and create or
planning, decision-making and service delivery. enrich databases to give you full data transparency.
In the past, when businesses needed to access and And because the relevant subject matter experts
understand information buried in their documents, train the underlying machine learning model, the
the only option was to read and review their IDP platform systematizes and augments their
contents manually. The nature, scale and deadline of knowledge in the process.
the exercise would dictate the resources required to
complete the review. If you had an internal team with
the right specialist knowledge, you could have them
review the documents; otherwise, you’d be compelled
to hire temporary staff or outsource the work.

Manual document review may not be too onerous or

expensive if you’re dealing with a small number of
documents but if you need to review 100s or 1,000s
of them, the hours and costs quickly add up. And
if the nature of the review requires the judgement
of a lawyer, the critical eye of an analyst or the

7
Say hello to repeatability and increased productivity

By using the Eigen platform, clients can scale the need remediating or repapering following a change
specialized knowledge of one or more experts in standards or regulations. In both instances, the
across entire teams. For example, a credit analyst Eigen IDP platform produces a comprehensive set
at a bank can train Eigen based on their expertise of data for 1000’s of documents utilizing the specific
to provide data and answers to questions that are knowledge of the relevant subject matter experts in
vital to monitoring the performance of loans across a fraction of the time it would take to do so manually.
an entire portfolio. Or a paralegal can train the The diagram below shows the typical workflow using
platform to identify and triage all contracts that a solution like Eigen.

Example workflow steps

(1) A subject matter expert determines the data and (5) Once production ready, all the documents are uploaded
answers required. to the platform.

(2) A sample of documents are collated and uploaded to (6) Machine processes all the documents and flags any low
the platform for training purposes. confidence answers for human review. Subject matter
expert reviews and corrects responses.
(3) A subject matter expert trains the platform using the
results of steps 1-2. (7) Machine outputs the data and answers from each of the
documents.
(4) A subject matter expert verifies the training results. If
results need improving, the subject matter expert (8) Data can be exported directly from the platform or sent
provides more training examples until the performance via APIs to other systems or users.
improves sufficiently to move to production.

Figure 2: Example Intelligent Document Processing workflow

Combining advanced natural language processing new documents and data requirements giving you
(NLP) technology with subject matter expertise gives a flexible and scalable solution that can be applied
organizations the ability to tap those knowledge to multiple use cases, further increasing the return
sources in a highly efficient way that minimizes on investment. Companies can now afford to
the human effort. By reducing the human effort, extract and organize data from all their documents
you take out considerable cost, especially in to give them a near real-time view of previously
those circumstances where you would typically unknown risks and opportunities. But is the quality
employ external experts to manually review your and accuracy of data the trade-off when using IDP
documents. The platform can be trained to handle technology versus human manual review?

8
Achieving better data accuracy with As the saying goes, computers are only as good
machine plus human as the data we put into them. They are not 100%
accurate. But then neither are humans. Humans are
not built for repetition and multitasking. When we
perform repetitive or complex tasks that require the
same parts of our brain over and over, we become
stressed, and consequently, our performance suffers.
We’ve witnessed this first-hand by conducting
large-scale exercises in which our platform
reviewed hundreds of thousands of documents and
benchmarked them against human performance.
The exercises showed that human accuracy usually
tops out around the 80% mark.

AI doesn’t suffer from the same cognitive

impairment that we do when completing repetitive
tasks. Conversely AI produces better results with
more use, ‘learning’ as it processes. By using cutting-
edge NLP and machine learning (ML) techniques,
Eigen can achieve results way above the 80%
accuracy threshold achieved by humans and at a
much greater speed. The AI will also ‘own up’ to not
knowing something and will flag data points and
answers as ‘low confidence’ when it’s unsure. These
flagged responses can be easily checked and, if
necessary, corrected in the platform. This flagging
and follow-up process produces results that are more
accurate than human or machine alone. In fact,
when a human subject matter expert performs this
final review step, the results achievable are closer to
the 100% accuracy mark.

