Research
What the evidence actually says.
Spendassay publishes the research its method rests on, including the findings that cut against us. Every study is linked to its primary source and reported with its effect size, so you can check the claim rather than take ours.
Reviews
Does AI actually make developers faster?
The published evidence, linked and dated, with effect sizes and a comparison table. Includes the results that cut against the case for AI tooling, because a review that only cites the favourable studies is marketing.
The sources we cite
8 published studies underpin the method. Each is listed with its finding on the methodology page, and compared side by side in the review above.
| Source | Year | Measures |
|---|---|---|
| METR | 2025 | Task completion time, with and without AI tools |
| Faros AI | 2026 | Delivery quality against AI-accelerated throughput |
| GitClear | 2025 | Code churn and duplication over time |
| DORA | 2025 | DORA delivery and stability metrics under AI adoption |
| CACM · Ziegler et al. | 2024 | Perceived productivity vs. acceptance rate |
| LinearB · APEX | 2026 | Adoption curve and payback window |
| Microsoft · Viva | 2026 | Minimum group size for reporting |
| DX (now part of Atlassian) | 2026 | Longitudinal engineering velocity across dimensions |
Measure your own numbers, not an industry average.
The research says the effect varies by team. That is the argument for measuring yours.