Change management and learning design
Change Management: AI Training at Foundation Education
An internal course I proposed to leadership and built from scratch to upskill the whole workforce in what AI is, where it fails, and how to use it responsibly in an RTO.
- Proposed to leadership and built from scratch
- For the entire workforce
- Aligned to ASQA's AI principles
- Capstone game: The Black Box
- Role
- Product Design and Innovation Lead
- With
- Foundation Education
- When
- 2026
- Practice
- Learning design, Interaction design
Project
An AI course for a whole workforce, built for a compliance-driven organisation
At Foundation Education I proposed to leadership, and then developed from scratch, an internal AI training course. It is designed to upskill the entire workforce in what AI is and how it works, its strengths and weaknesses, and responsible AI use in registered training organisations (RTOs) and other compliance-driven environments.
The problem
AI tools arrived in the workplace faster than a shared understanding of them. In an RTO that gap carries real risk. Staff handle student information, make decisions that affect students, and produce training and assessment materials that have to stand up to audit. A confident invented reference, a biased course description or an unchecked automated message is not a small error in that setting.
The opposite risk is just as real. An organisation that avoids AI out of caution misses the time and quality it can give back when it is used well. People across the business were at very different starting points, from daily users to the openly sceptical, and few had been shown how the technology works underneath.
The solution
I treated it as a change management project with a course at its centre. The course gives everyone, in every team, the same foundation:
- What AI is and how it works. How a generative model gathers data, learns patterns and builds a response one predicted word at a time, taught through interaction instead of jargon.
- Where it fails, and why. Hallucinations and bias are explained as the same mechanism: the model fills a gap with the most expected pattern. Learners also meet limited context windows and data that leans American and historical.
- How to use it well. Specific prompts, the right model for the job, documented and measured processes, careful handling of personal information, and efficiency judged across the whole cycle instead of the first draft.
- Who stays responsible. People do. The course surfaces ASQA's principles for AI use throughout: decisions that affect students stay with a qualified human, and anything sent to a student is checked by one first.
Every example is set in the learner's own world. One interaction walks through eight areas of an RTO, from marketing and enrolment to validation and compliance, and shows what AI used with integrity looks like in each, what misuse looks like, and what each really costs.
My role
- Proposal: Identifying the need and making the case to leadership for organisation-wide AI training.
- Learning design and writing: Designing the curriculum and writing all of the content.
- Interaction design: Designing and building the custom interactions, using my Rise Code Block skill and iterating each one through review.
- Visual design and build: Building the course in Articulate Rise 360 in the Foundation Education brand, themed with my Rise custom styling extension.
Process
1. Start with how it works
Most guidance on AI at work is a list of rules. Rules are easier to follow, and to apply to a situation nobody predicted, when you understand the reason behind them. So the course begins with the mechanism. In "How generative AI learns", learners watch the scale of the data being gathered, see that the model keeps patterns and not pages, and then generate a sentence themselves, with the likely next words and their probabilities shown at each step. Run it twice and the result changes, which makes the point about variability better than a paragraph could.
2. Teach the failures as consequences
Once learners understand that the model predicts the most likely continuation, the failures stop being mysterious. "Recognising bias in AI" builds directly on that: a gap in the facts is filled with a hallucination, and a gap about a person is filled with a stereotype. The harder idea is that bias is not always simply wrong. The pattern can be real and still be bias when a majority is treated as the whole. The block ends with an AI-drafted course blurb in which learners find five biased phrases among fair ones, including an American "fall intake" that gives away where the copy came from.
3. Make it about their job
"AI across the RTO lifecycle" is a wheel of eight sections of the business. Each opens to show AI done well and AI misused, with the time saved now set against the cost later, and arrows show where good work or risk flows next. A final comparison adds it all up: misuse looks fastest on day one and becomes the most expensive option once rework, complaints and audit findings arrive. Whatever team a learner works in, their part of the organisation is on the wheel.
4. Consolidate with a game
The course ends with a capstone game, The Black Box. The learner does not defeat the machine; they shape it. Across twelve realistic dilemmas, each good decision moves it toward a trusted ally and each shortcut warps it toward a rogue, with a character that changes form as the balance shifts. The dilemmas are deliberately hard: a real document with an invented clause, a de-identified case study that still identifies someone, a time saving that causes three weeks of rework. The answer options are written to a similar length so length is never a clue. The message underneath is the message of the whole course: AI is as good or as bad as how you use it.
5. Hold it to the same standard it teaches
The interactions follow the conventions I set for all Foundation Education course interactions: plain and consistent instructions, full keyboard access, reduced-motion support, layouts that work on a phone, and knowledge checks that guide the learner to the right answer instead of counting attempts. Green is used only to signal something correct or complete, always with a tick or wording, so colour never carries meaning alone.
Impact
Impact and outcomes
- A shared foundation. One course gives the whole workforce the same understanding of AI and the same language for talking about it.
- Risk addressed where it sits. The scenarios are the organisation's own: student data, competency decisions, validation, records and audit.
- Confidence in both directions. Cautious staff see where AI genuinely helps, and enthusiastic staff see exactly where it needs a human check.
- Aligned to the regulator. ASQA's principles for AI use are built into the scenarios, not left to a policy document.
- Proof of the tools. The course is built with the same extension and Code Block skill that staff can now use themselves.
Coming soon. The full course will be added to the course samples once final changes are complete. Four of its interactions can be played now in the Playground.
Try it
Four interactions from the course are playable in the Playground, including the capstone game, The Black Box.