The Downsides of Democratisation

“Democratisation” has been one of the defining buzzwords of 2024. In the context of Generative AI, it has been used almost exclusively as a positive: a levelling of the playing field, an opening up of powerful capabilities to anyone with an internet connection. Many new platforms and businesses have sprung up offering to “democratise” access to LLM capabilities, allowing organisations to leverage AI technology without vendor lock-in and without the need for deep technical expertise.

But it is worth pausing to examine what democratisation really means, what assumptions it carries, and whether its consequences are quite as universally positive as its proponents suggest.

The assumption of democracy

It is interesting that democratisation is so readily positioned as an unqualified good. According to recent analysis, only around 8% of the world’s population live in what could be classified as a “full democracy.” The assumption that democratic access is a natural, universal, and inherently positive model for organising anything, whether society or software, deserves more scrutiny than it typically receives. Democracy itself is hard won, fragile, and far from the default state of affairs.

This is not to argue against broader access to technology. But it is to suggest that uncritical enthusiasm for “democratisation” can obscure some important questions about what happens when powerful tools become widely available without corresponding structures for governance, education, and accountability.

The positives

The genuine benefits of democratised access to GenAI tools are significant and should not be dismissed.

Lowering barriers to entry. It is a fair assumption that the use of GenAI tools increases access to software creation. No longer is significant financial or human capital needed to create software businesses. An individual with a clear problem to solve and basic technical literacy can now prototype, build, and deploy applications that would previously have required a team of developers and considerable investment.

Creative expression. It is clear that Generative AI can enhance and provide easier ways for users to express creativity and control, especially in design and creative domains. Tools like Midjourney, DALL-E, and Adobe Firefly have genuinely opened up creative expression to a far wider audience, though it remains equally important to provide users with the tools to edit and manipulate outputs to suit their specific needs, rather than simply accepting whatever the model produces.

Addressing skills gaps. AI tooling can be used to fill critical skills gaps. As reported by Computerworld, four in ten C-level executives are looking to GenAI to address shortfalls in capability. But perhaps a better long-term approach would be to consider rapid upskilling and reskilling across the labour force. This, however, is fundamentally a social policy question at government level, and one that most organisations cannot solve alone.

The downsides

The downsides of democratised access to GenAI are less frequently discussed, but they are real and growing.

The amateur expert problem. At a basic level, in any organisation every employee can now become an “amateur expert,” able to cross their functional boundaries, for personal as much as organisational benefit. Individuals at any level can use ChatGPT or other chat services to learn about, and generate prescriptive answers to, business problems outside their area of expertise. This can create real friction within organisations that have historically relied on experience-based and skill-based functional separation across teams. When a junior marketing executive can generate a plausible-sounding legal opinion, or a sales manager can produce a credible-looking financial model, the question of who actually has the expertise to evaluate the quality of that output becomes pressing.

In most healthy organisations, there is some level of constructive disagreement, a culture of challenge and debate that drives innovation and problem-solving. Democratised access to AI-generated answers risks undermining this dynamic, replacing genuine expertise with fluently expressed but potentially shallow analysis.

Misuse and malicious applications. Democratised access to GenAI technology has the potential for misuse as well as positive applications. The ability to create realistic, fake content for phishing, social engineering, or outright criminal activity is becoming very real. As the cost of using these tools falls dramatically, and as access to unaligned LLMs and tooling capable of operating outside their original alignment and guardrails becomes easier, the likelihood of social media being weaponised for anti-democratic and criminal activities is very likely to increase.

Ethical concerns. There are also significant ethical considerations in the broader adoption of AI, as covered by CIO. Questions of bias in training data, intellectual property rights, environmental impact of model training and inference, and the responsibilities of organisations deploying AI tools at scale remain largely unresolved.

The labour market question. The longer-term impact on the job market, the need for reskilling at scale, and the shifting of economic power towards those who control AI infrastructure remains an important and unanswered question. Democratisation of access does not necessarily equate to democratisation of economic benefit.

The future: citizen programmers and the automation of creation

The longer-term trend is likely to move away from needing teams of skilled developers, with the concept of a “citizen programmer” emerging as a viable outcome. The most important factor will increasingly become solving a genuine business problem with “founder/market” fit, understanding the problem space deeply and personally, rather than “product/market” fit and having a team with the technical skills and experience to build a software service from scratch. The growth in AI-accelerated low-code and no-code platforms is one way that this trend is already being realised.

This use of automation to accelerate a technology transformation is nothing new. There are many examples throughout history. Vince Clarke, the electronic music pioneer behind Depeche Mode, Yazoo, and Erasure, has commented on how it is the sequencer, not the synthesiser, that has been most responsible for the music revolution from the 1970s to the 2000s: automation makes new forms of generation possible, as well as faster to create. Although anyone who has had to endure the infamous Page R on a Fairlight CMI might disagree.

The question for 2025 and beyond is not whether democratisation of AI will continue. It will. The question is whether the structures, policies, and cultural norms needed to manage its consequences can keep pace. History suggests they will lag behind, and the gap between access and accountability is where the real risks lie.