AWS re:Invent 2025-The AI revolution in customer support: Building predictive service systems-SPS315

Transforming AWS Support with AI: Building Predictive Service Systems

The Need for Transformation

• AWS support has traditionally operated in a reactive, "firefighting" mode, responding to customer issues as they arise
• This model is not scalable as cloud complexity and customer demands have increased
• Customers want near-zero downtime, maximum resilience, and zero-impact maintenance
• Reactive support is expensive and time-consuming, leading AWS to shift towards a more proactive, preventative approach

Evolving AWS Support with AI

• AWS is leveraging AI and automation to transform its support operations and deliver better customer experiences
• Key goals:
  • Detect issues before they impact production
  • Partner with customers to build resilient architectures
  • Use AI for accurate, contextual troubleshooting and faster incident response
  • Provide proactive guidance and security monitoring

Unified Operations: Delivering Mission-Critical Support

• Unified Operations is AWS's new top-tier support plan for business-critical workloads
  • Provides a dedicated team of engineers, TAMs, and incident managers
  • 5-minute response time for critical incidents
  • Leverages context about customer environments and workloads
• Focuses on:
  • Proactive risk management and resilience testing
  • Tabletop exercises and game days to identify gaps
  • Contextual incident management and retrospectives

Whoop's Journey with Unified Operations

• Whoop, a health/fitness wearable company, partnered with Unified Operations to prepare for a major product launch
• Engaged in a risk assessment exercise to identify potential issues
  • Allowed Whoop to scale up and stress-test their systems
  • Provided access to AWS experts for real-time guidance and troubleshooting
• Achieved 100% availability during the launch and subsequent peak periods
  • Reduced downtime by 70% in Q3 through faster incident response

Leveraging AI and Automation

• AWS has developed a multi-tiered approach to AI-powered support:
  • Chatbots and knowledge bases for basic assistance
  • AI-augmented human experts for complex troubleshooting
  • Autonomous AI agents for proactive monitoring and recommendations
• Key technologies:
  • Structured runbooks and workflows for deterministic automation
  • Graph-based knowledge retrieval and fine-tuned models for contextual insights
  • Orchestration layer to coordinate multiple AI agents and human experts
  • Automated reasoning checks for safety and compliance

The AWS DevOps Agent

• Newly launched tool that integrates with AWS Support
• Extracts application topology and relevant signals to provide root cause analysis
• Leverages insights from past issues to proactively recommend preventative actions
• Allows seamless escalation to human experts when needed

Key Takeaways

• Transforming support operations requires a shift from reactive to proactive, preventative approaches
• Leveraging AI and automation is crucial to scale support and deliver better customer experiences
• Context and historical data are key to providing accurate, actionable insights
• Partnering with customers and providing dedicated, expert-led support can drive significant business impact
• Continuous improvement and learning from past incidents are essential for optimizing support systems