A study of over 700 software firms found that AI coding agents significantly boost developer productivity. These agents lead to a notable increase in the amount of code generated, commits made, and pull requests submitted. However, this surge in raw code production does not translate into a proportional increase in shipped software. The primary reason for this discrepancy lies in the code review process. Human code review acts as a significant bottleneck, taking considerably longer after the introduction of AI coding tools. The time required for reviews balloons by nearly half on average. Consequently, pull requests are much more likely to require revisions, and reviewers leave a greater volume of comments. This increased review workload led to a 14 percent rise in the share of workers performing code reviews. Despite the use of AI in code review, humans remain responsible for the vast majority of this critical step. Ultimately, the study found little evidence that firms increased overall software output or reduced employment due to AI coding agents.