All the major AI platforms are loudly pushing “AI automation studios” where rank-and-file knowledge workers can create sophisticated AI-powered workflows across their enterprise workspace environments. These tools are genuinely impressive, as workers can now use plain language to describe what they want and how it should work, while AI handles the triggering, filtering, and decision-making that previously required a programmer’s mindset to build.

These tools are awesome.

But they are NOT the AI-powered future of knowledge work I’ve been writing about.

The automation studio assumption

The pitch behind automation studios (such as Microsoft Copilot Studio or Google Workspace Studio) goes something like this:

  1. Knowledge workers have repetitive workflows they’ve always wanted to automate but couldn’t because they weren’t programmers.
  2. Now AI lowers the barrier.
  3. Give them a visual canvas, connectors into enterprise apps & data sources, and natural language configuration.
  4. They’ll finally build the automations they’ve been dreaming of!

There are two problems with this:

  • Knowledge workers do lots of different things, and most of their work is NOT simple and repetitive enough to automate. The automation studio concept assumes workers have stable, repeatable workflows worth the investment of designing and maintaining. But most knowledge work isn’t like that. It’s fluid, reactive, and different every day.
  • Even for things that could be automated, most knowledge workers don’t have the time, wherewithal, or programmer’s mindset to stop working while they figure out how to build an automation. Most workers think like managers, not programmers. They don’t want to design workflows, they want to hand off tasks.

We’ve been hearing this same automation pitch for decades

Robotic Process Automation (RPA) was sold in part on the premise that workers could finally automate their own workflows without IT involvement. The tools got easier over time and the connectors multiplied. But after years of investment, most enterprises only have a handful of RPA specialists building bots while the vast majority of workers never touch the platform.

The same thing happened with low-code tools, citizen development platforms, and every other “democratize automation” initiative. The 1% of workers who think like programmers build amazing things. The other 99% keep doing their jobs the way they always have.

This isn’t a failure of the tools—they work as they should. Rather the failure comes from thinking knowledge workers will stop working to investigate, experiment, and try to learn those tools.

To be clear, “AI powered automations” will absolutely be a thing. But those automations won’t come from workers building workflows in studios. They’ll come workers interacting with AI, asking it to do things for them.

This is where AI as the interface becomes real. When AI platforms become the primary way workers interact with their digital environment, apps recede into the infrastructure. The automation piece will be part of that same shift.

AI’s “iPhone Moment” is coming

Today’s AI automation studios are powerful, and I love that they exist. My argument isn’t that they’re not useful, rather I’m saying don’t look at these AI studios as “the future”—they’re really more of a stepping stone on the larger path towards true human-AI collaboration at work. They’re the “boring infrastructure” plumbing that AI labs need to put in place now so that future AI agents have all the connectors, access, and context they need to function.

At some point, AI will be good enough that everyone will experience its full potential without having to stop working to learn new tools. This will be similar to how the iPhone changed the world. It was so intuitive your grandparents could use it, while hiding the decades of technical magic that made it possible.

AI will get there too, and everyday workers won’t need to use automation studios.

The bottom line

As I mentioned a few times, I like the AI-powered automation studios, and for the workers who use them, they’re great. But they solve a narrow use case for a narrow segment of workers.

The longer term AI-powered vision is not to turn knowledge workers into automation designers. They’re managers, not programmers. They want to hand off tasks, not build systems. The future isn’t a million workers hunched over workflow designers—it’s workers interacting with AI as a coworker, in the same environment where they’ve always worked, using all the same tools, processes, and workflows they always have.

Read more & connect

Join the conversation and discuss this post on LinkedIn. You can find all my posts on my author page (or via RSS).