AI in VET: Rethinking Assessment and RPL

In this presentation at the Velg conference, XponixAI founder Sharaf Goussous looks at the ethical, strategic and practical side of AI in vocational education and training. The focus is two of the sector's most tedious processes, assessment and Recognition of Prior Learning, and how scenario based role-play environments like KompetencyAI can counter cheating while handing assessors most of their time back.

Five ideas anchor the talk.

Understand where AI is heading

We are moving from narrow AI toward artificial general intelligence, systems as capable as humans collectively, and eventually toward artificial super intelligence, the point often called the singularity. Goussous's point is not science fiction, it is urgency: robust government regulation and institutional policies need to be in place well before the technology outruns them.

Rethink assessment to prevent cheating

Detecting AI generated text in written assignments no longer works, and it will only get harder. The answer is to change the assessment itself: flipped assessment models and simulated, scenario based role-plays where a student proves competency through live conversation, by responding to the situation and asking the right questions. That is exactly the environment KompetencyAI creates, and it is far harder to fake than an essay.

Streamline RPL

Recognition of Prior Learning is historically slow: unpolished evidence, thick files, hours of cross referencing. AI can automate the initial onboarding, map a participant's documents to specific TGA units, and identify the gaps before an assessor even opens the file. Done well, that saves 50 to 60 percent of an assessor's time on every candidate.

The human in the loop is non-negotiable

AI can give students instant feedback and cut an assessor's time per student from 25 minutes to around 5. What it must not do is own the outcome. Verifying feedback and approving results stays with a human assessor, always. The AI drafts, the human decides.

Future-proof VET with multi-agent systems

The next step is multi-modal AI and multi-agent systems: teams of autonomous agents, a bias auditor, an authenticity agent, a context expert, working together at the same time to validate learning and adapt to each student's needs. Instead of one model doing everything, each agent checks the others, which is precisely the kind of oversight regulated training demands.

The thread running through it all: AI can strip the tedium out of VET, but trust is earned through design, with scenario based evidence, transparent analytics and a human signing off on every outcome.

Watch the full presentation above, or get in touch to talk about what AI could do for your RTO's assessment and RPL workload.