AI is transforming how we work today, enabling

businesses to scale-up operations, automate
processes and capitalize on new opportunities as
steam-power did back in the 18th century. In the
case of IDP, it’s possible to significantly reduce the
time, human effort and cost involved in reviewing
and extracting information from documents. And as
a result, your experts can spend more of their time
on crucial higher-value work.

9
4. Industry adoption:
Who is using Intelligent Document Processing and how?

Companies have more data at their disposal than ever The platform is capable of processing virtually any text-
before, and IDP is becoming the go-to solution for firms heavy document and gathering data points for any use
who need to understand and analyze diverse types of case or user. Typical use cases include contract, policy
documents at scale. But who is benefiting from using this or deal onboarding and servicing, compliance, regulatory
technology and how? reporting, capital optimization and risk analysis. By
automating processes and leveraging data previously
Any business in any sector that handles large volumes unexplored, clients spend less time and money on these
of documents can benefit from the technology, but the tasks, freeing up their resources to tackle other work
earliest adopters have been in the contract-heavy world resulting in a significant return on investment.
of financial services where the cost and implications of
inaccurate, incomplete and out-of-date data can be huge. We’ve compiled a few examples below.
These firms need a solution that automates cumbersome
and costly manual processes as well as extracts and
structures data they’ve been unable to access and analyze
in the past.

Regulatory Compliance:
A global bank saves time and money by automating the identification
and routing of contracts impacted by a change in regulation

A multinational bank needed a quick, accurate and platform successfully extracted all the relevant
cost-effective way to identify credit agreements information which amounted to 7.5k data points and
referencing the LIBOR interest rate benchmark delivered them to the Front Office team to form their
scheduled for retirement at the end of 2021. The remediation plan.
bank’s back book extends to thousands of loans, and
each agreement needed to be reviewed, which was The net result of using Eigen was a reduction in
taking, on average, almost 70 minutes per document costs of 60% and a time saving of 75% as well as
to complete manually. an increase in data quality when compared to
manual review efforts. Eigen’s team of Legal Product
A large-scale regulatory exercise like this would Consultants carried out the model development
previously have involved a team of highly skilled and training, so the bank’s teams could focus on
and highly paid in-house specialists or external other priorities. It’s worth noting that bigger cost
consultants. On this scale, it would have been a savings could be realized by performing the machine
prolonged and expensive exercise to carry out learning model training in-house, a task easily done
manually yielding somewhat inconsistent and error- by non-technical users providing they are familiar
prone results. The bank used the Eigen platform to with the documents and data requirements. The
extract answers from all their loan documents to bank is now using Eigen to automate other processes
identify which ones needed remediation as well relating to loan operations and fulfilment to drive
as determine the relative priority of those changes further efficiency gains and cost savings.
and flag what exactly needed to be amended. The

10
Onboarding and operations:
A global investor makes revenue sooner by automating analysis, data
entry and cross-checking processes

The Operations team at a large, global investment Pleased with the results of the pilot and the
firm was overwhelmed with the volume of new platform’s ease of use, the firm has integrated Eigen
deals. Their manual processes were proving to be into their workflows for processing new deals.
inefficient and scaling them to meet increased
demand wasn’t cost-effective or desirable. The The efficiency gains using Eigen have been
rigorous review procedures and the inconsistent significant, reducing the average transaction
nature of the paperwork meant that processing processing time by two-thirds. The Eigen platform
a single transaction could take many hours and has enabled them to automate several processes,
involve several different people. The firm wanted including document analysis, data entry, cross-
to use AI to automate these processes but needed a checking and eliminating the need for multiple
solution that was easy to implement, maintain and approval steps. The Operations and Treasury teams
use so they wouldn’t need additional IT support or are now able to handle the busy periods around
in-house machine learning experience. month-end as well as other peak times, closing
deals faster to bring revenue in sooner. The firm is
During an initial pilot project, the client used Eigen to now using Eigen for a regulatory compliance use
accurately extract 50 key terms from a batch of 100 case, helping them to manage the document review
loan documents. The accuracy of the data compiled process in-house so they can avoid the cost of hiring
by the platform was on a par with levels attained by lawyers or consultants.
the team through manual review and processing.

