Portfolio — Case Study: Destination Voice Bot

Context

A major automotive client was developing an in-car voice bot to replace traditional call-center routing for navigation support. Rather than drivers calling in for directions or destination help, the bot would handle the entire interaction hands-free, in the vehicle.

I led usability research to evaluate the prototype's conversation design, identify where the experience broke down, and surface what else drivers expected a navigation assistant to be able to do.

Research Goals

Testing Communication, Breakdowns, and Real‑World Expectations

Goal 01

Evaluate Ease of Communication

How easily can drivers communicate with the bot in realistic driving conditions? Are the prompts intuitive, and does the bot respond in a way that feels natural?

Goal 02

Investigate Fallbacks

Where does the conversation break down? Which parts of the flow cause frustration or confusion?

Goal 03

Uncover Use Cases

What else do drivers expect from a navigation assistant, beyond the tasks we built for?

Approach

Evaluating the Voice Experience

Participants

10

Test rounds

2

Timeline

5 weeks

Moderated
usability testing

Remote

I recruited 10 participants and ran two rounds of remote moderated usability testing over five weeks. Participants performed tasks using the voice bot on mobile to simulate in-car use and were asked follow-up questions to understand their experience. I also asked about prior experience with voice assistants to understand expectations coming in.

Remote moderated sessions — participants interacted with the voice bot on mobile to simulate in-car conditions.

Example Usability Tasks

01

Basic Destination Request

"Imagine you are out driving and want to go to the nearest Starbucks. How would you ask the Destination agent for directions?"
02

Mid-Route Destination Change

"After getting directions to Starbucks, you also want to get directions to the nearest Target. How would you perform this task?"
03

Escalation to Live Agent

"Imagine you are asking for directions to Starbucks and the Destination assist is not working as expected. What would you say in order to get help from a live agent instead?"

Add script image

Conversation flow diagram mapping the bot's decision paths, fallback states, and escalation logic.

Key Insights

What Users Experienced

1

Conversation Changes Were Handled Well

Participants thought the bot handled fallbacks efficiently like changing destinations in the middle of a conversation and asking to quickly transfer to a live agent.

2

Perceived as Faster & Safer Than Typing

Speaking to the bot felt easier than typing on a phone or screen, especially if you were to be in an urgent situation.

3

Need for Destination Context

Users wanted the bot to confirm destinations with estimated time, distance, and other details before starting navigation. This would help provide more context when driving in unfamiliar areas.

4

Strong Appetite for Feature Expansion

Users described wanting to ask about business hours, nearby gas stations, attractions, and parking. These use cases reflected how the voice bot could enhance their navigation experience.

Outcomes

What the Research Produced

🚀

Launch Readiness Confirmed

The client planned to launch the voice bot to millions of their car owners. Research validated that the core conversation design was strong enough to ship, with targeted copy improvements identified before launch.

🗺️

Prioritized Feature Roadmap

Research produced a ranked list of future use cases drawn directly from user sessions, giving the product team a clear roadmap for improvement and new features.

⚡

Faster Iteration Through Dev Presence

Having developers observe sessions live allowed issues to be addressed between test rounds rather than waiting for the full synthesis report. This shortened the feedback loop significantly.

✏️

Revised Conversation Copy

I worked with the Convo UX Architect to rewrite the voice bot's dialogue based on research insights. Changes were validated in round two.

Reflection

What I'd Do Differently

If I were to do this project again under different circumstances, being able to test the prototype in a car or while driving would be the biggest opportunity. There are so many environmental factors to consider like background noise, other passengers, music, the cognitive load of driving and using the voice bot at the same time, etc. that could add more context to the research insights.

Next Case Study

Employee Loan Refinancing Experience

Re-designing an internal loan-refinancing system through discovery and evaluative research.

View Case Study →