Bottom line: yes, AI tools should be allowed in schools and workplaces, because banning them doesn’t stop the use, it just stops you seeing it. The organisations getting real value are the ones writing clear rules for how AI is used, not pretending it isn’t happening. The ones banning it are training people to hide it, which is a far worse outcome than the one they’re trying to prevent.
More on this here: Should You Put AI Tools on Your Resume? My Honest Take After Reading H.
Wrong question, right worry
“Should AI be allowed” is the wrong frame. It’s already happening, in every school and every office I’ve walked into for the past two years. The real question is whether you shape how it’s used or find out about it too late, usually from a plagiarism report or a client complaint about a report that reads like it was written by nobody in particular.
I’ve sat in boardrooms where the CEO announces a company-wide ban on ChatGPT, and watched three people in the room quietly close a browser tab. That’s not a policy working. That’s a policy being ignored by the people it’s meant to govern.
What’s happening in schools right now
JISC’s National Student AI Survey, run across UK higher education, found the majority of students were already using generative AI tools for their coursework, most commonly for summarising reading, generating ideas, and checking their own writing. Not cheating in the cartoonish sense of “write my essay for me”, though that happens too. Mostly it’s students using AI the way office workers use a spellchecker, except nobody wrote them a rulebook for it.
Universities responded in three ways. Some banned it outright and now spend enormous time and money chasing detection software that flags essays written by non-native English speakers as “AI generated” more often than it flags actual AI text, because the detection tools are unreliable and biased toward flagging simpler sentence structures. Some ignored it and hoped it would go away. It didn’t. The smartest ones rewrote their assessment criteria to require students to show their working: draft versions, source notes, a short reflection on what AI tools they used and why. That third group is producing students who leave university able to use AI as a tool rather than a crutch, which is exactly the skill employers now expect on day one.
Secondary schools are further behind, understandably, because a fourteen-year-old and a first-year undergraduate are not the same student with the same judgement. But even there, the schools doing this well are teaching AI literacy as a subject in its own right, not treating every use of it as an infraction to be caught and punished.
What’s happening in workplaces
Microsoft’s Work Trend Index found that the majority of knowledge workers were already using AI tools at work, and a large share of them were doing it on their own initiative, using personal accounts on personal devices, without telling IT or their manager. Microsoft coined a term for it: BYOAI, bring your own AI. It’s the same pattern as the early days of smartphones in the office, when staff started using their own phones for work email long before IT departments had a policy ready.
Here’s the uncomfortable bit nobody likes admitting in these conversations: a workplace ban on AI tools doesn’t reduce AI use. It reduces visibility. Staff still use it, they just use it on their phones, on personal Gmail accounts, pasting client data into a free tool with no idea where that data goes or who can see it. A ban aimed at protecting the company from risk very often increases the risk, because now the use is happening entirely outside anyone’s sight, with no training, no guardrails, and no record of what was shared.
A story I keep coming back to
I worked with a marketing agency owner, mid-thirties, twelve staff, good business, decent margins. He’d banned ChatGPT company-wide after reading a scare story about a data leak. Six months later he asked me to help him understand why his content output had barely changed despite the team supposedly not using AI at all. I asked to see a sample of recent client reports. They read like AI output, competent but generic, the kind of sentence structure you get from a tool with no house style fed into it.
I asked around, gently. Every single writer on his team was using AI. On phones. On personal laptops during lunch breaks. None of it logged, none of it reviewed, none of it fed with the agency’s actual brand voice or client history, because doing that openly would have meant admitting they’d broken the rule. He’d spent six months believing he had a policy. What he had was a team that had learned to lie to him about a completely normal part of their working day. We scrapped the ban, wrote three pages of actual guidance instead, and within a month his account managers were reviewing AI-assisted drafts openly, catching errors before they reached clients, something that had never happened when the whole thing was done in secret.
The case for allowing it, considered
- Staff and students who use AI openly can be trained, corrected, and supervised. Staff and students who use it in secret can’t be.
- Banning a tool that’s free, on every phone, and improving every three months is not a control, it’s a wish.
- The skill gap employers now hire for is judgement about when to trust AI output and when to override it. You only build that skill through supervised practice, not through pretending the tool doesn’t exist until graduation or your first day on the job.
- Allowing AI with disclosure rules gives you a paper trail. Banning it and hoping gives you nothing when something goes wrong.
