A hiring manager signs off on a planning document. A recruiter starts preparing outreach. Then the team discovers that the Greenhouse job is still waiting for approval. The document and the ATS were describing different states. Adding AI to the planning process can make that gap harder to see if a polished brief is mistaken for authorization.
Start with a smaller goal. Let an assistant organize the hiring request, identify missing information and explain changes. Keep the actual approval with the people and workflow configured in Greenhouse. The preparation controls below are recommended application design, not a description of built-in Greenhouse AI functionality.
Know which approval the team is waiting for
Greenhouse supports one-stage and two-stage job approval processes. With an approval flow configured, a new job begins in Draft and needs the required approvals before recruiting can start. The job creator or another appropriately permitted person can request the initial approval from the Draft job.
In a one-stage process, the final required approval opens the job for candidate recruitment. In a two-stage process, the initial approval opens it as Open (Approval Pending). Recruiting can begin, but a second job approval process is required before generating an offer for a candidate. Do not treat this as a substitute for checking any separate offer approval requirements your organization uses.
| Checkpoint | What to verify | AI preparation task |
|---|---|---|
| Initial request | Job details and authorized requester | Find missing information and conflicting instructions |
| Recruiting approval | Required approvals received in Greenhouse | Prepare a short review brief without asserting approval |
| Second stage, if configured | Current permission to generate an offer | Highlight changes since the previous review |
| After a material edit | Current job state and any renewed approval request | Explain the edit and retire the old preparation draft |
Build a hiring request packet before adding automation
Define what a reviewer needs to decide. A useful preparation packet includes the intended role, department, employment type, requested openings, the business reason for hiring and the person responsible for the request. Include the relevant job identifier so a reviewer can distinguish similar roles. Add private fields only when the review requires them.
Use AI to turn scattered manager notes into this structure. Require it to label missing facts instead of filling them in. If one note requests a replacement and another requests an additional opening, show the conflict. The assistant should not decide that both statements mean the same thing or invent a headcount decision.
The AI job description workflow covers the separate task of turning an agreed hiring brief into role copy. Keep that writing step downstream of the headcount decision. A better job description does not resolve an unclear approval request.
Keep the preparation state separate from the ATS state
Give the custom preparation tool its own simple states, such as needs information, ready for review and superseded. These are application labels. They are not Greenhouse job statuses. Show the current ATS state separately, with the time it was checked and the job it belongs to.
Record which version of the request a reviewer saw. Include the fields being proposed and the source notes used to create them. If the department, openings or employment type changes after review, mark the packet for another check before submission. Do not reuse an old approval just because the role title stayed the same.
Check permissions without spreading private information
Requesting approval and configuring approval flows are permissioned actions. Greenhouse documents specific Site Admin and Job Admin permissions for requesting job approval. Configuration of the default flow requires a Site Admin who can create and override default approval flows. Confirm the actual user's permissions before designing a handoff.
The preparation tool should not solve a permissions problem by granting broad access or sending the entire packet to a larger group. Keep the role brief, salary information and private job fields behind the appropriate audience checks. A notification can say that a request needs review without reproducing every private field.
When building AI recruiting agents, define the readers as carefully as the writers. A tool that can read restricted job information can still disclose it through a summary, export or notification. Keep review access attached to the underlying request rather than assuming anyone with the summary link may read it.
Handle changed job details explicitly
Greenhouse identifies Requisition ID, Openings, Opening IDs, Department and Employment type as fields whose changes can re-trigger approval notifications in the documented pending states. In the one-stage pending flow, the original approval notification is voided and a new notification lets the approver act on the changed request.
For a two-stage job that has received full approval, changes to those listed fields re-trigger the second stage only. Re-triggering the first stage in that situation requires setting the job to Draft, changing its details and submitting it again. Do not describe every edit as restarting the entire workflow.
After a one-stage job has been approved and opened, the one-stage process does not automatically require re-approval for later changes. If your policy needs another review, implement and document that additional control rather than claiming Greenhouse supplies it through the completed one-stage flow.
Test the handoff before using it for live hiring
- Prepare a request with a missing department and verify that the assistant asks for a fact rather than inventing one.
- Review a packet, then change its requested openings. Confirm the old packet is no longer treated as current.
- Use a requester without the necessary permission and confirm the tool reports a blocked handoff.
- Test both recruiting approval and the second-stage checkpoint so the messages do not confuse them.
- Simulate an interrupted submission. Inspect the current job before retrying rather than creating another request blindly.
Also test the ordinary case with an authorized recruiter. They should be able to see what changed, what is missing and where the decision must happen without reading a technical log. Keep the evidence brief enough to use during a normal hiring review.
How Remqa approaches job approval preparation
Remqa can build the custom preparation and review layer around your recruiting process. The scope can include structured manager intake, evidence checks, restricted review packets and a clear handoff to the ATS. The configured approvers retain the decision.
Start by walking through one recent hiring request and the places it stalled. Our recruiting workflow map can help separate preparation work from approval work. Bring the actual approval setup to a call so an engineer can discuss a bounded first workflow.