AI Assistant for Google Maps
At a glance A personal exploration of how an AI assistant could live inside Google Maps to help users decide where to go using saved lists, reviews, and real-time context.
↓ Jump to solutionProblem
Living in New York City, I rely on Google Maps' saved lists to keep track of places I want to try, organized by category (cafes, food, bookstores, and so on). Over time, those lists grew into hundreds of saved places, and instead of helping me decide where to go, they started creating decision fatigue. I'd scroll through long lists, forget why I saved certain places, and still end up manually checking reviews, ratings, and distance before deciding.
"What if Google Maps had an AI assistant that could help me decide where to go in the moment?"
Through a mix of AI interaction testing, Maps interface analysis, user conversations, and competitive review, four core insights emerged:
- Decision fatigue inside saved lists: Users see walls of saved places with no way to compare across them.
- Discovery, not decision-making: Maps surfaces new places well but doesn't help users choose between multiple saved ones.
- Forgotten saves: Users save places, then forget why. Newer saves quickly bury older ones.
- Untapped opportunity: Saved lists could anchor a more contextual, explainable AI experience.
Solution
An AI assistant inside Google Maps that helps users make faster, more confident decisions about where to go. The assistant supports three distinct entry flows that meet users where they already are in Maps and converge into the same AI experience.
Flow 1: General Query
From the main Maps screen, with a general topic in mind but no specific place picked yet.
Flow 2: Saved Lists
The user wants somewhere to study and eat, with a few saved lists that could help.
Flow 3: Specific Saved List
Already inside a saved list, looking to narrow down within it. This flow shows how the AI can sort through list tags within a saved list to surface the right pick.
Three core design moves
Across all three flows, three design decisions shaped how the assistant feels native to Maps:
Process
Research: AI patterns, Maps analysis, and user insights
I started from a personal pain point and grounded it in four parallel investigations, focusing on what users do today and where AI could plug in.
Synthesizing the findings, the goal became designing an AI assistant that supports contextual, explainable decision-making using saved lists, reviews, and real-time user needs.
Ideation: Entry flows, user flow diagrams, and lo-fi sketches
Research showed that users don't start a decision the same way every time. To support this, I mapped three distinct entry flows, each meeting users where they already are in Maps, with no new navigation patterns required. From there, I drew flow diagrams and lo-fi sketches to make sure the conversation pattern, ask, refine, recommend, act, held up across all three entry points.
General Query (Main Maps screen)
Open-ended exploration with no specific place in mind. The AI button lives on the main Maps screen, lowering the barrier to first use.
Saved Lists screen
Multiple saved lists, not sure which has what they need. Where decision fatigue is highest, and the AI bridges the gap across lists.
Inside a Saved List
Already in a list, narrowing down. The most context-rich entry point, the AI knows the list scope and can refine immediately.
Design: Hi-fi prototype
With the flows validated, I moved into Figma to build the hi-fi designs and an interactive prototype. The goal was to make the assistant feel like it belongs inside Google Maps: borrowing existing card styles, iconography, and palette throughout, so the chat reads as a Maps capability rather than a separate product.
Reflections
Learnings
This project pushed me to think deeply about how AI-powered features actually work, not just how they look, but how a conversational assistant should respond, surface information, and guide decisions. I also learned a lot about integrating AI into familiar UI patterns so the experience feels native, not like a separate product.
Challenges
The biggest challenge was finding the right balance between useful information and simplicity. I wanted recommendation cards to be rich enough to help users decide, but not so dense that they felt overwhelming. The same tension came up with explainability, how much reasoning is helpful versus how much just adds clutter.
If I had more time
I'd expand the assistant beyond single decisions to full trip planning, going from "where should I eat tonight" to "plan my day in Tokyo using my Japan saved list." I'd also flesh out chat history, the ability to edit queries after submitting, and other interaction refinements.
Takeaway
Conversational AI could help Maps evolve from a navigation tool into a decision assistant, one that doesn't just show you where things are, but helps you decide where to go. Google's release of Ask Maps validated that this is a real problem space worth designing for.
Other Projects
Phia x Design Meetup: Designathon Finalist
Designed for Phia, an AI-powered fashion shopping app, as a designathon finalist with work from the semifinals and finals.
Yammii: Online Order Revamp
Simplified and modernized Yammii's mobile order flow to create a more visually engaging and faster experience.
Hemi: A Personal Reflection Journal
Designing and building a digital journal in Claude Code, from a paper planner sketch to a fully deployed web app.