By Nayef Dagher, Founder & CEO, Amulet
This week, OpenAI confirmed it's shutting down Sora — the AI video generator that had the internet in a frenzy just 18 months ago. The compute is being redirected to robotics and autonomous agent infrastructure.
I'll be honest: I'm not surprised. And if you've been paying attention, you shouldn't be either.
The Demo Didn't Match the Need
Sora was genuinely impressive. Realistic, coherent video from a text prompt — the demos were stunning. But impressive demos and business value are very different things, and the gap between the two is where a lot of AI investment quietly dies.
The hard question was always: who is actually paying for this, and why? The answer turned out to be "not enough people." Content creators needed more control. Advertisers had compliance concerns. Enterprise buyers couldn't figure out where it fit in their workflows. The use cases existed, but they were niche relative to the infrastructure cost of running generative video at scale.
Meanwhile, something much less flashy was proving its value every single day: AI that handles work. Email triage. Document generation. Meeting summaries. Calendar management. Compliance reports. These aren't demo-reel material. But they save real hours for real people, and businesses are willing to pay for them.
The Agentic Cycle Is Here
What OpenAI's pivot signals is a maturation of priorities in the AI industry. The race to build the most visually impressive demo is over. The race to build AI that actually does stuff is well and truly on.
The redirected compute going into robotics and autonomous agents tells you exactly where OpenAI sees the next five years. Physical AI in the real world. Software agents handling knowledge work. Not video. Not image generation. Execution.
This is a meaningful shift. It's the difference between AI as a creative toy and AI as operational infrastructure. One generates content you look at. The other generates outcomes you need.
What This Means for Australian Businesses
If your AI strategy has been about exploring what's possible — watching demos, signing up for trials, experimenting with prompts — now is the time to get serious about what's actually valuable.
The businesses that will pull ahead in the next 12 months aren't the ones that found the coolest generative tool. They're the ones that deployed AI against their most painful, time-intensive workflows and didn't look back.
That means asking a different question. Not "what can AI do?" but "what does my team spend time on that AI could handle?" Email. Documentation. Scheduling. Reporting. Due diligence. Follow-ups. Client communications. The stuff that fills your day but doesn't actually require your brain.
That's where the ROI is. That's where the leverage is. Not generating videos.
Stop Chasing Shiny Demos
There will always be another impressive demo. Another model with a viral launch. Another capability that makes the rounds on LinkedIn. That cycle is relentless, and chasing it is a distraction.
The signal in OpenAI killing Sora isn't that video AI failed. It's that even OpenAI — the company best positioned to make it work — couldn't justify the cost relative to what agentic AI can deliver. That's a strong signal about where the value actually is.
At Amulet, we've been building in the agentic direction since day one. Not because we predicted OpenAI's roadmap, but because we asked Australian knowledge workers what was eating their time — and it was never "generating videos." It was email. Documents. Admin. Coordination. The invisible work that happens before and after every real piece of work.
That's what we automate. And that's where we think the AI story gets genuinely interesting for businesses.
The Bottom Line
OpenAI killing Sora is a clear sign: the industry is growing up. The shiny demo era is giving way to the infrastructure era. AI that does real work is winning over AI that makes impressive videos.
If you're still in the "exploring AI" phase, now is the time to pick a lane. The businesses getting actual value from AI right now are the ones that deployed it against real workflows — not the ones that watched the most demos.
Ready to put AI to work in your business? Join the Amulet waitlist →