Capital Optimization:
A FTSE 100 company benefits from lower capital requirements by
identifying eligible bonds using data automatically compiled

A large asset management firm managing assets on of eligibility for assets and compile the necessary
behalf of an insurance client wanted a scalable tech- data to demonstrate compliance as well as perform
led solution to determine which bonds within a pool benefit calculations. The actuaries defined the
of 1k were eligible for capital relief under Solvency information required to assess eligibility as well as
II rules. They wanted to speed up the process of the data required for the later stages of the process.
collating the relevant information to identify the They were able to quickly train the platform using a
assets meeting the necessary criteria. This involved sample of documents to analyze their contents and
reviewing product prospectuses and associated pull out those that matched the eligibility criteria
publications to pull out the correct data on asset and extract the required data. In the diagram below,
characteristics. They wanted a repeatable solution you can see an example of the results table showing
that leveraged the knowledge of experts and the the answers the platform retrieved as well as those
power of AI. requiring further attention by the actuary.

Applying Solvency II rules is a complex process

requiring the specialized knowledge of an actuary.
The asset managers engaged a consulting firm to
help them with their implementation project who
partnered with Eigen to automate the assessment

11
Figure 3: Example results table showing answers retrieved by the platform

Using the Eigen platform, 3k documents were leveraged the consultant’s expert knowledge and
analyzed during the initial implementation project, captured that within the platform in a way that can
and the assessment of those eligible under the be reused for other projects and clients in the future.
rules was successfully automated. Additional data
was compiled and exported for calculation and The beauty of intelligent document processing
compliance purposes. Because the platform flagged is its flexibility and scalability. There are almost
instances when it was uncertain if it had found the a limitless number of processes, workflows and
right answer and those data points were reviewed, manual tasks involving document review and data
checked and corrected by an actuary the quality compilation that the technology can automate or
of the results were far higher than those typically support across sectors and functions. Is your team
achieved through manual review. The results were spending time manually reviewing documents? Or
not only more accurate, but countless hours were are you employing lawyers, consultants, analysts or
saved in actuarial time required to complete the temps to tell you what’s in your contracts? If so, it’s
assessment. The end client, the insurer, is now able time to consider using an IDP platform like Eigen.
to claim capital relief. And the asset manager has

12
5. Who to Trust:
How to choose an Intelligent Document Processing
provider that’s right for your business?

With so many IDP, data extraction and contract analytics

platforms available – and new ones emerging all the
time – the landscape can be confusing. Many companies
looking for a better way to access and analyze large
amounts of data from documents are unsure how to
choose a solution.

In this section, we’ll guide you through the questions to ask

potential vendors and give you some tips on where to steer
the conversation to help you determine which solution
best meets your needs.

Vendor Question 1:

What can you tell me about the data accuracy

that you deliver?

Understandably, this is usually one of the first At Eigen, we deliver market-leading accuracy rates
questions we get asked by prospective clients. After typically of more than 90% because we apply the
all, improving speed and accuracy are the main best elements of the three approaches to machine
reasons, along with cost reduction, that companies learning to drive the best outcomes. In the diagram
turn to us for their data automation needs. First, be below, you can see our three-pronged approach to
wary of anyone claiming 100% data accuracy out machine learning. Our platform also guides users to
of the box. Unless your documents are completely review low confidence answers providing a feedback
standardized, and you only need to extract loop that ensures the model is continuously learning
quantitative data from them, 100% accuracy from the on the job.
outset is unrealistic.

