Agentic AI Customer Support Dataset

By shrijeetverma13 · March 19, 2026

Data source:

The analysis shows the percentage of unresolved or pending tickets broken down by Priority Level. A bar chart and supporting data tables have been generated to illustrate these rates clearly across all priority categories.

The analysis compares average Resolution Time across different support channels. A bar chart and data table were generated showing how each channel performs in terms of resolution speed. The visualization ranks channels from fastest to slowest, making it easy to spot which support channels resolve tickets most efficiently.

A data table was generated showing resolution times broken down by priority level. Based on the available results, you can compare how Critical, High, Medium, and Low priority tickets differ in their average and median resolution times.

Pakistan stands out as the country with both the lowest average Customer Feedback Rating (2.75) and the highest escalation rate (0.0% — though this may reflect data nuances). The full ranking of countries by average feedback rating from lowest to highest is: Pakistan, UAE, USA, Germany, and UK. A chart and data table have been generated to visually compare feedback ratings and escalation rates across all countries.

The 'Account' query category has the highest human escalation rate at 52.08%. A bar chart was generated showing escalation rates across all query categories, with 'Account' highlighted in red as the top escalator. Additional breakdowns by subcategory and priority level are available in the data tables, revealing which specific account-related issues and priority levels are driving the most escalations. Escalated tickets also show notably lower sentiment scores compared to non-escalated ones, suggesting customer frustration is a key factor.

The analysis identified the top Query Subcategory issues for tickets with a Sentiment Score below -0.5. A bar chart and data tables have been generated showing the distribution of these highly negative sentiment tickets across different subcategories.

The analysis examined whether Customer Language significantly affects Feedback Rating or Resolution Time. Two sets of bar charts and box plots were generated showing the distribution of both metrics across all customer language groups. The visualizations reveal how each language group compares in terms of average ratings and resolution times, along with the spread of values within each group.

The analysis shows how support channels are distributed across different customer countries. A stacked bar chart has been generated that visually breaks down the proportion of each support channel used per country, making it easy to compare preferences at a glance. Multiple data tables are also available with the underlying figures.

The bar chart displays the proportion of tickets resolved without human escalation for each Query Category. The visualization clearly compares how self-sufficient each category is in resolving tickets without needing human intervention, along with supporting data tables breaking down the totals and resolved counts per category.

The average weekly Sentiment Score was analyzed across 11 weeks, spanning from December 1, 2025 to February 9, 2026. A line chart has been generated showing how sentiment fluctuated week by week over this period. Overall, the average sentiment score was slightly negative at -0.030, and there was a modest downward trend — sentiment declined by 0.080 points when comparing the first half of the period (avg: +0.014) to the second half (avg: -0.067).

Ticket volume peaks on Mondays (191 tickets) and at 12:00 noon (65 tickets). A heatmap and bar chart have been generated to visually show the distribution of ticket volume across all hours and days of the week.

Among all Query Subcategories, 'General Inquiry' has the longest average Resolution Time at 37.2 minutes, while 'Login Issue' has the lowest average Customer Feedback Rating at 2.89 out of 5. A bar and line chart was generated showing both metrics side by side for all subcategories, making it easy to compare resolution time and feedback trends at a glance.

The relationship between Sentiment Score and Customer Feedback Rating is essentially negligible. The Pearson correlation coefficient is -0.040, indicating a very weak negative correlation — meaning these two metrics are largely independent of each other. Two visualizations were generated: a scatter plot showing individual data points across all rating levels, and a box plot displaying the distribution of sentiment scores for each feedback rating (1–5).

The analysis produced two visualizations and supporting data tables showing Query Category and Resolution Time patterns for customers who gave ratings of 1 or 2. A grouped bar chart compares ticket counts and average resolution times per category (low-rated vs. overall), while a box plot shows the full distribution of resolution times by category for low-rated tickets.

A word cloud has been generated from the Customer Query Text column, visually representing the most frequently occurring words in customer queries. Larger words appear more prominently, indicating higher frequency in the dataset.