ai-training-datasets/customer-service

Customer Service Conversations Dataset

This dataset contains 100 realistic customer service conversations between customers and support agents. Each dialogue is 11 turns long and covers a variety of common issues such as late deliveries, billing errors, account problems, and more. It is ideal for training and evaluating AI assistants, chatbots, and customer support models.

Dataset Structure

Each conversation is stored as a JSON object with the following fields:

• id: Unique identifier for the conversation.
• domain: Always "customer service" for this dataset.
• problem: Short description of the customer's issue.
• customer_type: Customer persona (e.g., "impatient executive", "calm and patient user").
• dialogue: List of alternating customer/agent messages. Each message has:
  • role: "customer" or "agent".
  • text: The message content.
• resolution: One-sentence summary of how the issue was resolved.

Example Conversation

{
  "id": 1,
  "domain": "customer service",
  "problem": "order never arrived",
  "customer_type": "impatient executive",
  "dialogue": [
    {
      "role": "customer",
      "text": "I placed an order 10 days ago and it still hasn't arrived. This is completely unacceptable."
    },
    {
      "role": "agent",
      "text": "I sincerely apologize for this delay. I'll look into this immediately. Could you provide your order number?"
    },
    {
      "role": "customer",
      "text": "Order number is #ORD-88421. I need this resolved today."
    },
    {
      "role": "agent",
      "text": "Thank you. I've pulled up your order. I can see it was delayed at the courier depot. I'm escalating this as high priority right now."
    },
    {
      "role": "customer",
      "text": "How long will an escalation actually take? I've heard that before."
    },
    {
      "role": "agent",
      "text": "I completely understand your frustration. I'm contacting the courier directly and will have a confirmed delivery window within 2 hours. I'll also apply a full shipping refund."
    },
    {
      "role": "customer",
      "text": "Alright. A 2-hour update is acceptable. What's the reference for this case?"
    },
    {
      "role": "agent",
      "text": "Your case reference is TKT-20394. I've also sent you an email confirmation with the courier's direct contact and your shipping refund confirmation."
    },
    {
      "role": "customer",
      "text": "I received the email. The refund is noted. I expect the delivery window shortly."
    },
    {
      "role": "agent",
      "text": "Absolutely. You'll receive an SMS from the courier within 90 minutes. Is there anything else I can assist with?"
    },
    {
      "role": "customer",
      "text": "No. Just make sure it's delivered. Thank you."
    }
  ],
  "resolution": "Order traced to courier depot delay. Case escalated with priority, shipping refund applied, and delivery window confirmed within 2 hours."
}
How to Use
Load with Python
python

import json

with open("dataset_customer_service.json", "r", encoding="utf-8") as f:
    data = json.load(f)
    conversations = data["conversations"]

for conv in conversations:
    print(conv["problem"])

Load with Hugging Face Datasets
python

from datasets import load_dataset

dataset = load_dataset("ai-training-datasets/customer-service", split="train")
print(dataset[0])

Customer Personas Included

    frustrated user

    confused beginner

    impatient executive

    elderly user

    tech-savvy user

    calm and patient user

License

This dataset is released under the MIT License. You are free to use, modify, and distribute it for both research and commercial purposes.
Citation

If you use this dataset, please cite:
text

@dataset{ai_training_datasets_2026,
  title     = {AI Training Datasets: Customer Service Conversations},
  author    = {AI Training Datasets},
  year      = {2026},
  version   = {1.0},
  publisher = {Hugging Face}
}

Contact

For questions or collaborations, please contact us at cybernovasg@gmail.com.

## 💼 Commercial licensing

This dataset is available for free under the MIT License for non-commercial use.  
If you need a commercial license, custom datasets, or the full version with all 600 conversations, contact me at:

📧 cybernovasg@gmail.com