The case against, taken seriously
I’m not going to pretend the risks are small, because they aren’t.
- Data leakage. Free AI tools can and do use input data to train future models unless you’re on a paid business tier with data controls switched on. A staff member pasting a client contract into a free chatbot is a real and common way sensitive information ends up somewhere it shouldn’t.
- Skill atrophy. A junior who outsources every first draft to AI without ever writing one unaided doesn’t develop the underlying judgement to know when the AI draft is wrong. This is a genuine long-term concern for entry-level roles in writing, coding, and analysis.
- Confident wrongness. AI tools produce fluent, well-formatted, entirely incorrect answers with total confidence. In a school essay that’s a bad grade. In a workplace report used to make a real financial decision, that’s a much bigger problem.
- Inequality. Not every student has a laptop or reliable internet at home, and not every worker has the same starting familiarity with these tools. A blanket allowance without support widens gaps rather than closing them.
None of these are arguments for a ban. They’re arguments for training, tiering, and disclosure, which is a completely different thing.
How to write a policy that works
I’ve helped several small business owners build these from scratch. It doesn’t need to be long. It needs to be specific.
- Name the approved tools. Not “AI” in general, actual named tools with business-tier accounts and data controls switched on, so nobody defaults to a free consumer version out of convenience.
- List what’s off limits. Client personal data, unpublished financials, anything under an NDA. Be specific, not vague. “Don’t paste client names, contract terms, or unreleased pricing into any external tool.”
- Require disclosure, not permission. Nobody wants to ask for approval every time they draft an email. Ask instead that AI-assisted work be marked as such internally, especially anything going to a client or a student’s final submission.
- Train, don’t just announce. A one-hour session showing real examples of good and bad AI use in your specific work does more than a page of rules nobody reads. I run these sessions for teams directly, and the questions people ask once they feel safe admitting they already use these tools are always more useful than the ones they ask in a policy rollout meeting.
- Review every quarter. These tools change fast. A policy written in January can be out of date by June. Put a date in the diary to revisit it, not because you expect huge changes but because you will get them anyway.
Businesses that skip straight to writing a policy without first understanding what their team is already doing tend to write rules that don’t match reality. That’s usually the point where it’s worth bringing in outside help to look at the actual tools in use and build something that fits, rather than something copied from a template. I cover this in my guide on working with an AI consultant for a small business, because most of the policy failures I see come from businesses trying to write the rules before they’ve done the audit.
What I’d tell a headteacher and a CEO, in the same breath
Allow it. Name the tools. Teach the judgement. Punish the dishonesty, not the tool. A student who discloses they used AI to structure an essay and then improved it themselves has learned something real. A student who hides it has learned to hide things, which is a much worse habit to walk into adult life with.
The businesses I respect most right now are the ones treating this the way Melanie Perkins treated design software at Canva, not as a threat to gatekeep but as a capability to hand to as many people as possible, with enough structure that the results are usable. I wrote more about how she built that mindset into a company in my piece on business lessons from Melanie Perkins, and the same instinct applies here: open access with real guardrails beats locked doors every time, because locked doors just get climbed over quietly.
Should AI tools be allowed in schools and workplaces? Yes. Loudly, with rules, with training, and with the honesty that a ban has never once stopped a determined fifteen-year-old or a busy account manager from opening a new tab.
Frequently asked questions
Should schools ban ChatGPT and similar AI tools?
No, evidence from UK higher education shows most students already use generative AI regardless of bans, so schools that teach disclosure and AI literacy get better outcomes than schools relying on detection software, which is unreliable and often flags non-native English writers unfairly.
Is it legal for employers to ban AI tools at work?
Yes, employers can set rules on which software staff use, including banning specific AI tools, but a ban with no enforcement or alternative usually just pushes use onto personal devices where the company has no visibility or data protection at all.
What’s the biggest risk of allowing AI tools in the workplace?
The biggest concrete risk is data leakage, where staff paste client information, contracts, or financial details into free consumer AI tools that may retain that data for training, which is why any workable policy names approved business-tier tools with data controls switched on rather than leaving staff to choose their own.
How do you write an AI use policy for a small business?
Name the specific approved tools, list exactly what data can never be entered into them, require staff to disclose AI-assisted work rather than seek permission for every use, run a short training session with real examples, and review the whole policy every quarter since the tools change fast.
