Canva cofounder Cameron Adams does not want employees to “figure it out” with one AI tool. In an interview with Rapid Response released on Tuesday, Adams said mandating a specific AI system for staff is exactly how you kill the experimental mindset companies need.

Adams was explicit about what he is not doing. “So, we're not mandating Claude or ChatGPT or Gemini or whatever tool you want to use in your part of the company,” he said. “You can figure it out.” The underlying logic is simple, but it has big consequences for how companies manage AI adoption, spend, and productivity: Adams said employees will use a forced tool “begrudgingly,” and that reluctance “hinders experimentation.”

This is one of those quietly consequential debates that is now sitting inside boardrooms and operational playbooks. As more companies chase AI at speed, they face a familiar management problem: standardize too much and you lose experimentation, standardize too little and you lose control. Adams’s stance is essentially “control the outcomes, not the exact vendor.” Canva staff get “their own AI budgets to try out tools, figure out workflows and processes, and tackle the problems they face.” That budget model is meant to turn AI from a corporate mandate into a set of experiments that map to real work.

And if you are wondering whether this is just theory, Canva’s process includes time allocation too. Another way to encourage experimentation, Adams said, was to give staff time off from their normal responsibilities to play with AI during Canva's “AI Discovery Week.” The company told people, “Please don't do your normal work. We want you to think of the problems that you have, the tools that you've heard about, the opportunities that you've heard from colleagues in other industries, and we want you to try those out for the entire week,” Adams said. In other words, Canva is not just letting employees choose tools, it is protecting the time needed to actually test them.

That matters because AI adoption is already proving harder than “buy the software, deploy the chatbot.” Business Insider notes that getting employees to adopt AI into workflows has been a challenge for many companies, and it points to several different approaches across industries. Some companies, like Duolingo, added employee AI use as a performance metric. Others, including JPMorgan and Disney, started AI token leaderboards ranking staff’s usage. Those strategies reflect a different belief than Adams’s: that behavior can be nudged through measurement and gamification.

But the real constraint is not just adoption, it is spend. The source flags AI spend as a “foremost challenge for enterprises,” and it notes that companies are trying to “spend less without throttling their employees’ AI ambitions.” One cost-saving tactic that has emerged is using different AI models and tools for different tasks, rather than being loyal to a single AI provider. That aligns with Adams’s argument against mandating one tool, even if it comes from a different angle. For many teams, the cheapest or best tool for a drafting task might not be the best tool for summarization, editing, or other internal workflows.

The competitive and cost context also extends beyond workplace rules. Canva has made AI a core focus of its business, including launching Canva AI 2.0 in April, a conversational platform that lets users turn simple prompts into designs. Business Insider says it “previously pitted Canva 2.0 against Claude Design to build the same slide deck” and found the final product was comparable to Claude’s. That kind of benchmarking helps explain why Canva cares about experimentation internally: if the company’s product strategy depends on AI performance, employees need room to explore tools and workflows rather than locking into one approach.

There is also an industry-wide “stack” reality that pushes companies toward flexibility. The source cites Coinbase CEO Brian Armstrong, who in June said that using multiple AI models and tools for different tasks helps prevent excessive AI tokens from being burned on simple tasks. Armstrong also recommended using cheaper Chinese models as defaults. Meanwhile, Vercel CEO Guillermo Rauch told TechCrunch earlier this month that companies are getting smarter about how to use different AI tools across their AI stack, including the model, harness, data platform, sandbox, and gateway.

For executives reading this, the takeaway is not just “let employees choose.” It is that tool mandates are not a neutral policy. Adams argues they change employee psychology, and psychology changes experimentation velocity. When AI spend is under scrutiny, and when leadership teams are searching for ways to reduce costs without slowing ambition, “single-tool enforcement” can become a hidden tax on learning. Canva’s model, as described here, is to fund experimentation directly through “AI budgets,” encourage tool discovery with “AI Discovery Week,” and keep the company’s stance practical: do not force Claude or ChatGPT or Gemini on every team. Make experimentation the default, and let teams align tools to their actual problems.