All micro-internships
Micro-internship brief ≈4h in one sittingno-code

Field-test voice AI on real local tasks

Voice AI still stumbles on local names, accents, and code-switching. Run a structured 10-task field test of a voice assistant and report like a QA engineer.

AI LiteracyContent StrategyConversational AIBusiness Communication

Posted by The AI Internships

The work brief

  1. 01Pick a voice AI you can use (Gemini Live, Google Assistant, Siri, Alexa) and design 10 spoken tasks.
  2. 02Include the hard stuff: place and people’s names from your region, accented or code-switched speech, a local query ("nearest public service office"), and background noise.
  3. 03Run all 10, twice each. Log verbatim what it heard vs. what you said, and whether the action succeeded.
  4. 04Score the results and write your findings: what should a first-time user in your family expect?
What you’ll produce4 deliverables

Submission standard

Submit your 10-task design, the full run log (both rounds), a screenshot of one notable transcription, and your findings. Add a link to a short recording of at least one task so the verbatim heard-vs-said log is verifiable.

  • The assistant tested and your 10 tasks (mark the deliberately hard ones)

    Written responseRequired

  • Full log: what you said, what the AI heard, did the action succeed — 10 tasks × 2 rounds

    Written responseRequired

  • Screenshot of one notable success or failure (the transcription visible)

    Screenshot must show the assistant's own transcription on your device with the app interface visible.

    Image uploadRequired

  • Findings: success rate, failure patterns, and advice for a first-time user

    Written responseRequired

You’ll complete these inside your private workspace.

What you must submit as proof1 proof item · Required

This brief requires evidence an AI can’t fabricate.

  • Link to a short recording of at least one voice task

    An unlisted screen/audio recording (Google Drive/YouTube) of at least one task, showing what you said vs. what it heard.

    Public linkRequired

Submissions without this evidence cannot be submitted.

Protect other people in your proof. Blur faces, names, phone numbers and email addresses before you upload, and refer to anyone you worked with by role or number ("Listener 1", "the stall owner"). Your proof is only ever used to check your work — it is never published, never appears on your certificate, and is never shown in your public portfolio.

How your work is evaluatedPass mark: 70/100

The passing benchmark is 70/100. Final score: Output 50% · AI process 30% · Evidence 20%. Required proof checks must also pass before certification.

Output criteria

Test design

20%

The 10 tasks systematically probe real weak spots (names, code-switching, noise).

Real execution

40%

The log shows verbatim heard-vs-said detail only real runs produce.

Findings

40%

Patterns and advice follow from the log.

Required proof checks

Voice runs really happened

Certification is held for review if there is no working recording, or it doesn't correspond to a logged task (invented heard-vs-said log).

Must pass
How we grade your AI usage30% of your score

Using AI is the point — it’s the skill this certificate proves. You’ll answer three short questions about how you used it: what you asked, what was wrong with its first answer, and what you changed. Specific, honest answers score high. “I pasted the brief and submitted the answer” scores near zero.