Outreach consoleCONTROLLED MODEL LEARNING

Training

Review → Curate → Train → Compare → Promote

Active model

Current inference baseline

Approved examples
0

Loading

Trainer
Not configured

Runs outside the live email agent

Promotion gate
Evaluation required

Candidate must beat baseline

Training runs

Dataset snapshots are immutable per run; promotion is never automatic.

No training runs yetCurate enough approved examples, then build a run.

Conversation review

Human labels and edited ideal responses become training data. Bad/excluded examples never enter a dataset.

No inbound conversations foundReal inbound threads will become available for curation here.