To optimize for accuracy, a machine learning model

needs to be trained based on your specific use case,
documents and data requirements. So, the answer to
your data accuracy question will differ by use case
and be dependent on factors such as variation in
documents and the complexity and granularity of
data required. The provider should be able to share
with you the accuracy levels they’ve achieved for
clients with the same or similar needs as yours. The
accuracy rate should come with some caveats on
the volume of training documents and the length
of time typically required to achieve that level of
performance.

13
 BETTER TRAINING BETTER MACHINE
BETTER FEATURE SCIENCE
WITH MORE DATA LEARNING ALGORITHMS

• Eigen leverages unsupervised and • Eigen has developed a deep • Eigen has a carefully curated set of
transfer-learning representations of mathematical understanding of features for its use cases.
language. multiple algorithms to unlock their
• Eigen also leverages proprietary
full potential.
• It learns inherent language and features such as advanced fuzzy
domain-specific features from large • It uses a fully automated model matching.
data sets and open source libraries. selection.
• As well as topological data analysis.
• These feed into a supervised • Dynamically employing the best
learning model to supplement the algorithm for the job in hand.
client’s limited data to solve their
specific use case.

Figure 4: Eigen’s three-pronged approach to machine learning

Vendor Question 2:

How can your IDP solution meet my

exact needs?

When sourcing any new technology, your starting If you’re in the market for an IDP platform, you
point is identifying the problem that you need to probably want to create efficiencies, drive down
solve as well as the people, processes and systems costs, automate and accelerate processes or analyze
that are impacted by those changes. A provider can’t previously inaccessible data. To achieve these goals,
address your needs if they don’t understand them you need a solution that can be finely tuned to your
so share with them as much information as possible exact needs but is also highly adaptable and able to
about your specific use case (with an NDA in place cope with evolving change. If you need the platform
if required). But no matter how thorough you are in to work with other systems and repositories, make
gathering and prioritizing your requirements, one sure you choose a provider who has a proven track
thing is inevitable; they will change. record in integrations and can offer APIs, plugins or
other options.

• Ask any potential provider what their process is for Eigen has successfully delivered customized
handling change requests. solutions and developed integrations for clients
• Ask them how often they launch new platform across a diverse set of use cases ranging from
features and functionality and how these are one-off regulatory exercises to full-scale digital
prioritized and rolled out to clients. transformation projects. Eigen’s models are purpose-
built to order to ensure they meet the exact needs of
• Ask them how they work with internal teams, third
each client. This bespoke approach makes it easier
parties and integration partners.
for the client to manage and tweak the model as they
• Ask them how easy it is for clients to scale their are familiar with its set-up and its data output.
platform usage up or down.

14
Vendor Question 3:

How do you ensure my documents and data

are secure within your environment?

Another common concern for clients is the security substantiate their claims.
of their data. And rightly so given the sensitivity
of their information, their customers’ expectations Don’t let the vendor blind you with science or
and the risks and regulations surrounding data baffle you with jargon. Information security is as
handling. If you’re entrusting a third-party with your much about processes and people as it is about
documents and data, you need to ensure they are authentication, encryption and data replication.
taking that responsibility as seriously as you do. You want to work with a provider who is clear,
and upfront about their solution and their working
When you ask the question ‘how do you ensure practises. If the vendor is offering a cloud-hosted
my documents and data are secure within your solution, find out who manages that service and
environment?’ the response should cover a broad which party is responsible for security.
spectrum of measures. These should include
platform access and security, data transmission and Eigen holds four ISO certifications, and we are
storage procedures, ongoing monitoring, reporting, subject to frequent independent audits to ensure
testing, and staff training. Ask them if they hold we meet the necessary standards of information
industry-standard certifications and undergo security management and business security possible
independent security audits. And don’t be afraid to to retain these certifications.
ask the vendor to provide details of their information
security credentials (e.g., ISO certifications) and
related policy documents (e.g., data protection,
disaster recovery and business disruption) to

Vendor Question 4:

What support and training will you provide?

