Publishing Meta

AI in Education and Publishing: Breaking Old Habits or Reinforcing Them?

Artificial intelligence in education and publishing has long been touted as the transformative solution to entrenched inefficiencies and outdated workflows. Yet, the conversation often circles back to an oddly human problem: resistance to change. The claim is that AI isn’t the issue—it’s the “old habits” of educators, publishers, and administrators clinging to legacy processes. But is this critique accurate, or does it oversimplify the real challenges of adopting AI in these spaces?

Let’s unpack what’s really at stake here.

The AI Promise vs. Reality

For years, AI has been marketed as the magic bullet for education and publishing: adaptive learning platforms that personalise instruction, automated editorial workflows that speed up production, and intelligent systems that help educators focus on teaching rather than admin tasks. The technology has undeniably matured, with generative AI tools like ChatGPT, Grammarly, and adaptive learning software making concrete strides.

But behind the glossy demos and success stories lies a growing disconnect between the promise of AI and its actual implementation. Much of the resistance isn’t rooted in sheer nostalgia or obstinance—it’s often a rational response to the risks and limitations of relying on AI.

Data Privacy: The Elephant in the Room

One of the biggest “old habits” that institutions are accused of holding onto is their hesitancy around handing over sensitive data. AI systems, particularly in education, thrive on data: student performance metrics, behavioural patterns, engagement levels, and even biometrics in some cases. The problem is that this data is often stored, processed, and analysed in ways that are opaque to the institutions providing it—and sometimes downright exploitative.

Take the perennial issue of student data privacy. Many AI vendors operate on business models that depend on collecting vast amounts of information, often without clear guidelines on how that data will be stored, shared, or monetised. Schools and publishers, wary of inadvertently breaching privacy laws or putting their students and readers at risk, are right to question whether the operational efficiency AI promises outweighs the potential long-term costs of eroded trust and regulatory scrutiny.

Rather than being “stuck in the past,” these institutions are often trying to navigate a complex, underregulated landscape of data usage and cybersecurity risks. It’s not outdated thinking; it’s risk mitigation in a market where AI vendors frequently prioritise growth over transparency.

Power Dynamics and Vendor Lock-In

Another reason AI adoption has been slower than anticipated is the growing concentration of power among a handful of major vendors. Big tech companies like Microsoft, Amazon Web Services, and Google dominate the AI infrastructure landscape, while edtech providers like Pearson or Blackboard increasingly tie their platforms to proprietary AI tools.

This consolidation creates troubling power dynamics. Schools and publishing houses risk becoming dependent on a small group of vendors that control not just the tools but the ecosystems within which those tools operate. Vendor lock-in stifles innovation, limits freedom of choice, and places organisations at the mercy of external business decisions.

Consider this: What happens when a vendor’s AI system is suddenly discontinued, or its pricing model shifts? Institutions are left scrambling, forced to either rebuild their workflows or pay escalating fees to maintain access. The narrative that institutions are simply “stuck in their ways” ignores these structural risks—risks that savvy decision-makers are right to be cautious about.

The Gap Between Marketing and Reality

AI adoption also falters because the technology often fails to live up to its marketing promises. For example, adaptive learning platforms claim to personalise the educational experience, but studies repeatedly show mixed results. Many tools rely on oversimplified algorithms that fail to account for the nuances of human learning. Similarly, automated editorial workflows in publishing can speed up rote tasks but struggle with the creative complexity that editors bring to the table.

This gap between expectation and functionality can breed scepticism—not because stakeholders are resistant to innovation, but because the tools themselves often don’t deliver as advertised. Users are understandably wary of disrupting their workflows for technologies that may not provide clear, measurable benefits.

What Needs to Change?

The real challenge isn’t just about breaking old habits; it’s about creating systems where adopting new ones feels safe, logical, and beneficial. Here are a few questions institutions should ask before diving headfirst into AI adoption:

How transparent is the vendor about data usage? If you can’t get a straight answer about where your data is going, walk away.
What’s the long-term cost of implementation? AI tools may reduce short-term labour, but what’s the price of vendor lock-in five years down the line?
What human expertise are you sacrificing? Automating tasks is tempting, but AI should augment human workflows, not replace them entirely.
Are you solving the right problem? AI adoption often starts with a solution looking for a problem. Institutions need to ensure they’re addressing actual pain points, not chasing tech trends.

The Path Forward

If AI is to truly transform education and publishing, it needs to be more than a shiny tool. Vendors must address privacy concerns, resist monopolistic behaviour, and build systems that genuinely empower institutions rather than eroding their autonomy. Meanwhile, schools and publishers need to demand more transparency, flexibility, and evidence of effectiveness before upending their workflows.

The old habits that critics love to deride are often the only thing standing between institutions and the risks of unchecked technological adoption. AI isn’t inherently the problem, but neither is human scepticism—especially when it’s rooted in legitimate concerns.

Change will come, but it needs to be thoughtful, measured, and, above all, equitable. Anything less risks replacing one set of problems with another.