Current AI – Design Reveals Intent, So Pay Attention

AI Isn’t Magic – It’s Built This Way

The rise of AI has been heralded as a technological revolution—systems that can write, analyse, predict, generate, and even converse with astonishing fluency. Billions have been poured into research, infrastructure, and productisation. And some of the results are, by any measure, impressive.

But alongside the hype, something peculiar stands out. Why do these tools—so powerful, so well-funded—come with such glaring gaps? Why can’t they remember what you told them last session? Why do they ignore what you’ve said you don’t want? Why is personalisation so often shallow, and redirection so often ignored?

These aren’t mere technical oversights. Maybe it's time to consider the possibility that AI is being built—and deployed—intentionally. Consider that this is just not necessarily for the benefit of the users, because in the emerging AI economy, the user is no longer the customer. The user is the source: of data, of training material, of behavioural signals, and the ultimate source of payment. The real customers are the platforms’ paying clients—advertisers, suppliers, and commercial partners.

This piece explores three connected observations:

• Why the most intriguing quality of AI is its ability to simulate choice while quietly removing real control.
• Why the most useful insight for businesses is that much of what’s sold as AI either isn’t new—or is very expensive to implement meaningfully.
• Why the most important shift is the transformation of the digital customer relationship—from Business-to-Customer (B2C) to Business-to-Customer-to-Supplier (B2C2S)—where the user is no longer the point of service, but a point in a supply chain.

1. The Loss of Agency: You’re Not Choosing—You’re Being Routed

Henry Ford once said you could have any colour car you wanted, as long as it was black. The AI version of this? "You can go anywhere you want, as long as its our road"

Today’s AI tools give the appearance of flexibility and personalisation, but systematically avoid giving users real control. There is no reliable way to say “don’t show me this again,” to opt out of certain genres or tones, or to persistently steer the system away from behaviours you find manipulative or irrelevant. Is the thumbs down as much as AI can handle?

The illusion of choice conceals the fact that you’re on rails. The loss of agency is not a technical flaw. It’s a product of business logic: if AI systems allowed you to define your own boundaries, they would gather less behavioural data—and generate fewer monetisable nudges.

2. Bundling: Old Tools Lead To Expensive Realities

Much of what is now marketed under the AI banner includes tools that have existed for decades: recommendation engines, predictive analytics, robotic automation. The bundling of these with newer capabilities lends an air of seamless innovation that is often more rhetorical than real.

Where AI has advanced—particularly with large language models and computer vision—the reality is that these systems are complex, costly, and require significant customisation to deliver real business value. Clean data, domain expertise, infrastructure, and continuous refinement are not optional.

AI cannot be sold as an app. In many cases, implementing something genuinely useful means building the equivalent of an in-house supercomputer—or at least maintaining the teams, systems, and strategy that can make use of one.

3. The Strategic Shift: From B2C to B2C2S

Perhaps the most important trend is not technological, but commercial. The dominant tech model in consumer facing businesses is no longer Business-to-Customer (B2C). It is Business-to-Customer-to-Supplier (B2C2S). Platforms serve paying suppliers—advertisers, product vendors, data clients—while using consumers as a source of behavioural data.

This explains many of the design choices in today’s AI systems. Features that would serve the end user—clear memory, persistent preferences, the right to refuse or redirect—are systematically underdeveloped or omitted. Instead, systems are built to capture attention, extract behaviour, and optimise conversion for the paying customer upstream.  The result is a set of tools that appear to work for you, but are in fact working on you.

Conclusion: Design Reveals Intent

It’s time to ask not what AI can or cannot do, but why it does—or does not do—what it does.  The limitations of current AI tools are not technical accidents. They are design choices. Each missing or frustrating feature reflects a strategic priority that puts someone else’s interest ahead of yours.

If we want AI to support genuine user agency, meaningful decision-making, and real business transformation, we have to push for a different model. One where the user isn’t a data source, but a stakeholder.

Until then, remember: the choice was yours—after someone else made it for you.