🤖 AI AGENTS
Make the whole company faster — without DS in the loop
Reclaim hundreds of human hours every month and pass the empowerment test: teams build and run their own agents. Hours must be measured (runs × time saved), not estimated — that's what makes the program credible to leadership.
LVL 2 · 180 h (July)
🏆 LVL 3 · 250 h (Aug)
⚡ ~135 h/mo
XP = verified productive hours/month from the agents catalog. Currently ~75% of the July level-up.
🚢 ETA MODELS
Make PETA decisively beat carrier ETA where customers feel it
Customers judge us on 0–7 days-ahead accuracy. This is the quarter the stack — new model heads, carrier-anchored rules, container-level trust — beats production by ≥5 pp weighted-OTPA on T≤7d, sustained 14 days.
🏆 VICTORY · +5.0 pp
+3.63 pp
Rules deployment already banked +3.63 pp (74.85 vs carrier 71.21). Transit Time, Port Stay & meta-models carry the rest.
🧠ETA-to-port model v2In prod since ~May 25 · +2–3% over old model in A/B
📏pred_clip_to_carrier (δ=0.75d)Winning rule live · +3.63 pp weighted-OTPA
🤖Code Review AgentEvery portcast-api MR · ~40 h/mo · ~$6/day
🔎Company Research bots ×2slack-er (~30s) + researcher (~8min) · merging in J3
📄Proposal GeneratorIn use by Sales · RFI/RFP extension queued (Y3)
📊Portcast JuniorPostHog metrics on call in Slack
🧮AI Analyst (beta)kpi_dashboard · grows into Self-Serve KPI Analyst (A1)
🛠️Prototypes MCPSelf-serve dashboards on Cloud Run · <$10/mo
👥Claude Team pilot10 users across teams · onboarding sessions running
🚧ML GatewayA/B router mid-build — single-config (ETA J3)