Prompt Engineering Playbook: Data Enrichment Prompts for Automated Workflows

Prompt Engineering Playbook: Data Enrichment Prompts for Automated Workflows

Data enrichment is a cornerstone of modern automated workflows, transforming raw, incomplete, or unstructured data into actionable business insights. With the rise of large language models (LLMs), prompt engineering has become a critical skill for automating and scaling data enrichment tasks across industries. This playbook provides a step-by-step, hands-on guide for designing, implementing, and optimizing data enrichment prompts in your AI-powered workflows.

For a comprehensive overview of prompt engineering’s role in end-to-end AI automation, see The Ultimate AI Workflow Prompt Engineering Blueprint for 2026.

Prerequisites

• Tools & Libraries:
  • Python 3.9+
  • openai SDK v1.2+ (PyPI link)
  • pandas v1.5+ (for data manipulation)
  • Command-line interface (Bash, PowerShell, or Terminal)
• Accounts & API Keys:
  • OpenAI API key (or Azure OpenAI key)
• Knowledge:
  • Basic Python scripting
  • Familiarity with REST APIs
  • Understanding of LLM prompt structure and limitations

1. Define Your Data Enrichment Objectives

1. Identify Enrichment Goals:
   • What missing or incomplete data do you want to fill? (e.g., company descriptions, product categories, address standardization)
   • What context or sources will the LLM need?
2. Example:
   • Enrich a customer database by adding industry categories and LinkedIn URLs based on company names.
3. Tip: Map out your input fields and desired outputs in a table for clarity.
   | Company Name | Industry Category | LinkedIn URL | |---------------|------------------|---------------------| | Acme Corp | ? | ? | | DataVision | ? | ? |

2. Prepare and Structure Your Input Data

1. Load your data into a DataFrame:

   pip install pandas

   
   import pandas as pd
   
   df = pd.read_csv('companies.csv')
   print(df.head())
         

   Screenshot description: DataFrame preview showing company names with missing fields.

2. Clean and normalize input fields:

   
   df['Company Name'] = df['Company Name'].str.strip().str.title()
         

3. Sanity-check for nulls and duplicates:

   
   print(df.isnull().sum())
   print(df.duplicated().sum())
         

3. Design Effective Data Enrichment Prompts

1. Prompt Template Example:

   
   You are a data enrichment assistant. Given a company name, return its industry category and official LinkedIn page URL.
   Format your answer as JSON:
   {
     "industry_category": "...",
     "linkedin_url": "..."
   }
   
   Company Name: {COMPANY_NAME}
         

2. Dynamic Prompt Generation:

   
   def build_prompt(company_name):
       return f"""
   You are a data enrichment assistant. Given a company name, return its industry category and official LinkedIn page URL.
   Format your answer as JSON:
   {{
     "industry_category": "...",
     "linkedin_url": "..."
   }}
   
   Company Name: {company_name}
   """
         

3. Best Practices:
   • Be explicit about output format (e.g., JSON, CSV, key-value pairs).
   • Provide clear instructions and sample outputs.
   • Consider including context (e.g., "Assume the latest public data is available.").
4. For advanced prompt patterns across multi-step data pipelines, see Prompt Engineering for Multi-Step Automated Data Pipelines: Strategies for Accuracy and Speed.

4. Automate LLM Calls for Batch Data Enrichment

1. Install OpenAI SDK:

   pip install openai

2. Set your API key as an environment variable:

   export OPENAI_API_KEY="sk-..."
         

   Windows:

   set OPENAI_API_KEY="sk-..."
         

3. Batch Processing Script:

   
   import os
   import openai
   import pandas as pd
   import json
   import time
   
   openai.api_key = os.getenv("OPENAI_API_KEY")
   
   def build_prompt(company_name):
       return f"""
   You are a data enrichment assistant. Given a company name, return its industry category and official LinkedIn page URL.
   Format your answer as JSON:
   {{
     "industry_category": "...",
     "linkedin_url": "..."
   }}
   
   Company Name: {company_name}
   """
   
   def enrich_company(row):
       prompt = build_prompt(row['Company Name'])
       try:
           response = openai.ChatCompletion.create(
               model="gpt-3.5-turbo",
               messages=[{"role": "user", "content": prompt}],
               temperature=0
           )
           output = response['choices'][0]['message']['content']
           data = json.loads(output)
           return pd.Series([data.get("industry_category", ""), data.get("linkedin_url", "")])
       except Exception as e:
           print(f"Error: {e}")
           return pd.Series(["", ""])
   
   df = pd.read_csv('companies.csv')
   df[['Industry Category', 'LinkedIn URL']] = df.apply(enrich_company, axis=1)
   df.to_csv('companies_enriched.csv', index=False)
         

   Screenshot description: Terminal output showing progress and sample enriched rows in the DataFrame.

