Founder & Head of Design
Determining what to automate and what to own
Clear the deck for big ideas, without giving them away.
Have you ever thought that AI is quietly taking your judgment? It rarely happens in big, obvious ways. It shows up when we stop questioning outputs and start offloading more and more of our thinking. First the busywork, then the decisions behind it.
We’ve already moved from asking if AI can do our jobs to asking how much of our calendar we can hand over to it. And that’s where a new tension appears: the Automation Paradox.
The paradox is simple. The more we automate the thinking parts of our process, the more vulnerable our products become. If we outsource our analysis entirely to AI, we’re losing the context, and the why of our work.
But you may think (and for good reasons): “If I save time and still have a job, where’s the danger?”
Even if AI will not replace you, over-automating might accidentally give away the very thing that makes you valuable: critical thinking.
How AI transforms jobs
There is a lot of noise about AI destroying roles. However, in reality, it’s just reshaping them.
According to the Stanford 2026 AI Index Report, we are no longer experimenting with AI. We are living in it. 88% of organizations have fully integrated AI into their daily workflows, leading to a massive 26% productivity boost in design and software tasks.
But there is a catch.
The report highlights a 20% drop in employment for junior-level roles (ages 22–25). And this happens because the grunt work that used to be the training ground for juniors is now being handled by machines. To stay relevant, you have to move from being a task-doer to an architect of intent.
If you only do what the AI can do, you're competing with a tool that is 26% faster than you.
The risk of over-automating comes down to how these models work. The Stanford 2026 AI Index makes it clear. While AI is great at basic tasks, its success rate in complex planning and multi-step reasoning often sits below 50%. This happens because LLMs are probability engines, not reasoning engines. They just guess the most likely next word or pixel based on what has already been done a million times.
And to be totally honest, the goal should be to use the machine to clear the deck so you can work on the high-value tasks that a model simply cannot grasp, not to work less.
Handing over the volume, but keeping the value
To stay relevant, you have to divide your to-do list into two distinct buckets. And, according to research published by the University of California, Berkeley, the most effective strategic approach is balancing automation (replacing tasks) with augmentation (enhancing human work).
You can think of this balancing like this:
- High-Volume, or the AI territory, needs automation;
- and High-Value, or the human territory, needs augmentation.
The high-volume bucket is something like generating 20 layout variations or resizing 50 assets for every breakpoint, or drafting the first version of JIRA tickets, PRD outlines, or boilerplate CSS. On the other hand, the high-value bucket is about deciding how your Design System tokens actually map to business logic, or explaining the reason behind a flow to the dev team so they build the right logic.
Reality check
Depending on your role, the line between Automate and Own is where your career will be won or lost. As Microsoft’s 2025 Future of Work Report suggests, the next frontier is collective productivity, how teams use AI to bridge gaps without losing the norms of human collaboration.
In design, AI is a master of the 90% solution, as it can deal with building components and generating variants at scale. But the final 10% is where the product actually lives or dies. This is the realm of the edge case, and the real-world test that no model can fully simulate. True insight requires a human brain to sit with the friction of a problem until a solution emerges. If you let a model summarize your user's pain, you’ll only ever solve for the symptoms, never the cause.
Even in technical implementation, where code and logic generation have become the norm, the role is shifting toward high-level oversight. You can automate the boilerplate, but you must own the architecture, ensuring the trust, ethics, and traceability of a system that will outlive the current AI trend.
Ready to stop guessing and start architecting your workflow? Explore the AI for Product Design Workflow.
