AGI Analytics
Top 10 apps
The apps production users run the agent in most. Together they are 43.0% of production traces in the last 14 days.
| # | App | Share of production traces | Traces | Share | Pass rate | vs 14 days before |
|---|---|---|---|---|---|---|
| 1 | YouTube com.google.android.youtube | 232 | 8.9% | 90.6% | +263% | |
| 2 | Chrome com.android.chrome | 196 | 7.5% | 84.7% | +513% | |
| 3 | Settings com.android.settings | 149 | 5.7% | 81.5% | +473% | |
| 4 | Calculator name:calculator | 135 | 5.2% | 95.7% | +238% | |
| 5 | com.whatsapp | 115 | 4.4% | 95.5% | +360% | |
| 6 | Tiktok name:tiktok | 66 | 2.5% | 93.0% | +187% | |
| 7 | Gmail name:gmail | 61 | 2.3% | 86.1% | +408% | |
| 8 | Phone name:phone | 58 | 2.2% | 89.7% | +346% | |
| 9 | Maps name:maps | 55 | 2.1% | 91.9% | +450% | |
| 10 | Google Play Store name:google play store | 53 | 2.0% | 84.4% | +783% |
Success rate
The share of traces that passed over the trailing week, by where they came from. Interrupted runs (stopped by the person, closed with the app, or never finished) are not counted.
Production
83.3%
+1.0 ptsvs 82.2% last week7-day success rate, as of Nov 14
82.8% overall · 2,194 passed · 455 failed · 2,649 traces · 41.1% of all traces
Partners
69.5%
+4.9 ptsvs 64.6% last week7-day success rate, as of Nov 14
65.6% overall · 754 passed · 396 failed · 1,150 traces · 14.4% of all traces
Benchmarks
83.4%
+8.5 ptsvs 74.9% last week7-day success rate, as of Nov 14
79.2% overall · 3,464 passed · 908 failed · 4,372 traces · 44.5% of all traces
Failure analysis
Why traces failed, and in which apps. 1,772 failed traces in the last 14 days.
Agent mistake
The guide and plan were right (or silent) and the agent did the wrong thing: a wrong tap, pushing through a verification, stopping early.
64.8%
of failures · 1,148 traces· show traces
Couldn't be done here
No step was wrong: the account had no data, the request was ambiguous, or the app blocked it, and the agent handled it correctly.
29.3%
of failures · 519 traces· show traces
Planner mistake
The agent's plan dropped or changed something the guide covered.
2.5%
of failures · 44 traces· show traces
All causes
Share of failed traces, with the change against the 14 days before. Click a cause to list its traces.
- Agent mistake64.8%−8.6 pts
- Couldn't be done here29.3%+8.7 pts
- Planner mistake2.5%−0.5 pts
- Guide gap2.3%flat
- Unknown0.4%flat
- App or device0.4%flat
- Unclear task0.3%flat
Apps with the most failures
Share of failed traces that happened in each app. Click an app to list its traces.
Eats7.6%+7.6 pts
Settings7.4%+3.5 pts
YouTube7.3%+4.8 pts
Disneyplus6.6%+6.6 pts
Uber5.5%+2.9 pts
Instagram4.9%+3.7 pts
WhatsApp4.0%+1.4 pts
DoorDash3.4%+1.5 pts
- 166 other apps53.2%−10.0 pts