Enhancing Customer Support with Predictive Forecasting: Leveraging Databricks ai_forecast

By: Vik Malhotra

Co-Authors: Animesh Raj Jha and Sunmin Lee

In our previous blog, we explored how Databricks AI functions (ai_classify, ai_extract, ai_gen) streamline support ticket management and unlock actionable insights for business teams. Building on that foundation, this post dives deeper into predictive forecasting — using the powerful SQL **ai_forecast** function—to help anticipate trends and optimize operations before issues arise.

Dashboard

Why Predictive Forecasting Matters

Modern IT and support operations are not just about reacting; they’re about planning ahead. Predictive analytics allow teams to anticipate spikes in ticket volume, forecast resolution demands, and align staffing or technical resources for peak periods. With Databricks’ table-valued function **ai_forecast**, generating future projections from historical support ticket data is now fast, scalable, and accessible directly within SQL Warehouses.

Getting Started with ai_forecast

The **ai_forecast** function takes your time series ticket data and produces forecasts for the desired future horizon. It’s remarkably simple—just specify your table, time column, value column (such as ticket counts), and how far ahead you wish to predict. This abstracts the heavy lifting of model development, letting data teams focus on action over configuration.

Sample SQL snippet:

-- An example of Ticket Volume Forecast
WITH daily_ticket_counts AS (
                      SELECT
                        DATE(created_date) AS ds,
                        COUNT(ticket_id) AS ticket_count
                      FROM quickstart_catalog_vkm_external.classify_tickets.raw_tickets
                      WHERE created_date IS NOT NULL
                      GROUP BY DATE(created_date)
                      ORDER BY ds
                    ),
                    max_date AS (
                      SELECT MAX(ds) AS last_date FROM daily_ticket_counts
                    ),
                    horizon_date AS (
                      SELECT DATE_ADD(last_date, 7) AS horizon FROM max_date
                    )
                    SELECT 
                      'Ticket Volume Forecast' as forecast_type,
                      'Next 7 Days' as forecast_period,
                      *
                    FROM ai_forecast(
                      TABLE(daily_ticket_counts),
                      horizon => (SELECT horizon FROM horizon_date),
                      time_col => 'ds',
                      value_col => 'ticket_count'
                    )

With grouping capabilities, forecasting by priority, affected system, or team is also possible, powering more targeted insights.

Dashboard Insights: From Past Data to Future Trends

Our enhanced Streamlit dashboard now features an “AI Forecasting” tab, visualizing projections for several key metrics:

AI Forecasting

• Ticket Volume Forecast: Anticipate load and resource needs.

Ticket Volume Forecast

• Urgent Tickets Forecast: Plan for critical incident management.

Urgent Tickets Forecast

• Resolution Capacity Forecast: Align staffing and response times.

• Team Workload, Priority Distribution, and System Issues Forecasts: Help leaders balance assignments and prevent bottlenecks.

Team Workload Forecast

Visual forecasts:

• Interactive selection of forecast types
• Line charts with confidence intervals, revealing best/worst scenarios
• Immediate, business-ready dashboards suitable for planning meetings or executive briefings

These predictive dashboards transform historical trend analysis into forward-looking intelligence, enabling proactive support and technology operations.

Real-World Impact and Takeaways

By plugging Databricks **ai_forecast** into your support data workflows, teams can:

• Plan for future spikes before they happen
• Reduce resolution bottlenecks
• Optimize spending and staffing
• Create business scenarios for incident, customer, or system management

This hands-off, SQL-driven approach democratizes predictive analytics, putting advanced forecasting in the hands of support leaders and analysts — not just data scientists.

Next Steps

• Try **ai_forecast**, review the official documentation, and integrate forecasting SQL into your own Databricks Warehouse.
• Explore the enhanced dashboard: Experiment with different forecast types and parameters to uncover hidden trends.
• Share your feedback: How would predictive forecasting change your support strategies?

References

• Original blog: Automated Ticket Classification with Databricks AI Functions
• ai_forecast function documentation
• For this project, you can find all the latest predictive forecasting code in the **predictive branch of the databricks-classify-tickets GitHub project**.