Today’s spark

This week was a heavy one. I had undergrad lessons to prepare, a keynote to finish, and a new course discussion on top of that. I got through it by working very closely with my “super‑analyst” co‑worker—Perplexity. It felt a lot like my Accenture days where I had the help of excellent junior team members. But a question kept nagging at me: my analysts used to learn from these interactions. Now I talk to Perplexity, and so do my peers who remained in the firm. So how do entry‑level people learn?

My conversation with AI

I asked AI what replaces the old apprenticeship model when seniors work mainly with AI instead of junior staff. The uncomfortable answer is that nothing replaces it by default. If organisations simply automate junior work and keep seniors plus AI, they risk breaking the talent pipeline.

New knowledge and perspectives

Apprenticeship hasn’t vanished; it just went invisible.

When seniors push their routine work to AI, juniors will gradually lose the messy, formative tasks where they used to build judgment. Unless we design new apprentice tasks on top of AI outputs—“What’s wrong with this answer?”, “How would you change it, and why?”—we’re not just saving time; we’re also deleting a generation of learning.

Entry‑level people need a structured way to argue with AI.

Teaching them to prompt is not enough. They need repeated practice critiquing AI’s answers against frameworks, client context, and logic, with a human pointing out where their critique holds up or falls short. In my client‑project course, this gives me a concrete idea: let students first learn the frameworks and think independently, then insert AI into the middle and have them compare, challenge, and refine its outputs in a visible, structured way—instead of everyone quietly using gen AI on the side.

Good design turns AI into a shared teacher, not a private shortcut.

If every student or junior uses AI alone on their laptop, I never see their thinking and I can’t coach it. But if I build AI explicitly into the workflow—generate, critique, discuss in teams, then review with me—AI becomes a common object we all look at together. That keeps the benefits of the tool and restores the apprenticeship dynamic: peer learning plus expert mentorship, just with a new actor in the room.

My next question

Today’s ah‑ha was that AI can sit between frameworks and independent thinking as a deliberate teaching tool, not a hidden crutch. My next question: how can I design one concrete assignment in my course that bakes this in, so students practise critiquing AI instead of quietly outsourcing their work to it?

Sources surfaced by AI for this post