How to Create Data Stories with LLMs: Practical Examples and Effective Prompts

Introduction: Why data storytelling is crucial today

How to turn data into stories with LLMs

1. Define the story’s purpose

Before drafting any prompt, clarify the objective. Ask yourself:What decision does the audience need to make? A business-focused story might highlight a growth opportunity, while a science-focused one could emphasize an unexpected discovery.

2. Prepare data in a structured format

Upload the dataset in a readable format (CSV, JSON) and, if possible, add metadata describing the columns. This helps the LLM understand the context.

3. Craft a storytelling-focused prompt

A well-structured prompt combines instructions, constraints, and the dataset. Here’s a basic template:

Analyze the following dataset and create an engaging story: - Highlight trends, anomalies, and key insights. - Structure the narrative into [section1], [section2], [section3]. - Use a tone suitable for [target audience]. Data: {data}

Effective prompts for data storytelling

Prompt 1: Executive sales summary

Objective:Summarize a month’s sales data into a three-slide presentation.

Generate an executive narrative for the following monthly sales dataset: 1. Provide an overview of performance (total, growth %). 2. Highlight top and bottom-performing products with brief explanations. 3. Suggest two strategic actions to improve performance in the next quarter. Dataset: {sales_data}

Prompt 2: LLM-guided visualization

Objective:Obtain chart descriptions and visualization suggestions.

For the provided dataset, suggest the most appropriate chart type for each variable and write a caption that can be used directly in a visualization tool (e.g., Matplotlib, Plotly). Dataset: {dataset}

Practical example: From CSV to narrative

Imagine you have a CSV file with weekly sales data from three regions. Here’s a workflow to create a story ready for a meeting.

1. Load the CSV:Use Pandas to load the data.
2. Generate the prompt:Insert the data into a prompt like the one above.
3. Invoke the LLM:Use LangChain or a direct API client.
4. Review and format:Adapt the response into slides or a written brief.

Here’s a minimal code example that combines these steps:

import pandas as pd from langchain.chat_models import ChatOpenAI from langchain.schema import HumanMessage # 1. Load the data df = pd.read_csv('weekly_sales.csv') sales_csv = df.to_string(index=False) # 2. Prepare the prompt prompt = f""" Analyze the following weekly sales dataset and create an executive narrative: - Provide a summary of total sales and weekly % change. - Identify the region with the highest and lowest performance, explaining the reason. - Suggest an inventory strategy for the next week. Dataset: {sales_csv} """ # 3. Invoke the LLM llm = ChatOpenAI(model='gpt-4o', temperature=0.3) message = HumanMessage(content=prompt) story = llm.invoke([message]) print(story.content)

2026 trends: Edge and multimodal LLMs

The LLM landscape is evolving rapidly. Here are three key developments shaping data storytelling:

• NVIDIA Jetson Orin Nano 2:Brings LLM processing power directly to drones and robots, enabling real-time analysis of visual data and sensor-based stories without cloud connectivity.
• Gatik and its autonomous trucks:Generate petabytes of data from cameras, LiDAR, and onboard sensors. Integrated LLMs transform this data into actionable safety and efficiency reports.
• Alibaba’s Qwen3.8-Flash-Next:A 125B multimodal MoE model with only 6B active parameters, ideal for generating narratives from text, images, and time series in a single pass. Its efficiency makes it perfect for edge data storytelling pipelines.

When designing your workflow, consider whether an edge model like Qwen3.8-Flash-Next can replace a cloud-based LLM to reduce latency and costs.

Key takeaways: Concrete actions to get started

• Define the story’s purpose before writing any prompt.
• Use a consistent prompt format that includes context, actions, and data.
• Experiment with edge LLMs (e.g., those integrated in Jetson Orin Nano 2) for real-time analysis.
• Combine structured data with multimodal sources (images, sensors) for richer stories.
• Monitor API costs and consider using open-weight models like Qwen3.8-Flash-Next for large-scale workloads.

Conclusion

Data storytelling with LLMs is becoming an essential tool for transforming complex data into clear, action-oriented narratives. By following the practical prompts and workflows outlined here, you can start creating data-driven stories today, leveraging the latest innovations such as edge and multimodal models introduced in 2026. Begin with a clear goal, experiment with prompts, and adapt your stack based on emerging trends to stay ahead.

**IMPORTANT:** We are in 2026. All temporal references must be current. Do not mention past years such as 2024, 2023, etc. The content must be fresh and relevant to the present. **CURRENT CONTEXT (August 2026):** Here are some recent trends and news to inspire you: - Alibaba’s Qwen Team Releases Qwen3.8-Flash-Next: A 125B Multimodal MoE With 6B Active Parameters Previewing the Qwen4 Architecture: We examine Qwen3.8-Flash-Next, Alibaba’s open-weight multimodal Mixture-of-Experts model and an early preview of the Qwen4 architecture. We break down... [2026-08-26] - Google Research Introduces GlucoFM: A 0.72M-Parameter Dual-Stream Foundation Model for Continuous Glucose Monitoring: Google Research and UNSW Sydney released GlucoFM, a self-supervised foundation model that splits a CGM trace into a slow physiological stream and a tr... [2026-08-27] - Orchestration is the new challenge for CX in the age of AI agents: Presented by Tata Communications Enterprises are deploying AI agents, voice AI, and automation across messaging, voice, and digital channels faster than ever... [2026-08-26] Use this current information as inspiration to create an original and relevant prompt for 2026.