Adoption
Orlo can be adopted as open-core primitives, a managed platform, a hosted SaaS product, or a self-hosted/on-prem control plane. The right path depends on what the organization needs to prove and operate.
The Main Adoption Question
Ask:
Are we exploring Orlo's primitives, or do we need a shared operating system for production domain AI?
If you need a package, start with Open Core. If you need a repeatable operating model across teams, start with Orlo Platform.
Choose Your Path
| Situation | Best Starting Point | Why |
|---|---|---|
| One technical team wants validation or adapters | Open Core | Low-friction adoption inside an existing stack |
| A platform team is standardizing AI across teams | Platform | Shared tasks, evaluations, deployments, monitoring, and governance |
| A domain team needs to prove a model works | Platform Quickstarts | Task-first workflow without building an eval harness |
| Risk or compliance needs evidence | Governance | Policy-to-control mapping and operational proof |
| You already use an LLM gateway | Orlo And LLM Gateways | Boundary between model traffic management and decision evidence |
| Agents can use tools or change state | Agent Governance Overview | Step-level policy, approvals, runtime limits, and traces |
| Data residency or sovereign deployment matters | Platform Overview | Provider-agnostic and self-hostable operating model |
Adoption Stages
Stage 1: Prove Fit
Bring one task and representative examples.
- define the task
- upload a dataset
- evaluate two or more candidate models
- inspect confidence intervals and failure cases
- confirm whether Orlo's task model fits the workflow
Stage 2: Prove Production Control
Deploy the best acceptable candidate behind Orlo.
- freeze task version and deployment configuration
- enable validation
- capture inference logs
- inspect routing and debug/audit metadata
- submit feedback and verify promotion flow
Stage 3: Expand Across Teams
Move from one task to a shared operating model.
- create tenant/org structure
- issue scoped API keys
- define user roles and reviewer responsibilities
- standardize datasets, rubrics, routing policies, and validation patterns
- monitor feedback pressure, validation failure rates, latency, and cost
Stage 4: Govern Agents And High-Risk Actions
Add step-level governance when agents can use tools, retrieve sensitive context, or change external state.
- create agent sessions
- define tool policies
- set runtime limits
- require approvals for high-risk actions
- promote reviewed traces into datasets
- score trajectories, not just final text
Stage 5: Operational Assurance
Use Orlo as evidence infrastructure.
- review deployment snapshots
- inspect validation and retrieval attribution
- audit approval decisions
- monitor drift and circuit breakers
- export or sample traces for internal review
- feed incidents and corrections back into evaluation