Every few years, a profession gets declared obsolete by AI.
In 2017, it was radiologists.
The prediction was confident: image recognition had advanced enough that human diagnosticians would be unnecessary by 2020.
Eight years later, we employ more radiologists than before.
Why? Because AI handles discrete tasks exceptionally well, but it struggles with the complete bundle of activities that constitute most jobs:
The consultation
The synthesis
The contextual judgment that happens between the measurable outputs
And once we frame this accurately, our strategic planning becomes much, much sharper.
Thanks for reading,
Robbie Allen
Founder & Managing Director
Automated Consulting Group
PS: If you’re thinking about equipping your team with agents or other AI-powered workflow tools, hit reply — I’d love to hear what you’re working on and share what we’re learning from client transformations.
Key Takeaways:
• AI automates tasks, not roles. Most jobs consist of task bundles plus human judgment (and it resists full automation).
• The real pressure is efficiency compression, not job elimination. The goal should be to get fewer people per unit of output, not zero people.
• Data synthesis is still a barrier. Full automation requires capturing organizational context that the current infrastructure can't provide.
• Framing matters for strategy. Preparing for compression requires different planning than preparing for elimination.
What AI Actually Automates
AI excels at accomplishing discrete tasks. It executes a series of steps with increasing complexity and performs them well.
But a task, by definition, is bounded.
It has clear inputs, defined processes, and measurable outputs.
So if your entire job consists of accepting an image and producing an assessment, AI can eventually handle that sequence.
But most jobs aren't structured this way.
The Task-Bundle Problem
On paper, the radiologist’s work seems simple. In reality, it’s anything but.
Radiologists:
Consult with other physicians,
Integrate multiple data streams,
Navigate ambiguous cases where guidelines don’t neatly apply,
And communicate findings in ways that adapt to patient context and clinician needs.
The diagnosis may be automatable, but the surrounding work that makes the diagnosis actually useful remains deeply human.
You see the same pattern in early-career roles.
Take interns: they take notes, conduct research, prepare materials, join discussions, and offer early assessments. Some tasks are now automated (for example, transcription), but the bundle that creates value doesn’t evaporate.
It adapts.
The Real Conversation We Should Be Having
The anxiety about mass unemployment comes from thinking:
Task automation = role elimination
AI is reshaping how work gets done, but the shift looks more like:
Task automation = Teams can do more with fewer people.
A department that once needed ten people may only need six now. That’s meaningful; it changes hiring ratios, org design, and promotion pathways.
But it does not mean we stop needing developers, creators, analysts, or radiologists. It means their productivity increases, and the structure of their role shift.
It's very important to accurately frame this.
If you assume radiologists are becoming obsolete, you stop training them. But if you understand they’re becoming more efficient while still essential, you invest in augmentation tools and workflow redesign.
The Takeaway for Leaders
The path forward is redesigning roles around human judgment, synthesis, and contextual application, while using AI as a productivity multiplier.
That's how you position yourself correctly for workforce planning over the next decade.
And with that in mind, I'm curious: if AI compresses workforce requirements rather than eliminating jobs entirely, how does that change your team capacity planning over the next few years?
Send me a message if you have questions as you head into 2026.
– Robbie