AI in the Hot Seat: Canva’s Bold New Interview

Canva just changed what a coding interview looks like, and others will copy it fast. AI use in their coding round is not banned, and not merely allowed. It is required. Candidates complete a live product challenge using their AI model of choice, and they narrate every step: prompt, refine, ship, explain. The textbook algorithm drills are gone, replaced by something that looks like the actual job.

It is a smart move. Most engineers hated leetcode theatre, and a large majority of developers already use AI daily. Real teams do not ban it, so why should the interview. Canva is the first major Australian unicorn to formalise AI collaboration as a core skill, and dev, product, and data roles elsewhere will follow.

So the question is no longer whether you use AI in an interview. It is whether you can be seen to use it well.

What they are actually scoring

The winners in this format are not the fastest typists. They are the best at directing the tool. A sharp interviewer is now watching for three things:

• Prompt clarity: concise, testable asks, not vague wishes.
• Critical override: you catch the hallucination before it ships.
• Narration: you explain your decisions in plain English as you go.

That last one is the whole game. Working with AI well is a thinking-out-loud skill, and it is trainable.

How to practise this week

Treat every practice session like a paired-programming stream. Speak your prompt aloud. Say why you are accepting or overriding the AI's suggestion. Keep the before-and-after of what it gave you versus what you shipped. Run that loop a few times a week and you build the exact habits the new format rewards: transparency, judgment, iteration.

A few concrete drills:

• Build a prompt vault. Save your best prompts for scaffolds, test stubs, and docs, so you are not improvising under pressure.
• Rehearse out loud. Talk through your prompts and edits while you code, on a Loom if it helps.
• Add guard-rails. After the AI suggests code, prompt it for edge cases and complexity, and check its work.
• Debrief. After a real interview, write three lines on where AI helped and where you over-ruled it.

Role-specific reps

• Product: feed an AI a mock PRD, have it surface five launch risks, then narrate your mitigation.
• Data engineering: ask it to optimise a Spark job for cost, then explain why the changes cut shuffle.
• Data science: hand it a clean dataset and prompt for three baseline models plus metrics, then justify your pick.
• Front-end: have Copilot build a virtualised React table, then critique its accessibility gaps.

It is already showing up in results

Over the past six weeks, three of five Careersy clients who practised this AI-collaboration skill made it to second-round interviews. Not because they blindly trusted the tool, but because they showed judgment, current awareness, and real examples that landed.

Canva's mandate is hiring 2.0: prove that you and AI are one team. Practise narrated prompting, audit the model's work, and document your decisions like a senior engineer. This is the start of a shift that will spread across rounds, roles, and industries. The candidates who start now will look fluent while everyone else is still adjusting.