AWS Applied AI Solutions GeoStrands Delivery Agentic Teammate

At the end of 2025, the Geospatial team approached the Applied AI Solutions (AAIS) UX team to request help preparing for a high level review. We had just over one month to create an example application for their vision of the future. This application is speculative in nature and not intended as a real product.

The focus of the AIIS UX org since mid 2024 has been to re-imagine how we present AI based designs to users. The goal for each project is to create an agentic teammate that handles work the way a human teammate would.

A human teammate and I worked on this project together, looking at the workflow for a fictional last mile delivery service. She tackled the planner (not linked, but in the prototype) workflow, while I handled the dispatcher workflow. The day to day job of a dispatcher on last mile delivery is largely spent communicating with drivers and fighting fires. Our AI Agent, Geovanni (Geo, get it?) handles the day to day issues. As you start the prototype, Geovanni informs you that everything is running smoothly. Soon, though, things start to happen (this is not shown in real time).

Image of a map depecting delivery routes and vehicle locations. — A traffic disruption begins to block traffic in north Seattle. The agentic teammate is aware, and is able to reroute vehicles as the disruption grows. It knows their routes and if the vehicles will be impacted.

Geovanni detects a traffic disruption in North Seattle. At first it’s not a big deal. Slowly, the disruption spreads. Geovanni knows the routes of the vehicles and knows which vehicles will be impacted by traffic.

Image of the AI agent surfacing an issue to the user. — Geovanni, the agentic teammate, asks escalates a decision to the human in the loop.

Soon, traffic is severely impacted. Geovanni notices that one of the vehicles is carrying perishable items, and if it is delayed by traffic the items could spoil. It escalates the situation to the human dispatcher, offering options to solve the problem.

Geovanni will offer to remember the user’s choice so that it can take the same action in a similar scenario in the future, allowing the teammate to grow in usefulness to the people who work with it.

Geovanni is always available in the chat, but, this demo wasn’t made to be very chatty. Newer demos have built in chat bots that can respond and role play.

Ask me about the easter egg in this prototype. It’s a fun little secret I added before we finished.

Image of the agent announcing the daily deliveries are complete. — At the end of the day, Geovanni announces the results. Today, everything made it where it was supposed to go, despite the difficulties.

Prototype

The prototype is a web application built using the Kiro IDE.

View The Prototype

User Experience Designer

AWS Applied AI Solutions - Geospatial

Early 2026

Challenge: Design an agentic teammate to handle mundane delivery tasks, allowing the human distpacher to focus on higher level issues.