AfterMe

Inspiration

Most assistants are good at answering questions, but much less useful when the user simply says, "Make sure this gets done." Real-life goals are rarely a single task. Preparing for a trip, getting a house ready for guests, or preparing for an important presentation involves multiple steps, decisions, reminders, and follow-ups. We wanted to explore a different kind of assistant: one that understands the outcome the user wants, figures out the work required, takes action where it can, asks for approval when necessary, and remembers the context for later. That idea became AfterMe.

What it does

AfterMe is a web-based simulated Alexa+ trip-planning experience. The prototype shows an example Delhi trip plan with five demo actions. Users can run simulated weather, packing, and reminder actions, while ride and home actions wait for approval. When the user says their parents are coming too, the demo updates the traveler count and packing list. The conversation and action status are saved in the current browser. No real-world service is contacted.

How we built it

AfterMe uses Next.js, React, and TypeScript for its web interface and plan logic, with Lucide React for icons. The current prototype uses simple rules and browser storage to demonstrate the conversation, plan updates, and action approvals. It does not connect to a live Alexa+ service, calendar, travel provider, smart-home system, or AWS service.

Challenges we ran into

The biggest challenge was deciding how much autonomy to give the agent. An assistant that only responds to questions is limited, but an assistant that blindly executes every action is difficult to trust. We therefore designed AfterMe around explicit action boundaries: low-risk actions can be handled automatically, while actions requiring user involvement are surfaced for approval. Another challenge was maintaining context across multiple interactions. A useful agent should not forget the original goal every time the user adds a new piece of information. We also had to balance ambitious agentic behavior with the limited time available during the hackathon, prioritizing a polished end-to-end experience over a large number of disconnected integrations.

Accomplishments that we're proud of

We’re proud of making the simulated experience clear and interactive. The demo shows a trip plan, distinguishes actions that can run from those that need approval, and updates the traveler count and packing list when new information is added.

What we learned

We learned that building an agent is less about making the model produce better text and more about designing the system around the model. Planning, state, tools, permissions, execution, and user trust all matter. We also learned that the best agent experiences start with the user's desired outcome rather than forcing the user to think in terms of individual commands.

What's next for AfterMe

AfterMe could eventually connect to more of the services people already use every day, including calendars, email, travel services, messaging, smart-home devices, and purchasing workflows. The longer-term vision is an assistant that can take responsibility for an outcome across multiple services while keeping the user in control of important decisions. AfterMe starts with a simple idea: Don't just answer the user. Help get it done.

Built With

• alexa+
• app
• next.js
• personal-assistant
• productivity
• react
• typescript
• web