SL Soft AI IDP service

2026: Additional Document Image Quality Control Mechanisms and AI Search

SL Soft has released the next release of the IDP service (the product is part of the SL Soft AI cluster). The updated version has improved the user interface, expanded the tools for quality control of incoming documents for recognition, and added support for intelligent search of the array of processed documents. Softline announced this on January 14, 2026.

Added a mechanism for assessing the quality of scans or photos of documents using the DPI metric (dots per inch). Now the IDP service automatically detects low-quality files and predicts the likelihood of errors when extracting data. Employees receive timely notifications of potential problems and can promptly request digital images of higher-resolution documents from counterparties. The metric complements the text layer quality indicator implemented in the previous release. This allows companies to reduce the cost of processing documents by early detection and elimination of low-quality files.

The attribute recognition confidence threshold is controlled. Users can set a minimum confidence level for specific fields: anything below the thresholds will automatically be highlighted for additional validation. This reduces the risk of errors in the processing of documents, prevents the transfer of incorrect data to accounting systems, helps to avoid confusion in payments and failures in related business processes, and eliminates potential financial and reputational losses.

The functions of training the ML model to recognize documents and highlight key attributes of documents in the IDP service interface are enhanced. Predefined templates for key details - position, signature and signature decryption are presented. The mechanism allows you to interactively train the model on specific areas of the document without programming and involving ML engineers and improve the accuracy of extracting legally significant fields.

Intelligent search SL AI Search is also integrated into the platform. It works in a dialogue format: the user formulates a natural language request and receives an accurate answer, a relevant fragment of the document or links to files with the necessary information. This greatly simplifies the navigation of the document array and speeds up access to the necessary information.

Additionally, the license control module has been improved: the system displays the validity period and the current use of licenses. This will allow them to be extended in a timely manner and prevent a potential shutdown of workflows. Also, the IDP service can now be delivered as a widget with a personalized interface containing only the necessary tools. The widget integrates easily through iFrame and adapts to the customer's business scenarios.

We strive to provide the market with the most reliable and functional tool for intelligent document processing. The added mechanisms for quality control of photos and scans of documents, AI search and other improvements optimize the labor costs of customers and allow building stable, unmistakable automation processes, - said Pavel Borchenko, Advisor to the Managing Director of SL Soft AI of SL Soft.

2025: ML Model Training by Document Field Directly from IDP Service Interface

The updated version of the SL Soft AI IDP service of SL Soft FabricaONE.AI (shareholder - Softline Group of Companies) allows customers to independently scale its application within the company. Softline announced this on November 6, 2025. Among the main changes are automatic classification of documents using LLM and training the ML model to recognize new types of documents without involving programmers.

The key possibility of the version is to train the ML model by document fields directly from the IDP service interface. Now, ML engineers or markup specialists are not required to configure model recognition and training. For this, an operator is enough to mark up documents in an intuitive interface and verify the data offered by the system. The update makes it easier to scale the product for processing any corporate documents: legal, accounting, human resources, financial.

Also, this version of the service implements the function of classifying documents using large language models (LLM). The technology allows you to determine the type of document by its attributes and context, which is especially important when working with unstructured information - contracts, statements, claims, incoming correspondence. This reduces manual sorting and speeds up document routing. The tool is effective even when learning from a limited dataset: two samples of a new type of document suffice. This solves the problem of companies with insufficient data to apply ML.

In the updated release, a text layer quality indicator has been added that indicates the unreadability level of scan images. Thanks to this, employees promptly receive notifications of problems and can request the correct files from the counterparty. The function is especially useful for companies with a large document flow - it helps to manage information flows and speed up the exchange of data between branches and counterparties.

Other updates include faster and more accurate comparison of document versions, making it easier to find and assess the significance of edits. In addition, the updated version of the IDP service includes the ability to centrally manage roles and accesses, which simplifies control over user rights and improves data security.

This version is a serious step towards creating a fully autonomous intelligent document processing service. Without deep technical knowledge, customer employees can independently post data using ML technology and teach AI models to new types of documents, scaling the product to any division. These functions directly affect the return on investment of our clients, reducing costs and speeding up processes, "said Pavel Borchenko, Managing Director of SL Soft AI of SL Soft FabricaONE.AI (shareholder - Softline Group of Companies).

The SL Soft AI IDP service helps companies process structured and unstructured documents using LLM and ML. Among the key advantages of the product: no-code mode and a convenient interface, which allows the customer to independently customize document processing, tools for quick training in working with a new type of documents, as well as high recognition accuracy of even complex elements such as barcodes, handwriting, tables and others.