Claim
AI adoption in engineering teams creates an asymmetric information problem: wins get celebrated publicly in demos and all-hands meetings, while costs accumulate privately in incident reviews, on-call rotations, and code quality retrospectives that AI enthusiasts rarely attend. Without deliberate feedback loop design, the gap between enthusiasts and skeptics compounds into an organizational reliability risk.
Mechanism
The information asymmetry runs in one direction. Productivity gains are legible, repeatable, and promotable. Reliability degradation, institutional knowledge loss, and increased on-call burden accumulate in forums the AI-forward engineers do not see. Skeptics experience the downstream costs but lack real-time channels to surface them. Enthusiasts observe the gains but not the full cost stack. The result is not a disagreement about values but about facts that land in different rooms. Bridging the gap requires engineering feedback systems deliberately, using the same discipline applied to code reliability: timely, precise, and relevant signals that reach the people who need them.
Conditions
Holds when: organizations have adopted AI tools fast enough that some teams are shipping AI-assisted code while others absorb the downstream reliability, quality, or knowledge effects.
Fails when: the team is small enough for full contextual overlap, or AI adoption is slow enough that costs and benefits land in the same forums and the same people.
Evidence
Fin (formerly Intercom) 3x'd engineering output in 9 months measured as merged PRs per R&D headcount. Product defect backlog shrank by more than half. Time from idea to shipped fell 39%. Downtime fell 35%. Majors attributes these results not to AI alone but to Fin's pre-existing engineering discipline, fast feedback loops, and measurement culture.
The 2025 DORA State of AI-Assisted Software Development report, attributed to Nathen Harvey:
"AI magnifies the strengths of high-performing organizations and the dysfunctions of struggling ones."
Majors:
"Feedback loops that are timely, precise, and relevant enable self-awareness."
"Both sides are grappling with a real, alarming, escalating threat to the company's existence."
Signals
- AI-assisted engineers report high productivity satisfaction while on-call engineers report increasing incident frequency.
- Code quality metrics diverge from shipping-speed metrics over the same period.
- AI skeptics are excluded from or underrepresented in forums where AI adoption decisions are made.
- Post-incident reviews name AI-generated code as a contributing factor but the finding does not reach product planning.
Counter-evidence
Majors does not claim skeptics are always right or that enthusiasts are reckless. Organizations with strong pre-existing engineering culture can adopt AI fast without the gap widening. The asymmetry problem is most acute where cross-team visibility and feedback loops are already weak. High-discipline teams like Fin achieved large productivity gains precisely because the discipline was already in place.
Cross-references
- (none in current corpus)