Engineering Flow
AI-first executive view · AI teams · last 30d
AI Bottom Line
Delivery volume was flat at 204 issues done, while active flow remained slow at 294.2h lead time and 144.8h cycle time, with client waits and aged intake continuing to hold back throughput.
4 cautions ›
- · ADS — AI cohort n=1 (<10) — Lead delta directional only
- · IMC — no AI-assigned completions in 30d — cohort comparison unavailable
- · TAA — AI cohort n=9 (<10) — Lead delta directional only
- · TDOT — no AI-assigned completions in 30d — cohort comparison unavailable
Headline KPIs · last 30d
AI insight
- Where AI helps:AI-owned work shows faster readiness and lead time in this period, and AI-touched work shows the same pattern at larger sample size.
- Where AI hurts:AI benefit is not broad-based yet: AI-owned adoption is concentrated in Test AI Agents and Platforms Evolution, while Advertisement is minimal and Tripledot and Incident Management are at zero.
- Next focus:Reduce intake age and active client-wait queues before adding more WIP, starting with Platforms Evolution, Advertisement, and Tripledot.
AI in teams · 5 teams
12 completed · 100.0% AI-touched · 2 active agents
Bottom line
TAA improved in throughput. AI-vs-Human cycle comparison is directional only (sample too small) (n=9/3); the main AI bottleneck is Ready For Review.
Issues done
12
n/a vs prev
AI-owned share
75%
+75.0 pp vs prev
AI-touched
12 (100%)
Lead Time (active)
3.1d
Cycle Time (active)
1.3d
TTR p50
0.1h
PR Review p50
—
WIP active
2
WIP blocked
0
AI insight
Where AI helps
No AI-owned metric is currently faster than the human cohort.
Where AI hurts
No statistically stable AI regression detected. Directional-only signals below numerically favor humans but sample size is below threshold.
Directional only (sample too small)
- · Cycle Time (active): AI 3d vs Human 11.1h (n=9/3, below 20)
- · Lead Time (active): AI 3.1d vs Human 1.6d (n=9/3, below 20)
- · Time to Ready p50: AI 0.1h vs Human 0.1h (n=9/3, below 20)
Top AI bottleneck
Active agents · top 2
84 completed · 38.1% AI-touched · 1 active agent
Bottom line
EVO declined (-44.4%) in throughput. AI-owned work moves faster on cycle (n=24/56); the main AI bottleneck is Ready To Merge.
Issues done
84
-44.4% vs prev
AI-owned share
33%
+13.4 pp vs prev
AI-touched
32 (38%)
Lead Time (active)
22.1d
prev 19.7d
Cycle Time (active)
6.7d
prev 6.3d
TTR p50
12.3d
prev 6.8d
PR Review p50
—
WIP active
42
WIP blocked
5
AI insight
Where AI helps
- · AI Cycle Time (active) 6.4d vs Human 7d (n=24/56)
- · AI Lead Time (active) 7.1d vs Human 30d (n=28/56)
- · AI Time to Ready p50 22.6h vs Human 16.9d (n=24/56)
Where AI hurts
No AI-owned metric is currently slower than the human cohort.
Top AI bottleneck
Active agents · top 1
AI Period: compare · intro 2026-04-02
42 completed · 38.1% AI-touched · 2 active agents
Bottom line
ADS improved (16.7%) in throughput. AI-vs-Human cycle comparison is directional only (sample too small) (n=1/41); the main AI bottleneck is Ready For Review.
Issues done
42
+16.7% vs prev
AI-owned share
2%
-8.7 pp vs prev
AI-touched
16 (38%)
Lead Time (active)
3.8d
prev 7.5d
Cycle Time (active)
2.9d
prev 1.8d
TTR p50
0.1h
prev 0.5h
PR Review p50
—
WIP active
30
WIP blocked
14
AI insight
Where AI helps
No statistically stable AI advantage detected yet. Directional-only signals below numerically favor AI but sample size is below threshold.
Where AI hurts
No AI-owned metric is currently slower than the human cohort.
Directional only (sample too small)
- · Cycle Time (active): AI 1.1d vs Human 3.8d (n=1/41, below 20)
- · Lead Time (active): AI 1.1d vs Human 3.9d (n=1/41, below 20)
- · Time to Ready p50: AI 0h vs Human 0.1h (n=1/41, below 20)
Top AI bottleneck
Active agents · top 2
AI Period: compare · intro 2025-08-01
3 completed · 33.3% AI-touched · 1 active agent
Bottom line
IMC improved (50.0%) in throughput. AI cohort too small for cycle comparison (n=0/2); no specific AI bottleneck stands out.
Issues done
3
+50.0% vs prev
AI-owned share
0%
+0.0 pp vs prev
AI-touched
1 (33%)
Lead Time (active)
69.1d
prev 23.4d
Cycle Time (active)
31.7d
TTR p50
0.5h
prev 15.5d
PR Review p50
—
WIP active
3
WIP blocked
0
AI insight
Where AI helps
No AI-owned metric is currently faster than the human cohort.
Where AI hurts
No AI-owned metric is currently slower than the human cohort.
AI cohort too small for direct comparison on this team.
Top AI bottleneck
no AI-owned cohort with stage data
Active agents · top 1
21 completed · 9.5% AI-touched · 2 active agents
Bottom line
TDOT declined (-61.8%) in throughput. AI cohort too small for cycle comparison (n=0/21); no specific AI bottleneck stands out.
Issues done
21
-61.8% vs prev
AI-owned share
0%
+0.0 pp vs prev
AI-touched
2 (10%)
Lead Time (active)
24d
prev 22.2d
Cycle Time (active)
4d
prev 8.6d
TTR p50
19.9d
prev 7d
PR Review p50
—
WIP active
28
WIP blocked
18
AI insight
Where AI helps
No AI-owned metric is currently faster than the human cohort.
Where AI hurts
No AI-owned metric is currently slower than the human cohort.
AI cohort too small for direct comparison on this team.
Top AI bottleneck
no AI-owned cohort with stage data
Active agents · top 2
Active bottlenecks · top 5
Top teams · by throughput
| Key | Team | Throughput | Lead Time (active) | Cycle Time (active) | Active WIP | Blocked WIP | AI % |
|---|---|---|---|---|---|---|---|
| EVO | Platforms Evolution | 84 | 22.1d | 6.7d | 42 | 5 | 33% |
| ADS | Advertisement | 42 | 3.8d | 2.9d | 30 | 14 | 2% |
| TDOT | Tripledot | 21 | 24d | 4d | 28 | 18 | 0% |
| TAA | Test AI Agents | 12 | 3.1d | 1.3d | 2 | 0 | 75% |
| IMC | [Engineering]: Incident Management | 3 | 69.1d | 31.7d | 3 | 0 | 0% |