AI Is Eroding Workplace Trust

Summary

This opinion essay argues that generative AI is disrupting a workplace assumption: people generally understand the code, briefs, explanations, or documents they submit. AI can now produce polished, and sometimes correct, work without the author having checked or understood it, making a finished-looking result harder to interpret. The author says this has changed code review from asking whether a change is good to first asking whether its author understands what they are sending. That shift increases scrutiny and can impose a lasting cost on collaboration, because trust previously allowed colleagues to focus only on the parts that needed another review. The essay also notes that stylistic signs of AI-generated writing or answers that fail to address a specific question may create suspicion, even when that suspicion is not always fair. The author acknowledges having merged a correct AI-written fix without understanding it, using that experience to argue that putting one’s name on work should mean understanding and standing behind the result. They recommend explicitly labeling early ideas and proofs of concept, explaining what was checked and what remains uncertain, and treating AI as an aid rather than a substitute for judgment. Engineers are urged to read their own diffs and documents, understand the cause of a bug and the proposed fix, and ask reviewers what was tested. Engineering leaders should make “ready for review” meaningful while keeping evidence proportional to risk. Finally, the essay calls for making unfinished work safe to share and for leaders to model the standard they expect, warning that sending unreviewed AI-generated material can force colleagues to redo the sender’s work before they can do their own.