Most people think AI fluency means writing a good prompt. It does not.
Prompting is a tool skill. Fluency is a decision skill.
Fluency is a decision skill
The 2027 Skills Horizon by The University of Sydney describes it as the foundation “to select the right system for the job, to know the difference between agentic and AI automation, and to anticipate AI cost now that you pay per use.”
Notice what is missing in that sentence? There is no mention of clever prompts. It is all human choices.
The report also asks us to avoid this question: “’Where does AI fit?’ is the wrong question to ask when you’re thinking about work design. Your day isn’t one continuous thing, so AI’s usefulness varies wildly across it.”
It splits the working day into four parts. Each one needs a different answer.
Plan. Calendars, inboxes, forms. The advice is to use AI “to compress this work, not to do more admin faster.”
Learn. Reports and briefings. “Productive struggle is the point. Use AI to support, not replace your learning.”
Make. Decks, spreadsheets, strategy papers. “Treat AI as a sharp junior colleague: hand over chunks but keep your hand on the wheel.”
Play. Brainstorming and testing ideas. “Make sure you’re the one doing the thinking.”
Same tool. Four different rules. That is fluency. And most of us are not there yet.
The report says most use is still shallow: “a prompt here, a draft there, invented from scratch each time and invisible to everyone else.”
The training gap is real. Only 24% of Australian workers have received AI training (Melbourne University, 2025).
88% of workers say they need major up-skilling in the next five years. Only 36% feel adequately skilled (BCG, 2026).
51% of AI users do not always know when to trust AI outputs and when to verify them (TeamViewer, 2026).
We are handing people a powerful tool with no map.
How we measure AI capability at Gradstack
We spent the last year building the map. Our Gradstack AI Fluency Index (GAFI®) measures AI capability across five categories.
It runs on our proprietary AI Capability Model, shaped by a team with 30+ years of combined experience across AI, machine learning, software engineering and people capability.
Here are the five.
AI Awareness. A foundational understanding of how AI works, its limits and its context.
AI Application. Practical, day-to-day use of AI in tasks, prompts and workflows.
AI Ethics. Privacy, bias, disclosure and the responsible use of AI.
Human-AI Collaboration. Task delegation, independent thinking and working with AI in a team.
AI Adaptation. Adopting new tools, transferring skills and building AI capability over time.
Read them as a journey. From AI Awareness to AI Adaptation.
Awareness is knowing how AI works and where it fails. Adaptation is the continuous learning that must occur to stay relevant, up to date and capable as the tools change.
And the tools change fast. So the ability to keep learning matters more than any one tool!
Everyone who completes GAFI® earns a GAFI® Certified badge at their level. A verified signal, not a claim on a resume with no proof.
Why we never rely on self-rating
Here is what we see in our GAFI® assessment data.
50% of the people we assess rate their AI capability about twice as high as they actually score.
That is a true blind spot, but you cannot fix what you cannot measure.
So we assess what people can do, not what they say they can do.
Because you cannot build a stronger workforce on a skill you cannot see or measure.
Source: The 2027 Skills Horizon, Sydney Executive Plus, The University of Sydney (plus.sydney.edu.au/horizon), pp. 38, 57, 58 and 103. Underlying research as cited in the report. Self-rating statistic: Gradstack GAFI® assessment data.