by Tom Mendham
An assignment can be fluent, well structured and confident, and still leave the real question open: what can this student actually do? Generative AI has sharpened that question in every discipline. Polished output is now easy to produce, which makes it harder to see what a learner independently understands and can apply.
The instinctive response, detecting and restricting AI use, addresses integrity but not evidence. Knowing how a document was produced still does not show what its author has learned. The more durable response is to ask for evidence that polish cannot substitute for: a skill demonstrated, reasoning explained aloud, an approach adapted when conditions change, a stronger attempt after feedback. When learners show their work in these ways, the learning itself becomes visible.
None of this requires abandoning written work. Essays, problem sets and projects still have real value, as artifacts of thinking and as occasions for it. The point is balance: when the stakes are high, written output needs a companion form of evidence that shows the person behind the work.
The urgency comes from where graduates are heading. They are entering workplaces where AI is part of how work gets done, and employers increasingly expect people who can use these tools well: knowing when to rely on them, how to question what they produce, and how to apply the results with judgment. That combination is AI fluency, and it has quietly become part of what workforce ready means.
Students do not pick up that fluency just by using the tools. Like writing or quantitative reasoning, it has to be taught deliberately, practiced repeatedly and assessed credibly. Employers are already telling institutions where the gaps are: in the National Association of Colleges and Employers’ Job Outlook 2025 survey, only 55.9 percent of employers rated recent graduates as proficient in critical thinking, and 53.5 percent in communication, two of the competencies employers consistently rank among the most important. Institutions that teach and assess these human capabilities alongside AI fluency will graduate students prepared for the workplace they will actually enter.
Authentic assessment is not a new idea. Oral examinations, clinical placements, teaching practice, language performance and presentations have always asked learners to show capability rather than describe it. What has changed is the need to apply that discipline more broadly, and the practicality of doing so at scale.
The resurgence is measurable. A three-year, NSF-funded project at the University of California San Diego ran more than 8,500 short oral exams across ten engineering courses; roughly 70 percent of students said the format increased their motivation to learn, and 72 percent said it increased their understanding of course material, as reported by The Chronicle of Higher Education.
The cycle that produces trustworthy evidence is consistent across disciplines: a clear task tied to a learning outcome, an observable demonstration against explicit criteria, specific feedback, structured reflection, and the opportunity to try again. A rubric shared before the demonstration tells learners what capability looks like and lets evidence be compared fairly across a cohort.
An illustrative example: in a nursing program, a student might submit a strong written care plan. The document is useful evidence of knowledge, but it cannot show readiness on its own. Ask the same student to talk through a deteriorating-patient scenario on video, respond to feedback on their clinical reasoning, and record a second attempt, and the program holds observable evidence of preparation, judgment and growth. (This scenario is illustrative, not a customer account.)
Used well, AI strengthens this kind of assessment rather than threatening it. It can give learners more frequent, lower-stakes practice opportunities than any instructor could provide alone, and it can help educators see patterns across many demonstrations. What it should not do is replace the educator’s judgment. Decisions about what evidence means, and what a learner should do next, remain human decisions, made within each institution’s own AI policies. Machines extend the opportunities to practice and observe; people decide what counts.
The argument for starting now is simple: the evidence question is already on the table, and the practices that answer it are proven. Institutions do not need to redesign every course to begin. One outcome, one task, one cycle:
Faculty will reasonably ask where the time comes from. Observing demonstrations takes more attention than grading documents, and workloads are already full. This is the practical hurdle, and it is where technology earns its place: asynchronous video makes demonstration possible without scheduling every learner in person, structured rubrics speed review, and AI-supported feedback can carry early, low-stakes practice rounds so educator attention concentrates where it matters most.
Those practical challenges, capturing demonstrations, giving timely feedback, tracking growth across attempts, are exactly what EchoReflect from Echo360 is built for: authentic, video-based practice and assessment that brings demonstration, structured feedback, rubrics, reflection and review together in one workflow, so institutions can make demonstrated capability routine rather than exceptional.
Echo360 is the global edtech leader enabling transformative learning experiences for education and business through the AI-powered Echosystem: the world’s first and only Learning Transformation Platform (LTP) that delivers dynamic creation, engagement, and assessment of education and training at scale worldwide. With innovative authoring (EchoInk), video (EchoVideo), engagement (EchoEngage), assessment (EchoExam), and video-based skills practice and feedback (EchoReflect) the Echosystem is an interoperable, modular platform. Learn more at echo360.com.
