VIDEE: Visual and Interactive Decomposition, Execution, and Evaluation of Text Analytics with Intelligent Agents

VIDEE is a new system that enables entry-level data analysts to perform advanced text analytics using intelligent AI agents without specialized NLP knowledge. The platform combines human-in-the-loop decision-making with LLM-powered execution and evaluation, demonstrated through quantitative experiments and user studies showing effectiveness across experience levels.

VIDEE addresses a critical accessibility gap in text analytics by lowering the technical barrier for non-expert users. Traditional text analytics required deep NLP expertise, limiting adoption to specialists. By leveraging large language models and human-agent collaboration, VIDEE democratizes advanced analytical capabilities—topic detection, summarization, and information extraction—making them available to entry-level analysts.

The three-stage workflow reflects broader trends in AI system design. The decomposition stage uses Monte-Carlo Tree Search with human feedback, enabling iterative refinement of analytical goals. This human-in-the-loop approach acknowledges that pure automation often fails on complex, ambiguous tasks. The execution stage generates executable pipelines, automating the technical implementation. The evaluation stage combines LLM-based validation with visualizations, allowing users to understand and verify results without deep technical knowledge.

For enterprise and institutional users, VIDEE reduces the cost of analytical talent while maintaining quality oversight. Organizations can expand their analytical capacity by augmenting junior staff with AI agents rather than hiring senior NLP specialists. This efficiency gain has implications for competitive advantage in data-driven decision-making across industries.

The user study revealing distinct behavior patterns across experience levels suggests that future intelligent analytics systems should adapt interfaces and workflows dynamically. As AI systems become more capable, the bottleneck shifts from technical execution to human judgment and validation. VIDEE's design validates this shift, positioning human expertise in guiding and evaluating rather than implementing—a model likely to influence broader enterprise AI adoption strategies.

• →VIDEE enables entry-level analysts to perform advanced text analytics by combining LLMs with human-in-the-loop decision-making across three workflow stages.
• →The system reduces demand for specialized NLP expertise while maintaining analytical rigor through human validation and interactive execution.
• →User studies demonstrate effectiveness across varying experience levels, with distinct behavioral patterns informing future interface design.
• →Human-agent collaboration emerges as a key design principle for making AI systems accessible without sacrificing quality or control.
• →Enterprise adoption potential lies in reducing analytical talent costs while expanding organizational analytical capacity through AI augmentation.