Intelligent Automation Demonstration

Automated Service Application and Approval

Quanton Intelligent Automation

Imagine the possibility of applying for a personal loan 24-hours a day, 7-days a week and receiving a real-time approval.

Scroll down for instructions on how to interact with the demonstration now.

Defining Quanton Intelligent Automation™

 

Noun:

The ability for technology to engage with people, systems, and things to interpret information, structure data and form probabilistic outcomes for information-intensive processes in autonomous work execution. 

– Quanton 2020

Quanton Intelligent Automation 2020

Intelligent Automation Demonstration Use Case

 

An existing customer for a financial services organisation applies for a new personal loan using a chatbot interface and obtains real time conditional approval (conditional approval based on application scoring and subject to final credit checks).

Intelligent automation demonstration solution

What We Built

Conversational Design

Conversational flow that identifies and manages a personal loan intent, could be deployed in multiple channels like a website chatbot, Facebook Messenger or a voice IVR and could engage in various languages including but not limited to English and Chinese.

Machine Learning Algorithm

Machine learning algorithm to assess the quality of loan applications based on information from existing customer records and new application information provided by the customer.

Automated Task Execution

End-to-end task execution using a combination of API’s and Robotic Process Automation technology.

How to Use This Intelligent Automation Demonstration

Step One:

Click the icon to activate the chatbot interface in the bottom right hand corner of the screen.

Step Two:

The use case scenario is for a personal loan, when prompted ask the chatbot to apply for one.

Step Three: 

The use case is based on an existing customer. Use customer ID 20001.

You will be asked a security question to validate your Identity. Use Date of Birth 13/7/1985 In any standard date format you want (like dd/mm/yyyy or 13th Aug 85).

You could also try customer ID 22011 with Date of Birth 3/7/88. This profile has lower quality finance attributes like lower income.

Step Four: 

You will be asked to validate existing information about ‘your customer profile’ in the system records. When asked ‘are your details correct?’ Try clicking ‘no’. You will only be able to update certain fields – enter you’re actual email in the field (this is not permanently recorded – for approved applications you will receive an email).

Step Five: 

Enter the remaining information required for the loan application.

You can enter your own variables of try one of these three scenarios:

Scenario One: 

Loan Reason: New Car

Loan Value: $20,000

Scenario Two: 

Loan Reason: New Boat

Loan Value: $45,000

Scenario Three: 

Loan Reason: Medical costs

Loan Value: $55,000

To refresh the chat session at any point click the refresh icon at the top of the chat bot window.

Intelligent Automation Demonstration Refresh Session

You will receive one of three outcome responses: 

Conditional Approval Response

If you see this response, your existing customer information in our record system and the new information you have provided were submitted to a machine learning algorithm to assess the quality of your application.

Business rules were then applied to reach a conditional approval outcome – you would be approved subject to additional checks like a credit check.

Intelligent Automation Demonstration Conditionally Approved

Additional Assessment Required 

If you receive this response, there is an aspect of your application which requires further assessment.

In real-world application the case will be placed in a queue to be assessed by a human case worker whose role could be to pro-actively work with the applicant to see how the application could be changed to gain an approval.

The chatbot interface could be configured to immediately hand the conversation off to a human agent to manage the conversation while it is live and work with the customer online.

Intelligent automation Demonstration Referall Outcome

Application Declined 

If you receive this response, your application would not be approved, based on the parameters of our Machine Learning model.

The reason for decline could be the presence of multiple negative factors like the amount of the loan relative to the applicants income, the purpose of the loan or unseen data like the number of recent applications.

Intelligent Automation Demontration

What’s Happening When You Engage With This Demonstration

We made this look simple – but there’s a lot happening in the background.

The Chatbot interface is underpinned by a Conversational AI platform that manages the customer conversation and initial application process, including the retrieval of information from record systems.

The information collected, and the existing customer information are passed into a machine learning algorithm to determine the quality of the loan application, which is then approved, declined or referred for further assessment based on the outcome score.

Customers are advised of the outcome. Assuming the loan application is successful, a customer is asked to proceed with the loan where external credit checks could be automatically completed.

Robotic Process Automation is being used to generate customer notification emails and in a real-life scenario would also be used to manage downstream processes, for example, executing credit bureau checks.

In Real-Time and On-Demand

  • Accessed and initiated on-demand by user.
  • Validates customer identification using security questions.
  • Collects new customer information required through conversation.
  • Validates existing customer information and updates record systems.
  • Autonomously approves, refers or declines customer application.
  • Generates approval outcome and communicates to customer.

