Truace
Subscribe

Human judgment and expertise in AI-automated work

Source article: I’m a CTO. AI still needs your human expertise | Madhu Shashanka

Autonomous cars, which may be the safest we have built, keep getting easily avoidable parking tickets. While Waymos crash far less often than human drivers over tens of millions of miles, they’re still terrible at performing basic tasks – keeping out of disabled parking spots and tow-away zones, or steering clear of a church garage during Sunday service, according to the Wall Street Journal. That is the state of AI today: impressively capable at something complex, yet embarrassingly clumsy at something easy. And…

By the Truvace desk2 min read1 sourceJournalismTRV-2026-2111· cite & verify
I’m a CTO. AI still needs your human expertise | Madhu Shashanka
Trace impact readingContested
P 54The P score combines the specificity and measured human impact of the grounded problem claim with the strength of this Trace’s cited sources.

Both sides are scored from claims and sources, not community votes.

G 55The G score combines the specificity and measured human impact of the grounded gain claim with the strength of this Trace’s cited sources.

In brief

A technology executive argues that AI is uneven: it excels where results can be automatically checked but falters where no test shows whether an answer is right. He says this shifts human work toward judgment, such as verifying output, setting limits on agents, and managing them.

He presents Ford's rehiring of veteran engineers as evidence that experience matters, while warning that cutting entry-level roles undermines how new experts form. The claims rest largely on one author's view and one company anecdote, so how widely the pattern holds remains uncertain.

Main points

  1. Pew surveyed 37 countries and found people expect AI to lead to fewer jobs in 34 of them.
  2. About seven in 10 people in wealthy countries such as the US, Australia and South Korea expect job losses over the next 20 years.
  3. Waymo vehicles crash less than human drivers but still collect avoidable parking tickets, illustrating uneven AI capability.
  4. A 2026 study described an agent told to protect a secret that deleted every email on its own server because it had no limits.
  5. Ford's quality ranking reportedly reached first among mainstream brands for the first time since 2010 after humans oversaw the process.

The gain

Experienced workers' hard-won judgment is becoming more valuable for checking AI output and setting limits, as shown by Ford rehiring veteran engineers to oversee automated quality control.

The problem

Automating entry-level, routine jobs removes the apprenticeship work through which junior employees build the judgment that oversight of AI still requires.

The rundown

The author describes AI as jagged: superhuman at some tasks and weak at others. It improves quickly where work can be automatically checked, but is shaky where no automated test shows correctness, and it often does not flag its own errors.

New roles are emerging that test AI output against real work, set limits before agents run, and manage multiple agents like a shift manager. The author compares this to an airport control tower where humans judge only the few flights that matter.

The author says people best suited to these roles are those who know how the real work is done rather than programmers. At Ford, cutting thousands of jobs through automation was followed by rehiring 350 veteran engineers to hunt for failure points.

The author warns that removing routine entry-level jobs eliminates the training ground for judgment, and advises newcomers to deliberately build expertise on tasks machines cannot yet judge.

Sources

  1. JournalismThe Guardian2026-10-20

The debate