From Hype to Workflows: Where AI’s Real Shift is Happening
We expected AI to change everything overnight. Instead, progress is hidden in pilots and processes. The sinkhole is forming beneath us.
I’ve been reading a lot lately on AI and work. One theme keeps coming up: people expected to see more visible change by now. After the hype of ChatGPT and Copilot, many organisations wonder, is this it? Is AI already stagnating? Or are we still on the edge of something bigger?
This article is a distillation of what I’ve been seeing in research, case studies, and my own work. I want to share why I think the disappointment is understandable, but also misleading.
The Horizontal Wave: Loud, Visible, and Sometimes Cluttered
The first wave of AI at work has been horizontal. Tools like ChatGPT, Copilot, or Claude spread quickly across teams. They’re easy to try, and they promise immediate help with tasks: drafting emails, summarising meetings, rephrasing text.
That visibility is exactly why expectations rose so high. Everyone could see colleagues using AI in small ways. Yet, as McKinsey notes, almost 80 percent of companies report using generative AI, but most still see no real bottom-line impact.
There’s another catch: horizontal AI doesn’t just make work easier, it can make more work. Because it’s so simple to generate content, companies risk drowning in drafts, slide decks, and documents.
Ethan Mollick joked about generating 17 versions of the same PowerPoint. It’s productivity on paper, but often it clogs the information stream instead of clarifying it.
Ethan Mollick
The Vertical Wave: Slow, Invisible, and Disruptive
The second wave is vertical. This is where AI doesn’t just sit on top of work but gets woven into specific processes and often requires those processes to be redesigned.
That’s harder. It demands governance, trust, and integration with messy company systems. Which is why most vertical AI use cases still get stuck in pilots.
But when it works, it’s not about saving minutes. It’s about restructuring weeks of work. AI agents drafting credit memos, coordinating supply chains, or even replicating complex academic research. The gains here aren’t incremental; they’re transformative.
This is where the “sinkhole” analogy fits. On the surface, not much looks different. But underground, workflows are hollowing out and being rebuilt. When the shift breaks through, it won’t look like a shiny new assistant. It will look like parts of the organisation suddenly working in an entirely different way.
Proof or Narrative?
Do we have proof this is happening? Not yet in full. What we have are signals.
- Tasks. AI models can now complete complex, hours-long tasks at near-human levels:contentReference[oaicite:4]{index=4}. That doesn’t replace whole jobs, but it reshapes them.
- Jobs. Stanford data shows that since AI adoption accelerated, early-career workers (ages 22–25) in AI-exposed occupations saw employment fall by 13 percent compared to peers. That suggests real labour market effects are already showing, concentrated at the entry level.
So no, AI isn’t just hype. It’s not stagnating either. It’s moving quietly, in ways that don’t always show up on the surface yet.
Why I Wrote This
I hear the disappointment a lot: “I thought by now AI would have changed everything.” It’s a fair reaction. Horizontal AI has been loud but shallow, and its benefits are often invisible in the balance sheet.
But I also see the deeper shift starting: the vertical integration of AI into real workflows. That’s harder to spot, and slower to scale, but it’s also where the genuine disruption lies.
So if you’re wondering whether the AI hype is fading, my take is this: the surface is calm, but the ground underneath is moving. The real wave hasn’t hit yet.
Horizontal AI reshapes how we do work today, for better and sometimes for worse.
Vertical AI will redefine what work is tomorrow.
Rob Hoeijmakers
Rob Hoeijmakers