How the Machine Learning Algorithm Works

The Machine Learning algorithm used in this demonstration is designed to determine the probability that a loan application will default; we call this our target variable. For example, a case which produced a 0.2 (20%) likelihood of default would be good when compared with a 0.6 (60%) likelihood of default.

Once a score is available for a case, business rules can then be applied to determine how the case is handled.

How Does the Algorithm Determine the Likelihood of Default?

 

The algorithm uses signal data, which are data points that have been determined to have a statistical influence on an outcome. The cumulative impact of all signal data points available are used to determine the probable outcome for the target variable.

The dataset we used to train the model has over 115 data points per transaction, which would be based on all available data about a customer and the transaction case. From the original data we determined that 18 data points which were signal data. The signal data points selected included but were not limited to:

  • Annual income
  • Number of recent credit enquiries
  • Instalment amounts
  • Loan purpose
  • Home ownership status
  • Total current balance
  • Employment title
  • Revolving balance
  • Loan amount

Some of this data, for example loan purpose and loan amount, are collected as ‘new’ information through the application process.

In this use case, much of the data used is ‘other’ information which may be contained in an organisations customer records, for example home ownership status or employment title, may be calculated as part of the application process, installment amounts for example, or obtained from a third party source, for example a credit score.

All this data is aggregated and served to the machine learning algorithm to determine a probabilistic outcome.

Could this Work for New Customer Use Cases?

Why did we choose existing customers and could this work for new customers?

We chose the existing customer scenario in our use case to demonstrate the ability for our solution to access systems and existing customer records, and complete customer identity verification using security questions.

 

The case is completely adaptable for new customer acquisition. 

 

By adapting the conversation flow to accommodate the collection of additional information required for new customers, for example, name, address, contact and employment details, the use case can be applied to scenarios involving new customer acquisition.

Depending on the requirements of an organisation Quanton Intelligent Automation™ capability could also support formal identity verification requirements.

A solution could also be created to differentiate between new and existing customers and accommodate both scenarios within the same process.

New or Existing customers

Real World Transferability

 

The demonstration use case we have used is applying for a personal loan, presumable from a finance institute. If we remove the specific nature of the service, personal loan application, the attributes of the solution are directly transferable to a wider range of uses cases across a wide range of sectors and services.

The scope of application for Quanton Intelligent Automation™ is far and wide reaching, in a range of B2B and B2C applications and is really only limited by your imagination.

Here are three generic use cases which directly leverage the attributes of Quanton’s demonstration solution:

Invoice Processing Automation

New Customer Account Application 

The ability for a customer to apply for a new billing account.

Any service that has a high volume of new customer account applications in a B2C or B2B environment. A good example could be telcos or

 

utility companies.

Invoice Processing Automation

Service Plan Enrolment or Switch Plans

The ability for a customer to add a new service to their billing account, change an existing plan or cancel a plan.

Any business that has a necessarily complex, high volume service offering in a B2B or B2C environment.

Invoice Processing Automation

Service Bookings and Booking Management

The ability for a customer to make new bookings or manage existing bookings that are capacity dependent.

Any business that delivers their services on a booking basis for example transport services or medical practices.

Additional Content:

Intelligent Automation – The Future for Business Process Automation – Go to article

Automation and Conversational AI – The Future of Customer service – Go to article

Top 5 Business Process Automation Trends for NZ in 2020 – Go to article

When automation programmes fail, this is usually why – Go to article

Ready To Have A Conversation About Quanton Intelligent Automation ™ In Your Business?

Get In Touch With Us – we’re here to help innovators destroy last century operating models with new ways of working.

Step One:

The first step is understand what you want to achieve and why. We can do this on a 45-minute call and give you everything you need to understand if Quanton Intelligent Automation ™ really could be your next big win and whether you’re business is ready.

 

Step Two: 

If we like each other and think there’s value the second step will be a workshop with your key people to determine the readiness of your business for Intelligent Automation and develop a strategy to determine your IA Roadmap and qualify potential benefit, but we can discuss that more when we talk.

 

Hit the form or pick one of us to reach out to.

DX Seminar Garry Green Quanton
Adam Taylor, General Manager Advisory and Transformation - Quanton
Russell Berg, General Manager Product and Emerging Technology - Quanton

Garry Green

Managing Director 

Adam Taylor

GM Advisory Services

Russell Berg

GM Product & Emerging Technology

Quanton Intelligent Automation 2020