Which types of tasks do you still rely on human expertise for, rather than automating with AI, and why?
The answers agree that tasks needing human judgment, empathy, trust-building, or contextual understanding are still best handled by people rather than automated by AI.
9 answers
Answers

Mukhtar Ahmed
Chief Executive Officer · BC Platforms
Mechanistic Plausibility
Mukhtar Ahmed says human judgment remains important for mechanistic plausibility, because models are trained on a documented record that contains publication bias and incomplete biology and rarely captures the tacit reasons experienced scientists stop believing in a mechanism.
The answer starts at 10:54.
The AI in Business Podcast · Listen on Apple Podcasts

Wade Foster
Co-founder and CEO · Zapier
High‑Emotion Communications
Wade Foster says anything involving high emotion, such as negative performance reviews or constructive feedback, needs a personal investment rather than AI, because outsourcing the most challenging communication risks empathy atrophy.
HBR IdeaCast · HBR IdeaCast · Listen at 11:49

Maik Taro Wehmeyer
Co-Founder & CEO · Taktile
Final Credit Decisions
Maik Taro Wehmeyer says the final credit decision should still be made by a credit policy and decision engine rather than agents, since agents suit pre-analysis tasks like reading documents but not the end-to-end decision.
Fintech Leaders · Fintech Leaders · Read

Jerry Kaplan
Adjunct Lecturer, Freeman-Spogli Institute for International Studies · Stanford University
Managing AI Agents
Jerry Kaplan says the key skill for humans in the future is managing AI agents, and that this is where we're heading.
The answer starts at 54:54.
The Good Fight · Listen on Apple Podcasts
Anders Hejlsberg
Technical Fellow · Microsoft
Understanding Business Context
Anders Hejlsberg says someone still has to understand what's going on and how it relates to the business problem and the organization, arguing it's wrong to view AI as having done it all.
Ryan Peterman · Watch on YouTube
Nancy Oriol
Faculty Associate Dean for Community Engagement · Harvard Medical School
Building Trust
Nancy Oriol says building trust is work only humans can do, arguing AI cannot sit with a bus driver and build enough trust in five minutes to get him to allow a finger prick, because these are human problems that need a human presence.
Imagination in Action · Watch on YouTube

Sasha Mitts
User Experience Researcher · Meta
Counting Tasks
Sasha Mitts says AI is not good at counting, such as telling how many times the letter R appears in strawberry, because the things models are good at differ from what human brains are good at.
The answer starts at 19:52.
The Curiosity Current: A Market Research Podcast · Listen on Apple Podcasts

Zeke Emanuel
Vice Provost for Global Initiatives · University Of Pennsylvania
Telling Cancer Diagnoses
Zeke Emanuel says a human physician, not an AI, should tell someone they have cancer, asking whether you would want an AI or a human deliver that news.
"Second Opinion" with Christina Farr, Ash Zenooz MD & Luba Greenwood JD · "Second Opinion" with Christina Farr, Ash Zenooz MD & Luba Greenwood JD · Listen at 48:22

Amit Spitzer
CTO & CISO · Glilot Capital
Prompt Engineering
Amit Spitzer says people still need human experience because you must know what to ask AI and how to ask it, a skill he calls prompt engineering and says is becoming more critical.
N/A · N/A · Listen at 26:39