Why Canva Won’t Force One AI Tool on Employees (And What That Means for Your Stack)

Why most adoption tactics miss the point

Plenty of companies are pushing AI hard, but often with sticks instead of carrots. Duolingo started tying AI use to performance reviews. JPMorgan and Disney built leaderboards that rank staff by AI tool usage. These approaches signal that AI is a chore—something you have to do, not something you want to figure out.

Canva’s method flips that. Freedom becomes a feature, not a checkbox. When employees own their tool choices, they treat AI less like a corporate mandate and more like a new superpower. They’re more likely to discover niche workflows, unexpected efficiencies, and edge cases a central team would never spot.

What this means for your AI stack

Enterprises are now waking up to the cost of monolithic AI loyalty. Using a single provider for everything burns tokens fast—and often wastes them on tasks a simpler, cheaper model could handle.

Coinbase’s CEO has openly talked about matching models to task complexity to avoid overspend. Vercel’s CEO sees companies getting smarter about layering different tools across the stack—model, harness, data platform, sandbox, gateway.

Canva’s approach feeds right into that multi-model reality. By letting teams naturally gravitate toward the right tool for the job, the company essentially crowdsources its own AI stack intelligence. A marketer might lean on Claude for copy, while a data analyst swears by Gemini for reasoning. Over time, that organic selection surfaces the real tool-task fit—no procurement committee required.

How to steal the playbook (without chaos)

Handing out unlimited AI budgets to a team of 500 isn’t always feasible. But the principles scale down elegantly.

• Give small, ring-fenced budgets. Even $50/month per person signals trust and funds real experimentation.
• Carve out exploration time. A half-day or a week where normal work is paused lets people actually touch the tools they’ve only read about.
• Share the wins. Create lightweight channels where teams post what they tried, what broke, and what surprised them. Knowledge spreads faster when it’s peer-to-peer.
• Resist the urge to standardize early. Let enough people hit walls with a tool before you even think about an official recommendation.

The real takeaway

Your AI stack isn’t a monolith—and it shouldn’t be a religion. Locking your team into one tool today is like buying them all the same screwdriver and wondering why nobody can hammer a nail. Give people the freedom to find their own fit, and you’ll end up with a smarter, cheaper, and far more enthusiastic organization.