Our Generative AI Future

“History never repeats itself, but it does often rhyme.”Probably Mark Twain

If my LinkedIn feed is anything to go by, we may already be at -or just past - peak ChatGPT. If so, I’m probably late to the party.

I’m not a technical expert on large language models. I can’t tell you whether GPT-4 is one release away from escaping onto the internet and going full Skynet, or whether the proposed AI “moratorium” from Elon Musk and friends is really just excellent tech PR—designed to inflate the value of companies that can plausibly claim AI capabilities while slowing down those that can’t. But I know which explanation I find more convincing.

What I can offer are some thoughts on how this moment compares to previous points in time when new technology rocketed up the hype cycle and began to challenge established ways of doing business.

Generative AI isn’t like last year’s obsession with the Metaverse. These tools are real, already useful, and operating at quality levels that justify at least some of the excitement. They deliver immediate value to both consumers and businesses, and they’ve found traction quickly because they don’t require dramatic behaviour change or expensive new hardware (seriously—does anyone actually know someone with a Meta Quest Pro?).

Where Generative AI does resemble the Metaverse is in its susceptibility to overhype. Both leap straight out of sci-fi and tap into a shared cultural disappointment: we were promised flying cars and robot helpers, and instead got internet dating and office productivity spyware.

To understand what might really happen next, it’s more useful to look further back - to technologies that have already played their hand and reshaped the world. Is ChatGPT the next Google Search? An iPhone moment? Or something closer to a fundamental shift in business software?

Looking at what it’s like - and unlike - may be more helpful than either breathless optimism or paranoid hand-wringing. And if I’m wrong and ChatGPT already controls the nuclear launch codes, then none of this matters anyway.

Is it like Google Search?

Google’s breakthrough wasn’t just PageRank, but the simplicity that made a rapidly expanding web navigable in the early 2000s. Its business model, however, was far less revolutionary—lifted straight from traditional media. Users got relevance; advertisers got attention.

This worked extraordinarily well. Google’s ad revenue hit $224bn in 2022. Broadly speaking, it’s been good for consumers (a single front door to the internet), good for businesses (one dominant media partner), and - monopoly concerns aside -remarkably stable.

Generative AI is already being framed as the next evolution of search. When Microsoft invested $10bn in OpenAI earlier this year, it was widely interpreted as a direct challenge to Google’s dominance, with Bing cast as an unlikely contender.

The promise is compelling: instead of searching, synthesising, and then manually turning information into something useful, you can delegate the entire task. The AI jumps straight to a first draft - text, images, actions - effectively collapsing search and execution into a single step.

In theory, this replaces not just search, but the intern, junior analyst, researcher, or PA. In practice - if recent headlines are to be believed - it may also replace them with someone who confidently makes things up and lands you in legal trouble.

If this really is an evolution of search, the implications for businesses are significant.

If Google was like paying someone to point customers to your shop, Generative AI is more like paying a virtual agent to sell on your behalf. Technically agile companies will adapt their platforms so AI can surface their products favourably. Deep-pocketed firms will pay for prominence.

But many businesses will find this threatening. If customers never visit your site or app, much of what we currently think of as branding and customer experience becomes less relevant.

This starts to feel uncomfortably familiar.

When Facebook Pages launched in 2007, there was a golden age of organic reach. Brands built audiences cheaply; publishers like BuzzFeed thrived almost entirely within the platform. Then Facebook captured both audience and advertisers—and slowly turned off the free tap.

We should expect a similar trajectory wherever Generative AI becomes a gatekeeper.

This could even disrupt former disrupters. I can never remember which streaming service hosts which show, and I have no emotional attachment to any interface. At some point, it’s easier to ask an AI to find what I want. When that happens, Netflix is just another content producer paying to reach me.

Is it an App Store moment?

The App Store was genuinely disruptive. It enabled deep integration with hardware and software, gave developers distribution, and triggered a Cambrian explosion of digital utilities. Entirely new categories - from Uber to Tinder to Strava - suddenly became mainstream.

ChatGPT shares some of this DNA.

OpenAI began with an open, non-profit ethos. While Microsoft’s investment has shifted the tone, the platform still feels unusually accessible. ChatGPT is free to use, and tools like GPT and Whisper can be integrated via APIs at relatively low cost.

This creates a future where companies that integrate with OpenAI’s technology inadvertently become features within OpenAI’s ecosystem - much like early apps enhanced the value of the iPhone.

But there’s a critical difference.

In the App Store world, developers fought for scarce shelf space—the limited grid of icons on a home screen. That scarcity made direct audience relationships valuable.

In a GPT-driven world, the shelf disappears entirely. The experience is closer to an infinite Argos catalogue with no index. You might ask for flights to Tel Aviv, select an option, and pay- without ever knowing which company powered the transaction.

This shifts power back toward genuinely differentiated products. A Emirates flight will still beat an EasyJet one, regardless of how seamless the booking is.

Is it a new era of business software?

Business software rarely captures the public imagination. Nobody drops out of university to build the next SAP module.

But it’s arguably the most transformative technology of the last fifty years. The move from paper to mainframes to cloud computing has reshaped how the world works far more deeply than search or smartphones. Globalisation, just-in-time supply chains, modern finance - all of it runs on enterprise software.

Until now, business systems have excelled at repetition and distribution. That’s created more work, not less: layers of people translating, reconciling, and interpreting outputs for decision-makers.

Generative AI threatens to break that pattern.

It can handle unstructured data, connect systems dynamically, and generate outputs readable by both machines and humans. It’s not obvious that it will create new roles at the same scale as the ones it absorbs.

In that sense, this feels less like Google - and more like Gutenberg.

But there are big caveats.

First: enterprises move slowly. Cloud computing has been viable for two decades, yet it took COVID to push many organisations off on-premise systems. Replacing software used by 10,000 people requires overwhelming proof or sudden urgency. Generative AI currently has neither. LinkedIn hot takes do not equal business cases.

Second: incumbents will resist. SaaS providers didn’t spend years embedding themselves in global organisations just to hand control to an AI layer. Interoperability wars will be fought quietly, contract by contract, for years.

Third - and more cynically - organisations tend toward hierarchy. Hierarchies require rungs. It’s hard to climb over an algorithm on the way to the C-suite, or blame one when something goes wrong. AI doesn’t just challenge workflows; it challenges management psychology.

A long, not-so-hot take

If you made it this far, apologies. I wish I had something more thrilling to offer. Many people seem oddly excited by the idea of AI going rogue.

Generative AI will increasingly act as a gatekeeper to audiences. That’s not new - it’s just the next version of an old pattern.

It will also dissolve layers of interaction between audiences and content. You won’t need a website or app - just a machine-readable database. Businesses that monetise the interaction layer will need to adapt fast or risk becoming middleware.

Enterprise adoption will be slow. There will be far more quiet bans than genuine productivity wins in the near term. But that won’t stop awards submissions.

As ever, we’ll overestimate what happens in two years and underestimate what happens in ten. And all of this depends on continued advances in chip manufacturing and stable global supply chains - both far from guaranteed.

So perhaps the bigger risk is not that AI will destroy us.

It may simply expose how fragile our systems already are.