4. Rate Limiting: Add time.sleep(1) between calls to avoid hitting rate limits.
5. Tip: For workflow orchestration and multi-step enrichment, see Prompt Engineering for Complex Multi-Step AI Workflows: Templates and Best Practices.

5. Validate, Clean, and Post-Process LLM Outputs

1. Validate Output Structure:

   
   def is_valid_linkedin_url(url):
       return url.startswith("https://www.linkedin.com/company/")
   
   df['Valid LinkedIn'] = df['LinkedIn URL'].apply(is_valid_linkedin_url)
   print(df['Valid LinkedIn'].value_counts())
         

2. Handle Nulls and Errors:

   
   df = df.dropna(subset=['Industry Category', 'LinkedIn URL'])
         

3. Deduplicate Results:

   
   df = df.drop_duplicates(subset=['Company Name'])
         

4. Export Cleaned Data:

   
   df.to_csv('companies_enriched_clean.csv', index=False)
         

   Screenshot description: Cleaned CSV file preview with all enriched fields populated and valid URLs.

5. For automated data cleansing prompt templates, see Crafting Effective LLM Prompts for Automated Data Cleansing Workflows.

6. Integrate Data Enrichment into Automated Workflows

1. Connect to Workflow Automation Platforms:
   • Airflow, Prefect, n8n, Zapier, or custom Python scripts
2. Example: Airflow DAG for Scheduled Enrichment

   
   from airflow import DAG
   from airflow.operators.python import PythonOperator
   from datetime import datetime
   
   def run_enrichment():
       # (Insert batch enrichment script here)
       pass
   
   with DAG(
       'company_data_enrichment',
       start_date=datetime(2024, 6, 1),
       schedule_interval='@daily',
       catchup=False
   ) as dag:
       enrich_task = PythonOperator(
           task_id='enrich_companies',
           python_callable=run_enrichment
       )
         

3. Trigger enrichment jobs automatically on new data arrivals.
4. For low-code automation approaches, see Prompt Engineering for Low-Code AI Workflow Automation: Templates and Pitfalls.

Common Issues & Troubleshooting

• LLM Outputs Invalid JSON:
  • Use json.loads() with try/except and fallback parsing.
  • Explicitly instruct the model to output only JSON.
• Hallucinated or Inaccurate Data:
  • Set temperature=0 for deterministic outputs.
  • Cross-check results with external APIs or reference datasets.
  • See Prompt Engineering to Reduce Hallucinations in Automated Document Workflows for mitigation tactics.
• API Rate Limits:
  • Throttle requests using time.sleep() or batch processing.
  • Monitor API usage and handle 429 Too Many Requests errors gracefully.
• Prompt Token Limit Exceeded:
  • Shorten instructions and input fields.
  • Chunk large datasets and process in smaller batches.
• Workflow Integration Fails:
  • Check environment variables and API keys in your automation platform.
  • Test scripts outside the workflow engine to isolate errors.
• For debugging and optimization, see LLM Prompt Debugging: How to Fix and Optimize Broken Workflow Automations.

Next Steps

• Template Library: Build a reusable library of prompt templates for different enrichment scenarios. See How to Build a Robust Prompt Library for Automated AI Workflows.
• Advanced Automation: Explore multi-modal and multi-step enrichment workflows. See Mastering Multi-Modal Prompts in Workflow Automation: Best Practices for 2026.
• Optimization: Experiment with prompt variants, model selection, and post-processing for accuracy and speed. For comparative analysis, see Prompt Engineering vs. Classic Automation Scripting: Which Is Better for 2026 Workflows?.
• Broader Application: Extend data enrichment techniques to marketing, sales, finance, insurance, and other business domains. See Prompt Engineering for AI Marketing Workflows: 2026’s Most Effective Templates and Prompt Engineering for Finance Workflows: Real-World Templates and Optimization Strategies.
• Stay Updated: As LLMs evolve, revisit your prompt strategies regularly. For ongoing best practices, refer to The Ultimate AI Workflow Prompt Engineering Blueprint for 2026.

---

Related Reading: Prompt Engineering for Workflow Automation: Tips, Templates, and Prompt Libraries (2026)  |  Prompt Engineering Tactics for Automated Marketing Campaigns in 2026