AWS re:Invent 2025 - Manchester Airports Group: Scaling operations with Agentic AI & MCP (SPS327)

AWS re:Invent 2025 - Manchester Airports Group: Scaling Operations with Agentic AI & MCP

Overview

• Manchester Airports Group (MAG) is the largest airport group in the UK, operating three major airports.
• MAG is on a journey to become the "world's most intelligent airport group" by leveraging Agentic AI to automate and optimize their complex operational processes.
• The presentation covers MAG's first Agentic AI use case focused on automating unplanned employee absence reporting and rostering.

Operational Challenges at MAG

• MAG faces significant operational complexity in coordinating 9,000 staff and 40,000 personnel across its airport campuses.
• Maintaining a "frictionless yet secure" passenger experience requires constant monitoring and management of various interdependent systems and partners (MAG staff, border force, airlines, ground handlers, etc.).
• Passenger safety and security is the top priority, often leading to higher operational costs.
• The COVID-19 pandemic greatly exacerbated these challenges, with only 1-2% of flights operating globally at one point.

Vision for a Digital Colleague Workplace

• MAG's vision is to create a "single unified experience for our colleagues powered by agents, removing a significant amount of the complexity in order to deliver value for MAG."
• This involves applying Agentic AI across a variety of use cases, starting with unplanned employee absence reporting and rostering.

Why Agentic AI?

• Agentic AI provides the necessary capabilities to handle MAG's complex, non-deterministic operational processes with dynamic coordination across multiple systems and functions.
• Generative AI assistants and automation tools lack the flexibility and reasoning abilities to adapt to the many exceptions and edge cases encountered in MAG's environment.
• Agentic AI allows MAG to iteratively build towards their vision of an "intelligent airport" where multiple AI agents can interact to optimize operations.

Unplanned Absence Reporting Use Case

• The first Agentic AI use case focused on automating the process of unplanned employee absence reporting and rostering.
• Key challenges included ensuring secure authentication, accurate recording of absence details, and reliable 24/7 availability - all while adhering to complex HR policies.
• The current manual process involved multiple steps, different procedures across locations, and required significant involvement from line managers and HR teams.

Technical Implementation

1. Adapting Tools for Agents: Ensured MAG's existing tools (APIs, knowledge bases, notification systems, etc.) were optimized for use by Agentic AI agents, including:
   • Implementing verbose error handling
   • Providing human-readable outputs
   • Combining sequential tool calls to reduce latency
2. Agentic AI Architecture:
   • Utilized Amazon Bedrock Agent Core Runtime to host the Agentic AI agent and provide scalability and security.
   • Developed a Model Context Protocol (MCP) server to decouple the agent logic from the underlying tools, enabling reusability and extensibility.
   • Integrated Amazon Lex (speech-to-text) to provide a conversational interface for employees to report absences.
3. Guardrails for Responsible AI:
   • Implemented custom, context-aware guardrails to ensure the agent's interactions remained within defined boundaries, considering the multi-turn nature of conversations.
   • Leveraged language models to detect and prevent potential misuse or social engineering attempts.
4. User Experience Design:
   • Recognized that the powerful backend capabilities alone were not enough, as users expected a responsive and engaging experience.
   • Designed a user interface that provided real-time feedback on the agent's progress, keeping the employee informed and engaged throughout the process.

Results and Business Impact

• Standardized the absence reporting process, achieving 99% consistency in data capture.
• Reduced the time to record an absence by 90%, unlocking opportunities for deeper analysis of absence patterns.
• Enabled the ability to dynamically re-roster staff, reducing overtime costs and improving operational efficiency.

Future Expansion

• MAG plans to expand the use of Agentic AI to automate additional operational processes, such as automated rostering, terminal capacity management, and more.
• The goal is to create a more autonomous, efficient, and effective airport operation by leveraging a growing ecosystem of Agentic AI agents and tools.

Key Takeaways

1. Adapt Tools for Agents: Optimize existing tools and systems to work seamlessly with Agentic AI agents, improving performance and reliability.
2. Implement Responsible Guardrails: Develop custom, context-aware guardrails to ensure Agentic AI agents operate within defined boundaries, especially for high-impact, mission-critical applications.
3. Prioritize User Experience: Recognize that a powerful backend alone is not enough; a responsive, engaging user experience is crucial for the successful adoption of Agentic AI solutions.
4. Start Small, Scale Fast: Begin with a focused use case that demonstrates the value of Agentic AI, then iteratively expand the solution to address additional operational challenges.