Formally is our long-term partner with whom we have collaborated for around four years and MagiDocs is part of it. Over the course of this partnership, we have contributed to multiple enhancements within their platform, and we were proud to be entrusted with the development of MagicDocs as a core AI-driven extension. You can explore the full case study of Formally with this link.

MagicDocs was designed to address a common challenge faced by professionals working with high volumes of documents, especially in legal and compliance-heavy environments.

BeeWeb was responsible for implementing the AI-powered document processing logic, building the frontend experience, integrating backend services, and ensuring synchronization with Formally’s platform architecture.

The primary technical challenge was processing extremely long legal documents, many of which spanned hundreds of pages. Generating accurate summaries and extracting meaningful insights from large files presented several difficulties:

    🔹 Token limitations in LLM processing

    🔹 Memory constraints when handling long document content

    🔹 Maintaining contextual accuracy across lengthy legal text

    🔹 Ensuring performance efficiency without degrading user experience

Since MagicDocs operates within a legal context, precision and reliability were critical. Any summarization logic had to preserve essential legal meaning while significantly reducing document length.

Beyond this, there were no major structural challenges, as we were already deeply familiar with Formally’s infrastructure and architectural standards. This prior experience allowed us to move efficiently and focus primarily on AI optimization rather than integration hurdles.

To address large-document processing challenges, we implemented a structured AI pipeline using chunking and contextual summarization strategies:

    🔹 Long documents were split into manageable text segments (“chunks”) within token limits.

    🔹 Each chunk was processed individually through the LLM for partial summarization.

    🔹 Intermediate summaries were then aggregated and re-processed to generate a coherent, high-level final summary.

    🔹 Metadata extraction was handled in parallel, enabling intelligent labeling and categorization.

This hierarchical summarization approach preserved context while staying within model constraints. Additionally, caching mechanisms and optimized API calls improved performance and reduced redundant processing.

To support intelligent labeling, we leveraged AI classification prompts designed specifically for legal document contexts, ensuring accurate categorization across document types.

The project was completed in around 3 months and involved 1 Full Stack Developer and 1 Project Manager.

As MagicDocs operates inside Formally, alignment was crucial. The Project Manager played a central role in:

    🔹 Synchronizing development timelines with Formally’s roadmap

    🔹 Maintaining continuous communication with the client

    🔹 Managing tasks and sprint planning

    🔹 Ensuring architectural consistency across both systems

Development followed an iterative approach, with ongoing feedback loops to refine summarization quality and classification accuracy. Since we were already familiar with Formally’s infrastructure, integration was smooth and efficient.

Technology & Architecture

    🔹 Next.js to build a fast and responsive interface for MagicDocs, ensuring smooth document viewing, AI summary rendering etc.

    🔹Firebase to take care of secure document storage and real-time collaboration features.

    🔹AI Integration (LLM) to analyze, categorize, and summarize documents.

    🔹 Intelligent Document Labeling: Automatically categorize and rename documents using AI.

    🔹 Concise AI Summaries: Generate clear, structured summaries for lengthy documents.

    🔹 Collaborative Document Management: Real-time updates and shared access for teams.

MagicDocs is now live and fully operational inside Formally. It enables users to organize, summarize, and manage documents effortlessly, significantly reducing time spent on manual review and classification.

Working with high volumes of documents?

We’ve already built scalable document intelligence solutions inside enterprise systems. Let’s transform your document workflows with production-ready AI. Book a free consultation with our CTO today!