Jev Task Routing for AI Agent Workflows | AutoJev

Use Jev task routing to choose whether an AI agent should proceed, review deeply, split work, ask for help or stop.

Last updated: 2026-09-18

Jev task routing chooses the next workflow path from options your application defines. It is useful when an agent has enough context to describe the situation but should not rely on an open-ended generation to decide whether to proceed, inspect more deeply, split the task or stop.

The workflow question

Given the available evidence and constraints, what should the agent do next?

The AutoJev route preset uses a bounded policy such as quick execution, deeper review, task decomposition or a blocked state. Your workflow can then map that answer to its own deterministic action.

Example task-routing input

json

{
  "task": "Resolve a request involving account access and a disputed charge",
  "evidence": [
    "Customer identity is verified",
    "Refund eligibility is unclear"
  ],
  "constraints": ["Do not change account state without confirmation"]
}

An illustrative result shape is:

json

{
  "recommendation": "review_deep",
  "confidence": 0.82,
  "guidance": "Inspect refund eligibility before selecting an action."
}

Actual answers and probabilities depend on the supplied state and configured Jev provider.

Where task routing helps

Decide whether to continue, ask the user or stop.
Route a support issue to the correct workflow.
Choose between quick inspection and deeper review.
Decide whether a large objective should be split into smaller tasks.
Escalate when the available evidence is insufficient.

Keep the answer space small and operational. If the agent needs to invent a plan or write the response, use its generative model after the route is selected.

Run the preset

Use POST /api/v1/decisions/route or open the task-routing example in the Playground. The task-router Skill can teach MCP-capable agents when this decision boundary has been reached.

Related Jev resources

Frequently asked questions

Is task routing the same as model routing?

No. Task routing selects the next workflow path. Model routing selects which approved model should handle a task.

Can I define different route choices?

Use the generic Decisions API when your workflow needs a custom answer space. The preset provides a stable policy for common agent task routing.

Does the route automatically execute the next step?

No. AutoJev returns the decision and guidance. The calling workflow applies it within its existing permissions.