Implementing any new technology or system takes the vendor to achieve the results and accuracy rate
time, effort and money and can be a daunting you’ve previously discussed.
prospect. In the case of AI, it’s also disruptive by
design. You’re not just replacing a legacy system However intuitive and easy to use the platform
you’re automating processes and changing appears you should be offered training, so your team
workflows that have been developed over time. Don’t understands how to develop models for your specific
underestimate the effort involved in implementing use cases. Beware of vendors who provide a self-
IDP and similar solutions successfully in ways that service solution without any support - you won’t
maximize your ROI. That said the platform itself achieve the best outcomes without training. Ideally,
should be intuitive and straightforward to use once you should be provided with a dedicated account
you’ve received training. manager, around-the-clock technical support and
training materials.
The exact services and support you require from a
vendor will be dependent on the scope and scale Eigen offers its clients a range of additional
of the implementation project. The complexity of technical and professional services including APIs,
integrations with other systems and your hosting plugins and model development as a service to
requirements will also have a bearing on the ensure they get the most out of the platform while
workload. The vendor should provide you with a only paying for what they need. All Eigen’s clients
delivery plan alongside their proposal that sets out have a dedicated Customer Success Manager, receive
the requirements and responsibilities on both sides regular platform upgrades and ongoing maintenance
with an indicative timeline to production. Be sure and support.
to clarify what’s required of your internal teams and

15
Vendor Question 5:

How long will it take to get up and running with

the platform?

There’s no simple answer to this question as the warmed-up models for specific use cases to
timeframe will vary depending on several factors give clients a head start. These models are built
including the variation in documents, the number based on the typical document structure and data
of data points and answers required, the complexity requirements for a use case and can significantly
of integrations with downstream and upstream reduce the time needed to train a model from
systems, the desired data accuracy etc. But this is scratch. However, the warmed-up model will need
not a trick question. A provider should be able to additional training to deliver a client’s specific
give you a relatively accurate estimation based on requirements and should not be considered a
their knowledge of your use case and experience substitute for developing their own model.
of implementing similar-sized projects for other
clients. Eigen’s IDP platform can be trained using very few
documents. It can take as few as 2 to 50 documents
The pre-production implementation phase can run to successfully train the platform while our
from a few weeks to a few months but should be competitors often require hundreds or thousands.
phased, and project managed, so you know where We’ve developed cutting-edge machine learning
you are in the rollout process at any given time. techniques that help us to significantly reduce the
Many providers, including Eigen, have developed time to value for our clients.

Always request a demo When you’re evaluating potential vendors asking the
right questions is only part of the process. Always
request a custom demo using samples of your own
documents or those of a similar length and format.
Be as specific as you can be about the data points,
sections and answers you want extracted from the
documents so the demo can include these and give
you a better sense of the work required to get you the
results you need.

16
6. Straight-through processing:
Integrating Intelligent Document Processing for
maximum efficiency

For use cases such as compliance or regulatory reporting

that require total precision, IDP can achieve up to 100%
accuracy when machine learning techniques are combined
with human review. But, for use cases where a small
number of errors can be tolerated, or stringent quality
control and model risk management protocols are built-
in, human review can be reduced or even eliminated to
complete the task even faster and cheaper. This is known
as straight-through processing (STP).

The origin of straight-through processing

STP methodologies were initially developed to telegraphic system. Beyond payments, STP has
automate electronic payment transfers, reducing revolutionized securities trading by automating and
the need for manual intervention and making digitizing previously complex and cumbersome
them virtually instantaneous. Before STP, sending manual trade settlement processes. Before STP, to
money involved multiple departments, both on the settle a trade, phone calls, emails and paperwork
initiation and receiving end of the transfer, and needed to be exchanged between the buy and sell-
took days to complete. The old process required side. The information would be re-entered and cross-
data to be rekeyed and rechecked, which was time- checked, and the trade wasn’t complete until receipt
consuming, prone to error and made transactions of the certificate and payment were confirmed by
more susceptible to fraud. With the advent of STP, both parties. Following the old manual process, it
the entire process was automated, eliminating would take up to 10 days to settle a trade whereas
duplication of effort, reducing the risk of errors and STP reduced the average to just 2-3 days.
fraud, accelerating the turnaround time and lowering
the processing costs. STP has many uses and can be applied to any
transaction, settlement process or financial product
STP development started in the early 1970s when the to eliminate the bottlenecks and costs inherent in
banking industry first began integrating computers human-led processing. Many banks and financial
and computer programming into their daily service providers use STP across multiple processes
operations, enabling banking payment transfers and functions to achieve paperless offices.
to be upgraded from a much slower, less secure

