Hey,
In late March, OpenAI announced it would shut down Sora—its generative AI video model that was supposed to revolutionize video production and disrupt Hollywood.
Instead, it became a very expensive slop machine.
Sora generated endless video content. It also generated… almost no revenue. The product made just over $2M in lifetime revenue while reportedly costing $15M a day to run—plus a mounting pile of copyright issues.
Not exactly “the future of film” they expected.
But that financial problem only scratches the surface. OpenAI has promised investors it will spend around $1.4T on AI infrastructure over the next decade—despite generating $13B in revenue last year.
Their strategy is simple (and nothing new): Burn cash, capture market share, and figure out profitability later. It’s the Amazon playbook. It’s also the WeWork playbook. Which one OpenAI becomes is still TBD.
AI feels like a bubble right now because in some ways, it is. We’re in a hype stage where:
- Costs are widely inflated
- Expectations are unrealistic
- Every product looks impressive
But reducing it to “just a bubble” misses the point.
We’re in the middle of a real transformational shift that looks a lot like the early dot com boom. We’ve seen this film before, folks.
Back in 1998, economist Paul Krugman famously claimed the internet would have no more impact than the fax machine—a prediction that aged like milk. At the time, the internet had only been public for around 3 years.
Generative AI has been mainstream for…*checks math* ah yes, about 3 years. Even with faster progress today, AI is still in its dial-up era.
AI is underperforming because it isn’t integrated yet. We know the tech is powerful, but it’s still very much optional—and that’s exactly what the data shows. A recent study from The National Bureau of Economic Research found that 90% of businesses report no meaningful impact from AI, with projected productivity gains of just 1.4%.
Those numbers sound bad until you look at how people are using AI.
Most AI tools today live in side tabs. It’s a tool you visit, not something your work depends on. So no, it’s not surprising it’s not driving results.
Over the next few years, however, the market will correct itself. The hype will cool down, most AI tools will die, and the winners will deeply integrate into core workflows and become hard to replace. This is when productivity gains will show up—not just because the tech improves (it will)—but because adoption finally catches up to capability.
Meanwhile, many marketers are focused on the wrong thing.
Right now, there's pressure to use AI to produce more, faster. Can I be honest? That’s irrelevant and ultimately distracts us from the real challenge of figuring out where AI actually fits into real work and how it drives outcomes.
Output isn’t the bottleneck anymore, integration is.
Here’s what we need to do instead:
- Embed AI into workflows. Instead of treating AI like a side tool for copy and ideation, we need to ask ourselves, “where in our process are we slow, repetitive, or blocked?” and then build it into how work actually gets done, like briefs, research, reporting, or testing.
- Tie AI into measurable outcomes. If it’s not improving speed, conversion, or efficiency, it’s just novelty. Move from “this tool is cool” to “this tool moves a metric.”
- Build systems, not one-offs. With AI, anyone can generate content. The advantage is creating repeatable processes—inputs, prompts, QA, distribution—that your team actually relies on.
Let’s entertain a thought for a second. Say AI disappears in a few years. What have you actually lost by learning how to use it? Some time and money? I spend both every month watching more hours of Real Housewives than I’d like to admit. At least with AI, you’re building a skill that’s relevant right now.
So yeah, we’re seeing AI tools rapidly launch and fold just as fast. And while the technology feels like a productivity tool that hasn’t delivered, it doesn’t mean it’s failing. It means it’s too early and not yet embedded in how work actually happens.
And that gap—between capability and real-world use—is where the advantage will come from.
Until next time,
Ines
If you found this useful, please be a helpful work friend and send it to a colleague, or share it on Slack. Did a friend forward this newsletter to you? You can subscribe here.
.png)
