Client’s Profile

Asah AI is dedicated to making human-machine interactions more natural and human-like. Their flagship product, DailyFriend, serves as a 24/7 AI companion designed to provide emotional support and assist with personal development.

The Asah AI team comprises experienced professionals with a collective background in developing technology solutions that have reached over 100 million users worldwide. The team is committed to leveraging their expertise to positively impact another 100 million users through innovative AI products. Asah AI’s mission aligns with their vision of creating AI solutions that seamlessly integrate into users’ lives, offering support and companionship through advanced technology.

Client’s Goal

Our client developed an AI-powered voice companion designed to support emotional well-being and personal development. The goal was to ensure the app functioned flawlessly across iOS and Android, providing users with a seamless and engaging experience. Given the app’s reliance on voice interactions and AI-driven responses, robust testing was essential to maintain quality of the app.

Our Contribution

Our team was responsible for implementing end-to-end (both in manual and automation) testing to enhance efficiency, catch critical defects early, and improve overall product stability. We:

• Designed and built an automation framework for both iOS and Android within four months.
• Developed smoke test coverage, including voice injection test cases, ensuring the AI’s response accuracy.
• Used Appium (Python) for mobile(Android and iOS) automation, integrating them with BrowserStack for cloud-based testing.
• Leveraged Proxyman for API testing and help improve existing API testing.
• Conducted black-box testing to evaluate user interactions, ensuring an intuitive and minimum bug experience.
• Use Deepeval as evaluation model tool to assess metrics like hallucinations, bias, relevancy.
• Uncovered important bugs in both the backend and app, for example authentication issues and AI response inconsistencies, which helped the client address key issues before release.
• Exploratory test to eksploit companion’s hallucination using technique like fake fact seeding, round trip prompting, contradicting prior context etc

Example of our contribution

Appium using Python language to automation Android version in Browserstack

Our Manual Test Run example

Test Cases for DailyFriends

Project Outcome

Our automation framework reduced manual testing efforts, allowing for higher test coverage. By achieving stable smoke test automation, we ensured the app’s core functionalities, including voice interactions. As a result, the client had higher confidence in their releases, knowing that key issues were addressed before reaching users.

Need expert testing for your AI-powered app? Let’s work together!