Most white collar work runs on trust and relationships, not correct answers. And that’s exactly why AI won’t replace it. That’s the core argument from Andrew Marble, as highlighted on Hacker News, in a piece that cuts through the noise around AI job displacement with a surprisingly simple framework.
Marble draws a distinction between two types of questions we answer every day. Type 1 is relationship-based: you ask because you value someone’s perspective, want to think out loud, or need to feel understood. Type 2 is transactional: you need a specific answer, fast. AI is exceptional at Type 2. For Type 1, it’s largely irrelevant.
The Strategy Consulting Test Case
Marble zeroes in on strategy consulting as the perfect example. For years, AI evangelists have claimed that better slide decks and faster data analysis would make consultants obsolete. But that misreads what consulting actually is.
“Buyers aren’t asking for a correct answer, they are asking for advice from someone whose opinion they respect,” Marble writes. Companies hire consultants for judgment, trust, and the cathartic process of explaining their situation to someone who gets it. You can’t replicate that with a chatbot, no matter how sophisticated the model.
This observation extends well beyond consulting:
- Most business decisions aren’t about finding the “right” answer. They’re about setting a plausible course and adjusting based on human judgment.
- Organizations run on social communication. Teams coordinate, negotiate, and build consensus through relationships.
- Government and military rely on human organization even more because they lack market feedback loops to course-correct automatically.
What AI Actually Replaces
The article doesn’t dismiss AI’s utility. Marble himself used a chatbot to solve a Python/pandas problem in seconds. That’s a clean Type 2 interaction: specific question, verifiable answer, move on.
What stands out here is how precisely this maps onto what we’re seeing in practice. Companies deploying AI successfully tend to use it for sub-processing: research synthesis, code generation, data transformation, draft creation. The work that requires someone to care about the output, stand behind a recommendation, or navigate office politics? Still entirely human.
The Misconceptions Worth Busting
Several popular beliefs deserve pushback:
- “AI will replace knowledge workers because it knows more.” Knowledge was never the bottleneck. Wikipedia didn’t replace professors, and LLMs won’t replace advisors.
- “Better AI means fewer consultants.” The value of consulting was never primarily informational. It’s relational. Telling your board “ChatGPT recommended this” doesn’t carry the same weight as “McKinsey recommended this.”
- “White collar work is mostly information processing.” Marble argues the opposite; most of it is social coordination dressed up as information work.
What This Means for AI Practitioners
If you’re building AI products or deploying them inside organizations, Marble’s framework offers a practical filter:
- Audit your use cases by question type. Type 2 tasks (lookup, calculation, drafting, code fixes) are strong AI candidates. Type 1 tasks (strategy, advice, relationship management) are not.
- Stop selling AI as a replacement for human judgment. Position it as a tool that handles the transactional layer so humans can focus on the relationship layer.
- Recognize that “how good AI gets” is beside the point for most white collar displacement fears. The constraint isn’t capability; it’s that humans want to interact with other humans on consequential decisions.
This isn’t a contrarian hot take. It’s an observation that’s been hiding in plain sight while the industry chases AGI timelines and automation percentages. The social fabric of work isn’t a bug to be engineered away. It’s the actual product.
Marble’s full analysis is worth reading for anyone building workforce strategy around AI. The original piece is available through Hacker News.