Banning AI From Your Team Does Not Protect Performance. A Two-Year Study Just Proved It.

Banning AI From Your Team Does Not Protect Performance. A Two-Year Study Just Proved It.

What if the most damaging thing you could do with AI is simply refuse to use it? That is exactly what a two-year university study out of Vrije Universiteit Amsterdam found, and the implications reach well beyond the classroom, straight into how small business owners should think about AI adoption right now.

Thibault Schrepel, a law professor at Vrije Universiteit Amsterdam, ran a controlled experiment across two consecutive years with his "Law of AI" course. He divided students into three groups: one banned from using ChatGPT entirely, one given ChatGPT-generated suggestions embedded in their work with zero training, and one given structured, hands-on training in legal prompt engineering and how to verify AI outputs for accuracy. All three groups worked in small teams of four or five, completed the same 20-minute task of improving a provision of the EU AI Act, and took identical exams. The study ran with 66 students in 2024 and was repeated at scale with 164 participants in 2025.

The results were striking and consistent across both years. The no-AI group finished last every time. Students in that group quickly hit what Schrepel calls "idea exhaustion," running out of substantive direction after 10 to 15 minutes and defaulting to minor wording tweaks. The untrained AI group made uncritical errors in the classroom exercise, accepting ChatGPT suggestions that included legally misleading terms without questioning them, but those errors did not carry over to the exam. Schrepel attributes this to something important: people who use AI tools on real, consequential work develop an instinct for where AI goes wrong, even without formal training. The trained group outperformed both in 2024, but by 2025, the advantage had nearly disappeared, because growing familiarity with AI tools in everyday life had largely closed the gap.

Schrepel's own starting assumption was that unguided AI use would cause measurable harm. "I was wrong," he writes. Banning AI produced worse outcomes than allowing it, even without any structured instruction at all. The study's conclusions carry a weight that goes beyond academia: keeping people away from AI tools does not protect them. It leaves them behind.

For small business owners, this research reframes a common instinct. Many operators hesitate to introduce AI into their teams because they worry about quality control, errors, or overdependence. Those concerns are real, but the evidence suggests the cost of avoidance is higher. Schrepel's study shows that even people using AI without guidance still outperform those who are blocked from it entirely, and they learn to self-correct through direct experience with real-stakes work. In business terms, that means your team members who are already using ChatGPT, Claude, or other tools on the job, even informally, are likely building a working instinct for what AI can and cannot do. Blocking that experimentation does not create a safer environment. It creates a knowledge gap that compounds over time.

There is also a direct marketing and customer experience angle here. If your team is not developing AI fluency, your content output, response times, personalization capabilities, and campaign efficiency will fall short of competitors who are actively building those skills. The study found that formal training still adds real value, particularly around ethical use and accuracy checking. Businesses that invest in teaching their teams how to use AI tools well, not just allowing access to them, position themselves to produce better work and catch the errors that untrained users will miss.

The second group in Schrepel's study accepted AI suggestions that "sounded better" without understanding the implications. That is exactly the failure mode that hurts businesses in marketing: publishing AI-generated content that is fluent but factually off, legally imprecise, or brand-inconsistent. The lesson is not to ban the tools. It is to train the people using them.

This week, pick one workflow your team handles repeatedly, whether that is writing email responses, drafting social content, or summarizing client notes, and run a structured session where you walk through how to use an AI tool for that specific task, review the output together, and document where the tool helps and where it needs human judgment. That single session begins building the institutional AI fluency that separates businesses that get results from AI from those still debating whether to allow it.

AI is not a shortcut that replaces good judgment. It is a capability multiplier for teams that know how to use it well, and the businesses that build that skill now will set the standard that everyone else scrambles to match.

Originally inspired by: Two-year university study finds banning AI from classrooms leaves students worse off (https://the-decoder.com/two-year-university-study-finds-banning-ai-from-classrooms-leaves-students-worse-off/) See how Leads to Conversion can help your team grow faster with AI strategy. Get your free AI audit

Your turn

What is your traffic actually doing?

Send us your details and we will come back with a short, specific read on what your traffic, your pages and your pipeline are doing today — and the first three things we would change. A real person reads every submission, and you get the read whether or not we ever work together.