ExB

FAQs

What problem does ExB solve?

ExB caters to the market of insurance companies and financial institutes that yearly deal with billions of unstructured documents (such as letters, emails, pdfs, invoices, customer requests, reports, financial documents, claims…) which need to be processed.
Such documents cover customer requests in policy management (contract termination, address change, updated bank information…), claims, reports, invoices and more.

Currently, 85% of these documents are processed manually with high costs (up to €18/document). And this also affects competitiveness and customer experience as it’s not uncommon for customers to wait two or three weeks (and often more) after sending in a request.

 

What goals can we achieve thanks to ExB?

ExB allows your company to process every document (even if completely unstructured) and customer request automatically in seconds instead of days.

Our Cognitive Process Automation (CPA) platform is an Intelligent Document Processing system that truly understands and extracts information from any document for any workflow (even the most uncommon ones) to feed process automation systems.

Automating input management with ExB leads to:

• Immediate insights
• Automated decision making processes
• Faster response time to customer requests
• Better CX and higher customer retention rate
• Higher competitiveness
• Lower costs
• Better employee experience (providing support for repetitive manual labor)
• Higher productivity and efficiency with less manual work
• Data accuracy
• Data centralization
• Higher ROI for RPA solutions (which rely on data to carry out processes) → no CPA=no scalability for RPA
• Higher ROI in BigData operations and Data Lakes (70% consisting of unstructured data)
• Improved Standard Operating Procedures (SOPs)
• Scalable framework for full-stack AIOps across the entire organization instead of single, one-off, isolated, AI projects

 

What kinds of documents does ExB process?

ExB processes every document type for every workflow.

With ExB you can process and extract information from any kind of document. From structured and semi-structured documents such as invoices, forms, and tables to completely unstructured documents such as emails, letters, assessments, medical reports.

 

What format should the documents be in?

ExB processes images and texts in one go.

ExB allows you to process and extract information from any kind of digital or non digital format: from email, to pdf, scanned or photographed pages, images,or text files of any kind.

 

How does Exb process a document?

ExB analyzes and processes documents exactly as a human being

1. Firstly, ExB looks at a document as a whole based on all of its parts (for example combining all the information contained in different components such as a cover letter, invoice and its line items, pictures, and reports to process a single claim)

2 Unlike other solutions that simply convert images to text to extract information, ExB analyzes then each component by looking at its layout and structure as a whole to understand relations between different parts

3 ExB then categorizes and classifies the kind of document being processed together with its intent (is it a medical report? A contract termination letter? A hospital invoice? A claim? A communication of an address change?…)

4. ExB then extracts all the information that you, as our customer, consider important based on your requirements from images (such as stamps), texts, and optical marks (such as tick boxes in forms) to allow you to have all the data points you need for routing the request to different agents or systems based on your internal processes

5. ExB uses the latest deep learning algorithms to truly understand spatial and contextual cues to immediately understand meaning and cause and effect relationships to differentiate for example between different dates contained in a document (e.g. issuing date, hospitalization date, discharge date, diagnosis date in a medical report) or to determine the role of different agents (e.g. the insured person as the patient, the doctor, the insurer)

6. All the information extracted is then verified for plausibility (does it make sense? Is there anything weird with the results?) and also validated through external sources to make sure that there are no errors, fraud, or inaccuracies.

7. If the document passes all the tests then the system returns a structured table that contains all the information organized the way you, as a customer, need to feed your process automation or input management systems.

 

What can I do with the structured data that is produced by Exb?

1. Feed your input management. ERP, BPM, or RPA systems

Once all the data is extracted and verified, you can use ExB to map each entity to specific internal processes. For example you can collect all the information contained in a contract termination letter (such as who, what policy number, from what date) have ExB determine if everything is okay and then map each element into your input management system in any specific way your process automation system expects data to be fed into to carry out the task.

2. Automate decision making processes

Shall we settle this claim or not? What’s the likelihood of a potential MDK reimbursement for this hospital invoice? Is the request compliant with current regulations?…

ExB can help insurers automate decision making processes through greater speed to insights.

3. Speed up research and time to insights (Expert Acceleration)

Do you need to isolate information about a single topic from an array of long documents? Do you need to analyze and summarize maintenance logs of one of your customers?

ExB can help you classify and extract only relevant information that you can use for research or verification processes converting long reports into bite-sized summaries.
Exb can look for and extract specific information across several documents and find relationships between all the extracted data.

 

How does ExB work?

Exb adopts a unique multimodal AI approach protected by several patents to boost automation from a current average of 25% to 70%.

The system generates machine learning modules that process specific kinds of documents in steps. Each module is specialized for a use-case.
For example, a specific module that you can implement could be a claim module, to process all incoming claims.

Each module is specialized to recognize different entities which are relevant for that particular workflow and trained to combine them to work on the use case as a human.

ExB offers a series of pre-trained modules for typical use cases in claims handling, policy administration, or medical reports.

