A leave request workflow: approve, decline, or send it back for different dates

SynaptaGrid8 min read

#workflow#hr

In short

A leave request workflow is the path a request for time off takes from the employee to a recorded decision: the manager checks that the dates work for the team, HR checks that the employee has enough leave left, and the request ends approved, declined or sent back for different dates. Most leave problems come from a missing third option and from unclear ownership. A manager who cannot say not those dates but another week has to choose between approving and declining, and nobody can tell afterwards whether the balance was ever checked. A good workflow gives the manager three choices, gives HR its own check, and writes the status onto a record. We built a small version in our own test workspace in which an AI tool, written for this job and used as an action in a normal step, reads the request text, names the type of leave and writes a short note on what was asked, and the workflow writes both onto the leave record before the manager reviews it. We ran it three times, once approved, once sent back and once declined for not enough leave, and checked the records afterwards.

What a leave request workflow is

A leave request asks for permission to be away on certain dates. A workflow writes down who looks at the request, in what order, and what the possible answers are. Without one, requests travel by chat and email, the balance is checked by whoever remembers, and the answer is lost in a thread.

The workflow does not decide who gets leave. The manager and HR do. It makes sure both of them look, that a request has more than two possible answers, and that every request ends in a state the employee and HR can find.

The version we built

An AI tool, used as an action in a normal step, reads the request and the workflow writes the leave type and a note onto the record. The manager can then approve, decline or send the request back. HR then checks the balance of an approved request. Automatic steps create the request record and update its status at each ending.

The steps to include, and who owns each

Record the request
When a request arrives, a record is created with the employee, the leave requested and a status of submitted. Every later step reads from and writes to that record.
Read the request
This is where an AI tool helps. It reads what the employee wrote, names the type of leave, such as annual or sick, and writes a short note of what was asked and anything the text leaves out. The next step writes both onto the leave record, so the manager sees them with the request. The tool does not approve or decline anything.
Manager review
The manager checks the dates against the team’s workload and cover. The manager knows what else is happening in the team, which HR cannot see.
A third answer
Besides approve and decline, the manager can send the request back for different dates. That ends this request with a clear status, and the employee submits a new one. Without this option managers decline requests they would happily approve for another week.
HR balance check
HR checks that the employee has enough leave left and that the leave type allows the dates. This is a different question from the manager’s, so it is a different step with a different owner.
Decline from either step
A request can be declined by the manager or by HR for a low balance. Both lead to the same ending: the record is marked declined.
Mark the result on the record
Each ending sets the status: approved with the balance updated, declined, or returned. The workflow does this itself.
Update the calendar and payroll
Showing the leave on a team calendar and passing it to payroll usually happen in other systems. Decide where an approved request goes next and add that as its own step.

Where approvals and checks belong

Put each check with the person who has the information. The manager has the team context, HR has the balance and the leave rules. Splitting them avoids the manager guessing at the balance or HR deciding on workload.

Name each approver by role, such as the employee’s manager and the HR team, rather than by person. If you need an extra approval for long absences, add it on your own policy; we did not build that.

The old way and the new way

The table compares leave requests handled by chat and email with the same requests run as a workflow.

First read of a request
The old way: The manager reads it cold
The new way: An AI tool names the leave type and notes what is missing first
Where the request is
The old way: In a chat or an inbox
The new way: At a named step, with an owner
Answers available
The old way: Yes or no, sometimes a long reply
The new way: Approve, decline or send back for different dates
Whether the balance was checked
The old way: Nobody can tell
The new way: A separate HR step that is recorded
The status of a request
The old way: Ask the manager or HR
The new way: Open the record and read its status
Finding past requests
The old way: Search old messages
The new way: List the records
Judgement
The old way: Still the manager’s and HR’s
The new way: Still theirs: whether the dates work, whether the leave is allowed

How we decided

We built the version in the diagram in our own test workspace in SynaptaGrid Automation, with a small record type in SynaptaGrid Data for the request, and ran it three times in November. The AI part is a custom tool we wrote for the job: a prompt that says how to review a leave request, an input that takes the request text, and an output with the type of leave from a fixed list and a short note. In the workflow it is an action in an ordinary step, set up like any action that calls another system, and the next step writes its result onto the record. In the first run the manager approved and HR confirmed there was enough leave. The workflow created the request record with the status Submitted, then updated it to Approved with the balance-updated flag set, and the run finished with the approved result. In the second run the manager sent the request back for different dates: the run skipped HR, marked the record Returned for different dates and ended. In the third run the manager approved and HR found not enough leave, so the record ended as Declined.

