Could an AI agent work for you - or does it really work for the tech bros?

AnalysisAgentic AI

Agentic AI promises to help with annoying tasks. But behind the fluffy characters is hardcore capitalism

SOME OF US are old enough to remember the original promises of the internet. It was going to democratise information, create a global community and put an end to middle men.

Social media was going to connect humanity. The sharing economy was going to liberate us from the tyranny of corporations. There would be multiple Arab springs. Information wanted to be free.

But it turns out nothing was free.

Instead, Google became the gatekeeper for information. Facebook became the owner of our online relationships. Amazon became the default online shop. Apple built an ecosystem that locked users into an ecosystem of high-end hardware. It was winner-takes-all economics. And the result was a host of walled gardens that privatised all the gains of the internet and socialised all the losses.

Platforms pocketed advertising revenue while legacy media kissed its business models goodbye. Social networks monetised engagement while society absorbed the costs of misinformation, polarisation and conspiracy theories. The gig economy generated fortunes for investors while transferring risk to workers. The business model was simple: keep the upsides, outsource the consequences.

Now the same Silicon Valley types are back selling the dream of artificial intelligence.

Sure, the productivity gains are clear for professionals; coders, lawyers, marketing professionals and anyone doing admin tasks can all see the benefits.

But that’s not enough; we’re also being told AI is for everyone. It’s not just B2B, it’s B2C too. We can all look forward to agentic AI: software that does things for us. Soon, we’ll be able to tell our AI assistant to find the cheapest electricity provider, negotiate a better broadband contract, book a family holiday or move your savings to whichever bank offers the best return. Instead of doing all the research and legwork, our assistants will do it all for us.

Could it possibly be that this technology might actually shift power towards the consumer? Maybe, this time it’s different?

Maybe not.

Why? Because the aforementioned walled gardens aren’t so keen on giving up the market share they’ve accrued. They’re blocking AI from doing precisely the things their developers promised to protect their positions.

Yelp restricts access to agents participating in its paid data licensing programme. Delta Air Lines has prevented agents from booking flights. JP Morgan Chase’s CEO Jamie Dimon recently said the bank will block external AI agents from logging in to customer accounts. Citi and Bank of America have followed suit. Amazon has been fighting Perplexity over its shopping agent. It has also blocked Meta’s Muse assistant. When he heard that Amazon may block X’s Grok agent, Elon Musk,was so incensed he suggested he might start his own Amazon competitor.

Blocking or partnering 

And when they’re not blocking each other, the corporations are forming strategic partnerships. Meta has done deals with retailers like Walmart, Gap, Sephora and Best Buy, allowing its assistant to shop with them. Other agentic alliances are sure to be formed. So instead of an agent that can scour the entire internet looking for the best deal, we risk getting one that operates within a network of corporate partnerships.

It’s like the streaming wars all over again. Netflix stole a march, amassed a strong library of content and promised that we’d never have to worry about television schedules. Then every studio with enough intellectual property decided it wanted its own streaming service. The shows and movies all disappeared behind competing paywalls, and consumers found themselves paying five separate subscriptions to watch their five favourite shows.

The same economic fracturing is likely to happen in the world of agentic AI.

Yes, we’ll have agents but with limited or no interoperability. Our AI agents might book a hotel, but only with participating chains. They may buy groceries, but only from approved retailers. They could compare financial products, but only from institutions that have negotiated access agreements. Our AI assistants could well become personal shoppers who only recommend shops that belong to a conglomerate.

This isn’t corporate skullduggery. In some instances the corporations are just looking after their shareholders. And there’s nothing wrong with that.

In other instances, they can legitimately say they are looking after their customers. Banks, for example, may be entirely correct to restrict access that would prevent an agent wiping out a user’s life savings, regardless of the instructions it was given. 

But if we can instruct an accountant to examine our finances, why shouldn’t we be able to instruct software to do the same, regardless of which platform it comes from? If we can hire a broker to find a cheaper mortgage, why should banks be able to prevent AI assistants from doing the same thing on our behalf? For users to realise the benefits of agentic AI, there should be common standards, transparency, clear permissions and accountability – not a collection of private agreements negotiated between billionaires. 

EU needs to take the lead

Maybe this is one area where the EU can take the lead. Europe has already spent years trying to prise open digital markets through competition law, interoperability requirements and consumer rights. If the same principles are applied to AI agents, European consumers could see real benefits. If not, we’ll end up beholden to Chinese and US AI companies where the appearance of choice conceals an elaborate system of corporate control.

Thirty years ago, the internet promised to give everyone access to the world’s information. We ended up with a handful of companies deciding which information we could see.

Now AI promises to give everyone the power to act. We risk ending up with a similar handful of companies deciding which actions we’re allowed to take. Sure, AI promises a host of efficiencies. But don’t be fooled: this isn’t technological idealism, it’s good old capitalism.

Steve Dempsey is a media and technology expert and commentator.