Home/Blog/Greenhouse Offer Approvals: Setup and AI Review
Recruiting operations

Greenhouse offer approvals: setup, changes and reviewed AI handoffs

Configure Greenhouse offer approvals, check job-specific coverage, handle new offer versions and add AI preparation without bypassing human review.

Remqa · November 21, 2026 · 5 min read

Short answer

Greenhouse offer approvals control when an offer can be extended to a candidate. Configure the approval flow, confirm it applies to the specific job, then request approval for the current offer version. A custom AI workflow can prepare a review brief, but it should not treat an old approval or a draft summary as authorization to send.

  • Check the specific job because default approval settings may not cover existing jobs
  • Pending remains until the configured approval conditions are complete
  • Changing a refresh-marked field creates a new offer version and affects existing approvals
  • Keep custom AI preparation separate from offer approval and candidate delivery

A recruiter has an approved offer, then the package changes. The team updates the candidate record, copies the old approval into a handoff message and assumes the offer is ready to send. That assumption needs checking. In Greenhouse, job approval, offer approval, offer versions and document delivery are separate concerns. A useful AI workflow makes those boundaries easier to inspect. It does not compress them into a single green status.

Start by agreeing what the workflow may do. Preparing a compensation brief is different from changing the offer. Requesting approval is different from approving it. Sending the candidate a document is another action. Keep those responsibilities explicit before adding AI recruiting agents to the process.

Check the job before trusting the organization default

Greenhouse distinguishes jobs with an offer approval process from jobs without one. Where the flow is enabled, newly created offers need approval before they can be extended. Without an offer approval process, appropriately permitted users can generate and send offer documents and mark candidates as hired from the Offer stage.

A default configuration is not proof that an older job is covered. Greenhouse notes that a job may have no offer approval process because the default was never configured or was not retroactively applied. For each rollout job, inspect its actual approval configuration. Record whether approval is required and who owns any gap. Do not present a missing flow as an approved offer.

Configure the approval steps and their scope

The configuration path is Configure, then Approvals. Creating or overriding default approval flows requires the corresponding Site Admin permission. Defaults can apply to all new jobs. Approval settings can also target an office, a department or a department within an office. That last choice is an intersection, so an office and department combination should not be described as covering either group independently.

Select the offer approvals section rather than the job approvals section. Add the approvers and decide how many approvals each step needs. Greenhouse allows individual users or a custom job field to identify an offer approver. Review the resulting people on a real job instead of assuming a field label resolves to the intended person.

DecisionNative behavior to checkRecommended operating check
ScopeDefault or office and department configurationInspect the specific job
Required approvalsConfigured number within each stepRecord the rule, not just the approver list
OrderAll at once or in orderMake the current blocking step visible
Offer changesRefresh-marked fields create a versionCompare the reviewed terms with current terms
DeliverySeparate action after required approvalVerify the document and intended recipient

Choose parallel or ordered requests deliberately

With all-at-once approval requests, Greenhouse sends requests to approvers across the steps at the same time. With in-order requests, each step must meet its approval requirements before the process progresses. Choose based on the organization’s decision process. If finance must settle the package before another reviewer acts, the order should express that dependency.

For a custom assistant, summarize the current step and outstanding conditions. Do not reduce a multi-step flow to a count of people who have clicked approve. The configured threshold and ordering matter. Treat an unavailable or ambiguous approval state as an exception for the recruiter to inspect, not a reason to infer completion.

Request approval for the current offer

After an offer has been created for a candidate in the Offer stage, Request Approval starts the process. Greenhouse keeps the overall status Pending until every condition of the job’s offer approval process is complete. Approval Details provides more information about progress. Once approved, Greenhouse displays the date of approval.

Offer management and approval actions require the relevant permissions. The documented Job Admin permission includes viewing and editing offers and approving or requesting approval on offers. The Site Admin route also includes private notes, salary information and offer management permissions. Avoid giving a broad service account those rights just because the assistant needs to prepare a short brief.