The role of IDP in straight-through processing

At the heart of STP is the near real-time transferral internal systems to help manage the loan over its
of information between systems and parties, using entire lifetime. This eliminates inefficient manual
technology as the enabler. STP is now the norm for processes and the risk of human error.
payments, but it’s also been applied more broadly
to repetitive processes that are transactional STP is beneficial for automating steps in long
and can be automated to drive efficiencies. For processes where information would otherwise be
example, IDP can remove the need to manually rekeyed between systems to be used by multiple
review new loan agreements and extract the salient teams. IDP can meet different functional teams’
information required to complete the deal and feed information requirements and gather insights

17
and transform data to eliminate steps in various Analyzing documents manually is time-consuming
processes. Wherever you have sets of documents and expensive, especially for organizations with
that are subject to periodic review, require some large volumes of documents to review regularly.
form of follow-up processing or where events can Ultimately, the goal of STP is to make end-to-end
trigger the need to dive into individual contracts, IDP operations less dependent on human intervention
can help. Using a technique called point extraction, and speed up processing times. IDP can play an
the technology pinpoints and extracts the exact integral role in the straight-through processing
word, number, phrase or sentence that each user or of documents. By extracting and compiling the
team in the process might need in the future. This necessary data and sending it to other systems, the
technique outputs precise results that can be used human effort required in processing your documents
as-is or can form the basis for further analysis that is minimal.
the platform can be trained to perform automatically.
By applying logic, IDP can interpret the extracted
data and provide actionable insights and answers to
complex questions.

The benefits of straight-through processing

your documents

IDP makes it possible to automate document systems and processes. Once you’ve identified all the
analysis with rapid and precise results, so your touchpoints, you can gather requirements from all
people don’t have to scour them repeatedly. The impacted parties and factor their needs into the data
Eigen platform can extract dozens of key terms flow to ensure you achieve the highest possible level
from hundreds or thousands of documents creating of automation. Where data formats are inconsistent
significant efficiency and reducing processing time across systems and information would usually be
and costs. It can automate processes to smooth the rekeyed, we’ve developed customized plugins to
transfer of data associated with document analysis, transform it before it’s sent to other systems via
data entry, cross-checking and eliminate the need APIs. For example, date formats can be located
for multiple approval steps. As a result, clients report and automatically reconfigured for international
a range of benefits including closing deals faster to transactions (e.g., changing the date format from the
bring in revenue sooner, saving on outsourcing costs, US to UK standard).
enabling teams to focus on higher-value work and
the ability to make better more timely decisions and The diagram below illustrates how IDP integrates
action requests faster. with other systems within an organization.

To achieve maximum efficiency, you should identify

every touchpoint over the document’s lifespan,
including all upstream and downstream users,

Figure 5: How Intelligent Document

Processing integrates with other systems
within an organization

18
Little-to-no human intervention

Thanks to STP methodologies and IDP, entire Gathering, analyzing and manually rekeying data to
processes can be automated, and the extracted data process documents is both ineffective and expensive.
can be integrated seamlessly with internal systems. Using IDP, you can empower your team to spend more
This enables documents to be processed with little- time on tactical, revenue-generating tasks and less
to-no human intervention. In addition, clients that time on time-consuming manual review procedures.
use Eigen for STP purposes have full visibility of STP holds the key to achieving maximum efficiency,
the documents being processed and the data being and you can harness its power through IDP for your
extracted so they can review and audit the results at information management needs.
any time. They remain in total control even when the
machine is doing all the work.

