Patented Digital Platform to AI-Assisted Disruption Recovery
Executive Hook
1
Conceived and co-patented Agent on Demand as a cloud-native remote service platform that transformed airport customer support into a globally shared operating model, then extended that platform with an AI-assisted disruption intelligence capability that applied LLM orchestration, GenAI, and retrieval-grounded decision support to help service teams recover travelers more intelligently during day-of-travel disruption.
The Problem We Solved
2
Airport customer service was traditionally constrained by local staffing, physical counters, and station-bound expertise. That model was expensive, operationally rigid, and difficult to scale when disruption or demand spikes hit the network. During irregular operations, customers often needed nuanced recovery options, but frontline teams were forced to work within rigid rebooking logic that optimized for system rules rather than the traveler’s actual outcome.
The deeper problem was not simply service access. It was decision quality under pressure. Standard disruption handling could rebook a customer to the original destination or co-terminals, but it did not reliably optimize for where the traveler truly needed to end up, how quickly they could get there, or what trade-offs they would accept. The opportunity was to redesign the workflow in two steps: first, virtualize service delivery; second, augment that service layer with AI so recovery decisions became more contextual, personalized, and outcome-based.
Technology Deep Dive & Architecture
3
Platform was architected on AWS and Cisco WebRTC (UCCE) as a real-time, responsive web, engagement and analytics platform designed to support high-volume airport customer interactions with strong operational observability. The delivery layer used Java and .NET Core services within AWS networking and load balancing, with event-driven processing patterns to capture agent activity, customer interaction data, and operational signals. Near-real-time metrics were persisted into Amazon RDS for operational visibility, while historical and analytical data flowed into Amazon Redshift and S3 for longer-horizon reporting and downstream analytics in platforms. This separation of live interaction processing from analytical workloads made the platform resilient enough for 24/7 airport operations while still creating a reusable digital data spine.
Get Me Close was designed as the next AI layer on top of that AWS foundation. The architecture combined airline system integrations for PNR retrieval, flight search, and rebooking with Google Maps APIs for airport proximity, distance, and drive-time calculations. The GenAI orchestration layer used Azure OpenAI with function calling, allowing the model to invoke enterprise services rather than generate ungrounded answers. For retrieval and contextual grounding, the design incorporated AWS Kendra for enterprise search, Amazon Neptune for graph-oriented relationship modeling across airports and routes, and Amazon S3 for structured and unstructured context. GPT-3.5 Turbo initially, then GPT-4 were used with prompts structured to gather traveler preferences, evaluate alternate-airport options, and return ranked recovery paths grounded in live enterprise and travel data rather than chatbot-style free text.
AI Transformation Pattern
4
Get Me Close is best understood as a domain-grounded AI copilot for a high-friction operational workflow. When disruption broke the planned itinerary, the system was designed to establish customer context, generate an empathetic response, collect traveler preferences such as earliest arrival or acceptable drive time, and then evaluate alternate options across nearby airports and ground transport trade-offs.
This is the kind of AI transformation pattern that scales across industries. The value is not in a generic chatbot. The value is in applying AI to a live decision bottleneck, grounding the model in enterprise data, constraining it through orchestration and retrieval, and keeping humans in the loop for execution. That is what makes the capability credible in a mission-critical environment.
Operational Resilience
5
Agent on Demand created resilience at the operating-model level by decoupling customer support capacity from local station staffing and allowing demand to be served through a shared digital resource pool. The architecture further supported continuity by preserving real-time responsiveness while maintaining historical visibility for analysis, replay, and optimization.
Get Me Close extended that same resilience principle into the decision layer. During severe disruption, the challenge is not only transaction volume but cognitive overload. AI-assisted disruption workflows reduce the service time and inconsistency involved in evaluating recovery alternatives, helping teams make better decisions faster while staying grounded in enterprise context.
6
Cross-Functional Leadership
This initiative required alignment across airport operations, customer service, digital product, engineering, enterprise platforms, and analytics. The work sat at the intersection of service model redesign, cloud architecture, operational reliability, and AI enablement.
Enterprise AI execution requires the ability to identify a high-value workflow, architect the digital foundation, align stakeholders around measurable business outcomes, and evolve the platform toward AI-assisted decisioning without losing control of reliability, governance, or customer experience.
Phased Rollout & Delivery
7
Agent on Demand moved rapidly from concept to production, demonstrating the ability to identify an urgent business problem, frame a differentiated solution, align teams, and execute in a live environment where speed mattered. The platform was built through agile delivery, then scaled as a new operating model for airport customer support.
Get Me Close represents the next phase in that evolution. It was observed within airport operations as a use case for dynamically assessing needs of customers in wait queues and using data alongside urgency to deliver differentiated outcomes. Achieved via prototype-to-platform extension: define the AI use case, identify the integration points, establish enterprise data access, and build the path from design to governed rollout. That pattern—digital first, AI second, scale third—is the same pattern that underpins effective enterprise AI transformation.
What’s Next / The Legacy
8
The lasting value of Agent on Demand was not only that it reduced cost and improved service quality. It created a digital operating layer onto which progressively smarter capabilities could be added. Get Me Close demonstrates that evolution clearly: remote assistance became AI-assisted disruption recovery, and a service channel became a platform for contextual decision support.