Bind the review brief to the terms it describes. As a custom engineering control, store the candidate, job, current offer version when available, reviewed terms and preparation time together. Invalidate the brief if those terms change before the next action.

Handle new offer versions without recycling approval

Greenhouse marks offer fields that create a new version with a Refresh icon. To edit an existing offer, open the Offer stage, choose Update, change the fields and save. Changing a marked field creates a new offer version. Editing an unmarked field does not create a version or re-trigger the approval process.

Creating a new version and re-triggering approval voids the original offer and its associated approvals. Request Approval is the documented action for requesting approval for the new version. Configured offer approvers are notified by email. Do not assume that saving a changed package proves that the replacement request has been sent or approved.

Also avoid the opposite assumption. An unmarked field can change without native reapproval. Your internal policy may still require a person to review a material document change. Keep that policy distinct from the software’s version trigger. Record both what Greenhouse requires and what the organization chooses to require.

Add AI preparation around the native approval flow

A custom assistant can assemble a proposed review brief from authorized information. Include the proposed terms, the approved hiring context, unresolved questions and a comparison with the last reviewed package. Label missing evidence plainly. The assistant should not invent compensation ranges, candidate commitments or approval reasons from similar offers.

Limit the audience of the brief. Salary details and private notes should not spill into a broad channel or an unrelated candidate record. Use the viewer’s permitted scope when selecting evidence. This is a recommended application control, not a claim that a generated summary automatically inherits every Greenhouse permission.

  1. Read the current candidate, job and offer context.
  2. Prepare the brief with links or references inside the authorized internal workspace.
  3. Ask the responsible person to resolve discrepancies.
  4. Recheck the current offer before a permitted approval request.
  5. Confirm the resulting state separately before any delivery handoff.

Test changes and uncertain actions

Test a job with no flow, a pending offer, an ordered approval and a changed offer version. Include a change to an unmarked field so the team sees that not every edit restarts approval. Compare the assistant’s description with the native record. The existing Greenhouse job approval workflow covers a different decision and should not be substituted for this offer check.

If a custom integration loses its response during a write, inspect the current record before retrying. Keep an action journal with the intended change and observed result. A timeout is not evidence that nothing happened. Never let an uncertain approval request become an automatic candidate email.

How Remqa helps with offer workflow handoffs

Remqa builds recruiting AI workflows around the systems and approval owners a team already uses. The work can include evidence preparation, permission-aware review queues, change detection and reconciliation after uncertain actions. Start with one offer workflow and its exception cases. Use a recruiting workflow review to identify the handoff that needs engineering before expanding the assistant’s authority.

FAQ

Questions people ask.

How do you request offer approval in Greenhouse?

After creating an offer for a candidate in the Offer stage, select Request Approval. The overall status stays Pending until all conditions in the job’s offer approval process are complete.

Does a default Greenhouse offer approval flow cover every existing job?

Not necessarily. Greenhouse notes that a job can have no offer approval process when a default was not configured or was not retroactively applied. Inspect the actual job before relying on the organization default.

Which offer edits trigger a new version?

Greenhouse marks fields that create a new offer version with a Refresh icon. Editing unmarked fields does not create a new version or re-trigger offer approval. Check the configured fields instead of assuming every edit behaves the same way.

Can offer approvals happen in parallel?

The approval configuration supports requesting approvals all at once or in order. In-order steps must satisfy their requirements before progressing. Each step can also have its own required number of approvals.

Can AI approve or send Greenhouse offers?

This guide recommends custom AI only for preparing evidence and checking proposed changes. Keep authorized people responsible for approval, and verify the current offer version and permissions before any separate delivery action.

Talk to an engineer

Want an engineer to look at your setup?

Book a free 30-minute call. We look at your tools and your team's week and tell you what we would do first.

Book a free 30-minute call
30 minutesFree, with an engineer
Your stackWe look at what you run today
Straight answerIncluding when we're not the fit