19
7. How to scale:
Leveraging Intelligent Document Processing technology
across your enterprise

While companies are always facing external challenges,

the past year or so has been especially formidable. The
global coronavirus pandemic has changed how many of
us work and has increased the number of economic risks
for many businesses. External pressures can quickly
change business priorities, and one way to manage these
more effectively is with technology. Businesses have
faced many challenges, including temporary closures
and supply chain restrictions due to the COVID crisis as
well as new regulations and rules. These issues have
triggered a renewed focus on digital transformation.

Many businesses are trying to cut costs, and these cost-

saving measures often result in reduced productivity.
But the savviest firms are embracing technology to help
them through these difficult times. Whether they need
to mitigate risk, manage regulatory compliance, or meet
other requirements, firms must find ways to do more
with less to remain competitive.

External pressures create huge

administrative burdens

The administrative burden for firms is enormous thorough job of compiling information, so you have
right now. Large firms and corporations have the salient data to hand but do it quickly so it can be
thousands of contracts and agreements they may acted upon before the situation changes or worsens.
need to review in relation to recent events and In an environment where you lack resources and
regulatory changes. This puts additional time- budget but need to act fast, manually reviewing
sensitive demands on organizations to analyze their documents is no longer an option. Fortunately,
documents. Data needs to be located and compiled IDP can help alleviate the burdens associated
from each contract, so the impact and business risk with reviewing, evaluating and gathering insights
of changing market conditions and reforms can be from thousands of documents. And once you’ve
accurately assessed and dealt with. successfully implemented IDP for one use case, it’s
quick and easy to apply it to other documents, data
And while external factors are fluctuating and needs and processes.
evolving, so are your data needs. You need to do a

20
Advancing your digital transformation efforts
with IDP

By using technology rather than people to do the specialist approx. 70 minutes per agreement to
cumbersome work of sifting through documents, review manually. It would have been a prolonged
firms are saving a substantial amount of time and and expensive exercise to complete manually.
money. Some of Eigen’s clients report a cost saving Instead, the bank used Eigen’s platform to extract
of 80% or higher compared to the cost of outsourcing answers from all their loan documents. The platform
manual document review. And time savings of up successfully extracted over 7.5k data points with the
to 75% have also been reported. Once clients have exercise completed in 75% less time at a fully loaded
seen the value they get from one application of the cost saving of 60%.
technology, they feel confident that it can handle
other functional areas, use cases and address more After this initial project’s success, the bank used
internal process pain points. In fact, starting with Eigen to automate its loan fulfilment processes
one use case and expanding into others has become to drive further efficiency gains and cost savings.
the norm for our clients. Eigen’s technology is now being rolled out across
other functions and use cases, delivering automation
Most businesses are awash with contracts, of repetitive, costly and human-error prone
agreements, reports and paperwork in general. tasks. IDP sits at the heart of the bank’s digital
With so many documents, the opportunity to solve transformation program powering its information
for multiple business challenges are plentiful. management, data quality, document processing
And those choosing to use AI to automate their and analysis automation efforts. Using APIs and
in-house document review capabilities for one plugins, data is seamlessly extracted, interpreted,
problem area, have taken a giant leap forward with transformed and delivered to systems across the
their digital transformation efforts. For example, a bank. And with the easy-to-use no-code platform,
global bank wanted to automate how they identified new data fields or questions can be added and
and managed contracts impacted by regulatory trained quickly as external pressures and business
change. They had thousands of loan documents to needs change.
review, and each one was taking a highly skilled

