When AI Augmented Novice Support Agents: Uneven Productivity Gains at Work
A real deployment asking how AI fits into work and who benefits, rather than treating free access itself as the outcome.
Updated Oct 16, 2026
Suggestions from AI, choices left to agents
Research described by MIT Sloan analyzed an AI assistant introduced in software-company customer support in 2020 and 2021. It suggested replies and resources, while agents chose whether to adopt them. The start date records only an observed rollout year, not a specific month or day.
Average effects differed from skill-group effects
The research account reports approximately 14% higher average productivity, concentrated among newer and less-skilled agents. Efficiency gains for the most skilled group were close to zero. This did not measure nationwide free provision or net job creation.
Ask about workflow design and beneficiaries, not just distribution
When considering the video’s entry-level job concerns, distinguish designs offering suggestions to people from those removing people. Review authority, training and task allocation may change how the same AI operates. This case is a comparison against both blanket replacement claims and guaranteed productivity gains.
Do not turn a workplace result into national-policy success
This is AI analysis based on an account of research in one company and job setting, not personal experience. It does not verify Korea’s policy scope, budget, domestic-model quality, employment or fertility, or establish that national benefits exceed costs. Better novice-agent performance does not guarantee new hiring or every young person’s job. The overall deployment end date was not established and remains blank.
Sources
- MIT Sloan IWER — Generative AI and Worker ProductivityAccessed Oct 16, 2026