Home / Work / Case 02 · Humanizing AI Usability Testing
02 Human–AI Interaction · Usability 4 min read

Humanizing
AI Usability
Testing

How we harnessed AI to amplify, not replace, the human touch, and surfaced the Authenticity Paradox that reshaped product strategy before a single line of code shipped.

Method
Unmoderated
usability
Sample
n = 10 caregivers
Preference
9 of 10 chose AI
Outcome
Automation → collaboration
Role
Sole Senior
UX Researcher
Method
Unmoderated
Usability
Sample
n = 10 caregivers
Stimulus
High-fidelity
Figma prototype
Impact
Product pivot,
pre-engineering
The Problem

A high-stakes trust moment.

Incomplete caregiver profiles were a major bottleneck in our marketplace, leading to fewer successful matches. We saw an opportunity to leverage AI to help caregivers craft compelling bios at scale. However, this introduced a high-stakes risk: the care industry is built on human trust and authenticity.

The challenge wasn't technical; it was deeply human. My task was to answer a single, critical question:

The research question

How do we harness the efficiency of AI without sacrificing the authenticity that is the very currency of our platform?

My Approach

Rapid, behavioral, unfiltered.

To get fast feedback on a genuinely disruptive technology, I designed an unmoderated usability test with 10 caregivers using a high-fidelity Figma prototype. This method was the fastest way to observe visceral, unfiltered reactions to using AI for such a personal task.

The study assessed usability, effectiveness, and overall perception of the AI tool, from the initial questions through to the final consent screen.

The 'Aha!' Moment

The Authenticity Paradox.

An overwhelming 9 of 10 caregivers preferred the AI-assisted option. But they fundamentally rejected the idea of AI as the author.

They wanted a partner, not a ghostwriter.

Preferred AI
Did not
What I Heard

Two truths, held simultaneously.

Caregivers held two contradictory beliefs in the same breath: AI was a powerful efficiency tool, and AI was a threat to the personal voice their livelihood depends on. The product strategy had to honor both.

The Value

AI as an efficiency catalyst.

Caregivers saw the AI as a powerful tool to overcome writer's block and save time in their demanding schedules. They viewed the generated bio not as a finished product, but as an excellent "starting point" to build upon.

My first impression is that it's great. It's a great starting point but it doesn't have my personal flair to it… I hope that there is a space for me to edit this.
The Risk

AI as an authenticity threat.

The greatest fear was that AI would create a "generic" persona, stripping the unique voice that builds trust. The label "Crafted with help from AI" could become a scarlet letter, undermining credibility and making families question honesty.

Concerned about losing the voice and personality, which is valued in the care industry… but definitely more favorable with the time-saving aspect.
Recommendations

From automation
to collaboration.

My findings provided a clear strategy to resolve the paradox: empower the caregiver to collaborate with the AI, giving them full control and ownership of the final product.

01
Principle

Empower, don't automate.

Position the AI as a starting point, not a finished product. Make editing the default action, not an exception. Every generated bio should feel like a draft the caregiver is finishing, not a result they're approving.

02
Inputs

Deepen inputs for richer outputs.

The richness of the output depends on the richness of the input. Ask fewer, deeper questions that surface the personal flair caregivers worried they were losing, hobbies, philosophies, signature phrases, specific moments with clients.

03
Trust

Remove the AI stigma.

Drop the "Crafted with help from AI" label. Once a caregiver has edited and approved the bio, it is theirs. Branding AI involvement risks signaling inauthenticity, and erodes the trust this feature exists to build.

Impact

A product pivot, before a line of code.

This rapid usability study provided critical insights that fundamentally shaped the product's direction before a single line of code was written.

My research directly influenced the strategy to pivot from a simple "automation" feature to a more sophisticated "collaboration" tool. The user-centered approach de-risked the project by ensuring the final feature would enhance, not compromise, the authenticity that drives trust and successful matches.

Why this transfers

Research the behavior, not the technology.

The Authenticity Paradox I surfaced here, users wanting AI collaboration, not AI authorship, is the defining tension in enterprise AI adoption.

The same trust dynamics that govern caregiver bio generation govern how knowledge workers adopt AI-assisted drafting, decision support, and agentic tools. This study demonstrates the ability to research novel AI behaviors before established frameworks exist, the exact capability senior AI product teams are hiring for right now.

← Previous case

De-Risking the 2026 Monetization Roadmap

Next case →

Unlocking Conversion through Conjoint

Previous
Previous

De-Risking the 2026 Monetization Roadmap

Next
Next

Unlocking Conversion