Tailored Modules for YOUR specific needs

But Exb doesn’t just offer a cookie-cutter solution for common use cases. After all, there are over 400 specific workflows that require input management each of which requires a very unique set of information for process automation.

With ExB, you can create ultra-specific modules based on your internal processes without any coding. Anybody in your organization, with very minimal training, will be able to create and train models simply using their mouse for every use case.

And if you don’t want to train modules yourself, we can combine your know-how on what information is needed and where it needs to be routed to to carry out a process and our know-how on how to quickly train a module to create your unique Cognitive Process Automation system.

 

Why is Exb so special?

ExB transforms your entire organization not just a process.

There are many solutions out there that process either images or text. Some of them are one trick ponies specialized in one single kind of document (for example invoices or licence plates…).

The majority of them uses a legacy approach that relies on rule-based systems that can’t cope with unpredictability and completely unstructured information. They deliver okay results when it comes to structured or semi-unstructured documents if they’re all in the same format. But that’s it.

Some of them are inflexible and force you to change your internal processes to adapt to the system while others are flexible but require months and lots of coding to adapt to your specific requirements.

ExB is the only system that processes both images and texts in a single process. The platforms can be easily trained to process any kind of structured or completely unstructured document and adapt to every workflow.

One single end-to-end solution, for every use-case, that anybody can train in days instead of months, without any coding or AI-related skills.

That’s why current solutions out there can guarantee a maximum of 25 to 30% automation grade in input management while ExB can bring the rate up to 70-75% with the highest level of accuracy.

 

What do I get from ExB?

ExB as SaaS, AIaaS, or self-hosted.

It really depends on how you want to work with our system.

A simple case might be one in which you only need a single specific module to process a single kind of document for a single workflow.

In this case, the use-case module is exported into a docker file which can be hosted on your own server, in your private cloud, in a public cloud (like AWS) or in our data center based on your requirements.

This “machine” (which we call TSM – “Thinker Service Module”) is now at your service and will accept documents as input to produce structured information in simple tables as output that you can export in formats such as JSON or XML to feed your input management or process automation systems.

We already have a series of pre-trained modules you can already implement today!

Each additional use case is an additional TSM that can be implemented. Each of them receives a particular kind of document which ExB can even easily classify and distribute.

You can add as many modules as you wish to transform your entire input management with cognitive automation and implement AI across your entire organization.

Other optional components of our system that customers can have access to include:

1) A quality gate: to block documents that haven’t been processed properly and pass them on to a human being for verification and correction (HITL – “Human in the loop”) based on a set of filters you define
2) A reviewer: to see what portions of the document have been recognized and how they have been interpreted
3) Our Cognitive Workbench: the platform that allows us to create modules.
If you decide to work independently on the creation and training of your own modules, you can in fact, work directly with our own Cognitive Workbench (in the cloud, or on-premise) and use the system to train as many modules as you wish without any coding.

What we offer:

Ready to use modules (R2U): Pre-trained modules for common use cases such as contract termination, bank information changes, invoices…
Built to order modules (B2O): Modules we can train for the customer for specific workflows based on their requirements of what data needs to be extracted or what documents need to be classified and processed.
Cognitive Workbench (CWB): We can provide customers access to our Cognitive Workbench (the tool we use to create and train modules) so that they can work independently on their own modules .

 

How do I start?

Let’s talk, analyze your needs, check your use cases, define a plan together.

First of all we need to define your specific use case. What documents are associated with the process? What’s their composition and logic? (emails and attachments for example) What information do you need to extract to automate the process?

For many use cases, you can already implement existing pre-trained modules. Deployment and optimization might require some engineering work, but once the system is in place, everything happens automatically and your clerks can leverage the system to automate a large portion of their work.

If there are no modules we can offer you right now for your specific use case, we only need you to provide us with some training data and we can create ad-hoc modules in just a few weeks for you. Then, the TSM (our “Cognitive Machine”) can be deployed and used right away.

Alternatively you can opt for a more end-to-end approach and get access directly to our cognitive workbench.

In this case we need to define the kind of deployment and our solution architects will work with your engineers to install our NLP engine and we can also offer training for your staff so that they can annotate training documents directly in our Cognitive Workbench to train the system on different use cases.

 

How much do I pay?

Flexible pay-per-use model, scalable solution with immediate ROI

It really depends on deployment, use case, volume, and optional elements you need.

One thing is for sure:
We’re cheaper than our competitors despite being able to tackle more use cases, offer higher accuracy, and provide a more tailored product for your operations.
And our pricing doesn’t punish growth!
The higher the volume and the more use cases, the cheaper ExB becomes. We don’t want to upsell you with a bunch of tiered packages for different use cases. We want to offer a scalable solution that covers every process in your company to completely automate every cognitive process.

We offer a flexible “pay per use” pricing model that won’t force you to deal with pre-defined pricing tiers.
Processing documents manually can cost you between €1 and €18 per document based on the kind of document and grade of complexity of the process.
With ExB, you can be sure you’ll only pay a few Euro cents per document and prove ROI from day one after implementation.