1) Extract KPI data

The KPI report PDF is accessed through the Tika Parser node.

GPT-4.1 nano extracts and formats the six KPI definitions, while the corresponding charts are organized into a structured table for analysis.

KPI Interpretation with Multimodal AI

This workflow uses OpenAI’s GPT-4 nano multimodal model to automatically interpret KPIs from a company report. It extracts text and charts from the PDF, generates concise insights and recommendations, and compiles them into a structured, AI-assisted report.

4) Report writing

The model’s interpretations are combined with the visuals, structured, and then written to a PDF.

3) Generate AI insights

The workflow connects to OpenAI via the Authenticator and LLM Selector, using the GPT-4.1 nano model.

The Message Creator combines the image, KPI definition, and prompt, and the LLM Prompter returns the interpretation and recommendations.

2) Prepare prompts for interpretation

The KPI definitions and plots are joined into a single dataset.

A structured prompt is created for each KPI, guiding the model to describe data patterns and provide actionable recommendations.