Sensitive work in-house. Volume with us. One dataset.
Hybrid lets you split the work by task type, sensitivity, or season — while every label, from either side, meets the same guideline and lands in the same dataset with the same provenance.
Most mature programs end up hybrid. Expert adjudication and anything involving proprietary processes stays with your team; high-volume, well-defined labeling flows to our managed workforce; raw data can remain in your storage under a hybrid deployment. The agentic feedback loop runs across both, so a guideline clarification made by your expert reaches our annotators on their next task.
What is included
- Task-type and sensitivity routing rules
- Managed capacity for defined task classes
- Your reviewers as final sign-off where you choose
- Optional hybrid deployment with data in your storage
- Unified quality dashboards across both workforces
- Single provenance trail for the whole dataset
Who it fits
- Enterprise programs with confidentiality or residency constraints
- Teams with strong experts but limited annotator capacity
- Programs with seasonal or release-driven volume spikes
- Organizations transitioning from in-house-only to scaled delivery
Getting started
Map the work
Decide which task classes stay internal and which flow to managed capacity.
Set routing
Rules route tasks by type, sensitivity, or queue depth.
Align quality
One guideline, one gold set, one set of thresholds across both teams.
Operate
Dashboards show the combined program; recommendations flow both ways.
Questions
Can raw data stay in our environment?
Yes, with a hybrid deployment: Agentuor orchestrates while data remains in your storage.
Who owns guideline changes?
You do. Clarifications propagate to managed annotators automatically.
Can the split change over time?
Routing rules can be adjusted at any time without moving data.