Scaling IDP beyond short term challenges

At Eigen, we work with innovation, digital and the correct data is delivered. The platform
transformation and data management teams to help guides non-technical users through the model
them solve the document-related aspects of their training process, which can be handled in-house or
digitization programs and scale data extraction and by Eigen’s Legal Product Consultants if resources
analysis across their organizations. We can help are unavailable. It’s easy to gauge the model’s
identify and assess which use cases, document performance with an in-built model evaluation
types and processes are good targets for IDP and tool showing you the accuracy levels attained for
meet with your engineers and functional leads to each field or question. A screenshot of the model
determine requirements and dependencies. We evaluation tool is shown below. Using the tool to
work with clients to prioritize use cases and plan evaluate the performance enables clients to retrain
the implementation effort to ensure that the time and tweak the results until the desired accuracy is
to value is minimized and that benefits are quickly achieved before moving into full production. Clients
realized. If required, we also have a network of have complete control over the data output so they
delivery partners and consultancy firms to call upon can handle future changes as necessary.
for subject matter expertise.

During the setup and training phase as well as

ongoing post-implementation, we provide support
and advice to make sure requirements are met,

21
Figure 6: Model evaluation enables users to quickly assess the performance of their models

The platform’s flexibility and simplicity and the

transparency of results make leveraging the
technology across multiple use cases easy. New
models can be trained to handle more document
types, and existing models can be updated as data
needs change. The platform can be used to scale
operations, automate processes and drive cross-
functional efficiencies. Companies handling large
volumes of documents have an almost endless list
of possibilities and functional areas across their
enterprise that can benefit from streamlined or
automated processes.

IDP is increasingly becoming part of many firms’

digital transformation tech stack, enabling them to
eradicate their teams’ need to review documents or
manually locate information buried in contracts. As
a result, these firms are reaping the cost and time-
saving benefits of automation of everyday tasks.

22
 8. Conclusion

The Intelligent Document Processing future is here.

According to the market research firm IDC, worldwide
spending on AI and cognitive systems will reach $57.6
billion in 2021. IDC also predicts that worldwide revenues
will surpass $300 billion by 2024, illustrating how AI
technologies are being used to solve problems across the
business spectrum.

The events of 2020 have reinforced the fact that

regulations, market conditions and business priorities
can change quickly and that companies must be
increasingly agile to overcome the next inevitable
challenge. Whether firms need to efficiently process large
volumes of documents, extract critical information, feed
data into other applications or automate cumbersome
processes, there is a renewed need and appetite for using
AI to cut costs and save time.

IDP not only helps enterprises achieve cost savings

but also improves workforce productivity. Repetitive,
manual document processing by experts is expensive
and inefficient. People are the most important resource
for any business. They should be focused on the
highest-value work: making informed decisions rather
than analyzing pages of documents. Automating the
answering of questions to assess risk and opportunities
and power processes is the most cost-effective and
efficient way to free up peoples’ time.

As enterprises seek to improve compliance, meet

regulatory changes and reduce the cost of processing
vast volumes of information, Eigen expects IDP adoption
to accelerate in the next couple of years. There is
incredible potential to reshape the way companies do
business in the not-too-distant future.

Are you ready to discover how Intelligent

Document Processing can help your
organization gain a competitive edge?

Request a demo of our platform to find out

how we can help you.

23
ABO U T EIG EN T ECHNOLOG IES

Eigen is an Intelligent Document Processing company

that enables its clients to quickly and precisely
extract answers from their documents, so they can
better manage risk, scale operations, transform
ways of working and navigate dynamic regulatory
environments.

Eigen’s customizable no-code AI-powered platform

uses machine learning and natural language
processing to automate the extraction of answers
from documents and can be applied to a wide variety
of use cases. It understands context and delivers
better accuracy on far fewer training documents,
while protecting the security of clients’ data.

Our clients include some of the most well-known

eigentech.com
and respected names in business, including Goldman
Sachs, ING, BlackRock and Allen & Overy. Almost half [email protected]
of all global systemically important banks use Eigen UK & Europe: +44 (0) 20 8132 4139
to overcome their document or data challenges. North America: +1 929 999 5908