Artificial Intelligence in Business Process Automation: How WEBCON Utilizes AI

Artificial intelligence’s influence is broad, from personalized shopping recommendations to automated home temperature management and summarizing search results. However, these kinds of applications pale in comparison to the benefits AI can bring to organizations, including bespoke business process solutions.

The WEBCON platform undergoes continuous evolution, with a significant focus on utilizing artificial intelligence. We apply AI where it delivers the highest business value, meaning providing greater opportunities without incurring greater risk. It would be easy to randomly add AI features for technology’s sake, but WEBCON’s approach involves aggressive investment in low-code development scenarios where the return on investment is clear and observable.

Examples of Artificial Intelligence Applications in Business Process Automation

We hope the following examples will illustrate what we mean by a clear and observable return on investment:

Effective Document Registration with OCR and AI

Documents are crucial in business processes, whether they’re electronic or paper-based and whether they’re internally authored or received from outside parties. Many intelligent automation processes involve shepherding these documents from receipt to disposition, including scanning and archiving.

When documents are received on paper or as graphics-only scanned images, the need for Optical Character Recognition (OCR) technology is obvious, but that isn’t enough. Interpreting recognized text requires elaborate neural networks that extract crucial data from scanned documents. You need to combine OCR with AI.

For example, consider registering cost invoices. An OCR engine such as ABBYY FineReader overlays a text layer on the scanned document and recognizes each character. Neural networks then automatically identify and populate key header information into forms, including amounts (net/gross), dates (issuance, sale, payment), bank account numbers, the contractor’s VAT identification number, and other relevant details.

WEBCON provides its clients with a pre-trained neural network for cost invoices, which can be further trained and improved with each use, enhancing the efficiency of automation based on the specific invoices received by the organization.

A universal neural network is used for most documents, and WEBCON provides the ability to handle exceptions when the layout of any given scanned invoice differs significantly from established expectations. Training the neural network is very simple and requires no coding or specialized knowledge. Users only need to identify the correct data from the scanned preview image and indicate its target field on the form.

Applying AI to OCR within a low-code business process allows for the mass registration of documents in WEBCON, leading to a reduction of up to 90% in the time employees spend on data entry.

Using AI for Anomaly Detection and Suggestions

Frequently executed business processes build up large repositories of historical data—a gold mine for inferential AI. Patterns can be gleaned from that data and used to examine each new unit of work for irregularities. WEBCON employs machine learning techniques to detect anomalies, analyze information, identify inconsistencies with historical data, and provide actionable suggestions.

How does this work in practice?

AI flags payment anomalies

The system reports when an entered payment amount significantly deviates from the usual range.

AI detects an unfamiliar bank account number

The system flags a bank account number that hasn't been previously used by a particular contractor.

AI identifies duplicate payments

Algorithms catch duplicate payments and warn against making the same payment again.

These examples are not limited to invoice processing or fraud detection. AI tools can also be applied in scenarios such as:

• Monitoring production machinery: AI can monitor equipment to prevent major breakdowns by detecting early signs of failure.
• Analyzing patient medical data: AI can detect potential health issues and errors in medical documentation.
• Providing process instance suggestions to users:
  The system generates reports based on data and process actors, identifying elements that require employee attention. This feature allows users to easily review the most important tasks and issues across the system in a single place.

Low-Code and Generative Artificial Intelligence

The potential of generative AI in low-code platforms lies in lowering the entry barrier for application development. It’s advantageous to use generative AI selectively, where qualitative improvements are most noticeable.

Even low-code environments use some code for specific tasks, and when they do, generative AI can be a big help. When a form’s behavior needs to be enhanced, and WEBCON’s graphical form rules aren’t sufficient, advanced rules can be authored using JavaScript. The fact that this is possible is useful, but for many low-code developers, it is not second nature, and properly trained generative AI can be a big help.

For example, very few low-code developers enjoy crafting regular expressions. A power user configuring a validation for a form field, such as a VAT number, needs to enter a regular expression that specifies the required format. Instead of manually crafting this expression, the user can have ChatGPT (for example) quickly provide the desired expression.

Continuing in this vein, while a great deal of logic and data access can be accomplished using data sources and rules in WEBCON , sometimes it’s necessary to query process data using Transact-SQL. For many low-code developers, this can be daunting. Generative AI can assist in both creating and optimizing these queries.

While WEBCON prides itself on delivering a world-class low-code platform, “low-code” does not mean “no code.” When code is (hopefully only occasionally) needed, generative AI like ChatGPT can be invaluable. As good as generative AI has quickly become, it’s a very good idea to maintain a “trust but verify” approach, where any generated code is manually validated.

Low-Code and Conversational AI

Another application of artificial intelligence is in communication, known as conversational AI. This technology enables devices or applications to interact with humans using natural language. WEBCON clients also use the capabilities offered by conversational AI to optimize case management, both within the organization and in interactions with external parties (clients and stakeholders).

For example, consider a chatbot that answers user questions about ongoing issues. This interactive agent relies on a database populated with specific information related to a particular specialty, such as legal documents, contracts, or process descriptions. By combining machine learning with a chatbot, you can create a versatile and user-friendly knowledge base. This supports internal users in making informed decisions and provides external users with easy access to desired information without searching through various data sources (websites, documents, regulations).