We checked the three records in Data afterwards. The approved one showed Approved with the flag set. The returned one showed Returned for different dates, and the declined one showed Declined, both with the flag not set. All three carried the AI check: the leave type annual and a note restating the request, a family wedding with the handover agreed. The AI tool did not work first time: our earlier attempts to call it failed until we named the model in full in the tool’s settings, and our first version did not use its output in a later step. One run also failed at the AI call with an error we did not investigate; a fresh run with the same input passed. The version described here is the one whose three outcomes we checked.

One path was not run on its own. A decline by the manager leads to the same declined step that HR’s decline uses, and we ran that step through HR’s decline only. We chose a third answer for the manager and a separate HR check because those are the places leave handling usually goes wrong. The choice of steps is ours and a judgement; it is not a standard.

Six limits, so this guide does not claim more than the runs showed. The tool saw one request sentence, so we did not test other leave types, unclear dates or a request that conflicts with a leave policy, and the note only restates the request. One test account held every human step, so the runs show the sequence and the branching, not hand-offs between different people. HR’s check was a person’s decision: the workflow does not read a real leave balance, and the balance-updated flag is a stand-in, since we did not connect an HR, calendar or payroll system. We did not build an extra approval for long absences, a limit on overlapping requests in a team, or a way to link a returned request to its replacement. We did not test reminders when a request waits too long or notifications to the employee. And the template carries one test account as every approver and the actions tied to our test record type, so to reuse it you would change each step to a role and point the actions and the tool at your own, which we did not do.

Standards and accepted practice behind this

These are the practices the guide rests on. We align with them. This page does not claim a certification.

Documented decisions
Employers are often expected to be able to show who decided on a leave request and when, particularly if a request is disputed. A recorded step and a record that carries the status provide that.
Statutory leave rules
Minimum leave and how it may be refused are set by the law of your country and sometimes by contract. Check yours before you decide what a manager may decline.
Human oversight of AI
Where AI helps with a decision, accepted practice is to keep a person accountable for it and to let them see the AI’s input and output. Here the tool names the leave type and drafts a note, and the manager and HR decide.
Source: EU AI Act (Regulation 2024/1689), Article 14: human oversight (opens in a new tab)
Segregation of duties
Separating the manager’s judgement from HR’s balance check means one person does not both approve the request and decide whether the entitlement exists.
Source: NIST glossary: separation of duty (opens in a new tab)
BPMN
The standard notation for drawing a process. The diagram above is a plain flowchart, and a BPMN diagram of the same steps would show the same decisions.
Source: OMG: BPMN 2.0 specification (opens in a new tab)

Where SynaptaGrid fits

SynaptaGrid Automation lets you write the steps as a workflow, assign each decision to a person, use an AI tool you wrote as an action in a step, pass its result to the next step, create and update records in SynaptaGrid Data as part of the run, and see where every request stands. You can start from the version above, change the steps to match your own leave policy, and keep the result as a template.

Check the limits above before you rely on it: whether each step is assigned to a role rather than a person, whether the balance check needs to read your HR data, whether approved leave must reach a calendar or payroll system, and who should be told when a request waits.

Frequently asked questions

What steps should a leave request workflow have?
Record the request, a manager review with a way to send it back, an HR balance check, a way to decline from either step, and a status written on the record.
Why let a manager send a leave request back?
Because the choice is often not yes or no but different dates. A third answer ends the request cleanly and avoids a decline for something the manager would approve another week.
Who should check the leave balance?
HR or whoever holds the entitlement records, as a step separate from the manager’s review.
Can AI read a leave request?
It can read the request text, name the type of leave and note what was asked. In our test run it named the leave as annual and restated the dates and the reason. The manager and HR still decide.
Can a workflow update the leave record for me?
Yes. In our test run the workflow created the request record when the request arrived and set its status at each ending. You need a record type for the request and each action pointed at it.
Share

SynaptaGrid

Build your leave request steps

Start with the steps above in a test workspace and run one request through each outcome before you use it for real.