Session
Can You Measure AI Fluency?
Moving beyond access and usage to understand how people actually work with AI
PAWorld26 San Francisco · 23–24 September 2026 · San Francisco
← Back to agenda
– · Plenary
About this session
Tom Zhang will examine how employers can obtain observable, defensible evidence of AI fluency without mistaking confidence, familiarity with one tool or a polished AI-generated answer for genuine capability. Using one non-engineering role and an authentic work task, he will set out a multi-dimensional rubric covering problem framing, tool use, iteration, verification, domain judgement and explanation. He will then address the validity, fairness, accessibility, privacy and accountable-human-review tests that should be met before such an assessment influences live hiring. This is a rigorous assessment blueprint, not a claimed employer case or validated universal score.
Learning outcomes
- Define AI fluency for a non-engineering role in observable, job-relevant terms.
- Use an authentic task and multi-dimensional rubric to gather stronger assessment evidence.
- Identify the validation and governance tests required before using the method in live hiring.
This session will explore
- Why conventional applications, interviews and self-reported AI skills provide weak evidence of real capability.
- How to define AI fluency for one specified non-engineering role and job outcome.
- One authentic work task in which AI use is permitted, relevant and observable.
- A rubric covering framing, tool use, iteration, verification, domain judgement and explanation.
- The validity, fairness, accessibility, privacy and human-accountability questions that must be answered before use.
Context
The second Day 2 keynote: a 20-minute prepared framework followed by a 10-minute moderated challenge, with both keynote speakers joining the shared 10-minute Q&A in 204P.
Speakers and contributors
Tom ZhangPresenterSr. Director, Head of HR/People Ops, NIO USNIO