When a System Auto-Triggers a Performance Improvement Plan

A business rule can now open a performance improvement plan on its own. The law still expects a person to have decided to open it.

In its first half 2026 release, SAP SuccessFactors added a set of business rules to its performance forms that automate decisions managers used to make by hand. One of them can launch a performance improvement plan form automatically when an employee's overall rating lands on Below Expectations. SAP describes the new rules as reducing manual oversight and building fairness, accountability and consistency into the process. Read the same feature the way an Australian employment lawyer would, and it removes the one thing the Fair Work Act cares about most in a performance dispute: a human being who decided, and who can say why.

This is not an argument against the software. It is an argument for configuring it deliberately. The safe rule is short. A system can flag. Only a person can trigger.

What SAP's 1H 2026 release actually changed

According to SAP's release notes and the partner summaries that track them, the first half 2026 (1H 2026) release reached production in May 2026, after a preview period from April. Among its Performance and Goals changes is a new group of business rule scenarios that, in SAP's own framing, automate key decision logic and reduce the manual steps in running a performance form.

One of those rules is the one that matters here. As reported across SAP's release documentation and independent partner write-ups, administrators can now configure a performance form to launch a performance improvement plan form automatically when an employee's overall performance rating is recorded as Below Expectations, with the option to show a custom message in the confirmation dialog. A companion rule can launch an individual development plan when the overall competency rating is low. SAP positions the change as a way to apply the same standard to every employee, reduce bias and free managers to spend less time on administration and more on coaching.

The appeal is obvious. Consistency and less paperwork are real benefits, and a rating of Below Expectations is a reasonable thing to want the system to notice. The shift that matters is quieter. The decision to start a formal performance process moves from a manager's deliberate act to a condition evaluated by a rule engine. Whether that rule notifies a human or opens the form itself is a configuration choice, and it is the choice this article is about. Agent and automation building is now moving inside the core HR system rather than sitting with IT, a shift covered in our piece on how your HRIS now lets anyone build an agent, and the same governance questions apply to a business rule that can act on an employee.

Why an auto-launched PIP is a Fair Work problem

A performance improvement plan is not a neutral administrative artefact. Under section 342 of the Fair Work Act 2009, adverse action by an employer includes injuring an employee in their employment and altering the position of the employee to the employee's prejudice. A PIP, depending on how it is run and what it leads to, can do both.

That connects the PIP to the general protections. Section 340 prohibits an employer from taking adverse action against an employee because the employee has, exercises, or proposes to exercise a workplace right. Section 341 defines a workplace right broadly, including an entitlement under a workplace law, a role or responsibility under one, and the ability to make a complaint or inquiry in relation to employment. Australian courts have accepted that placing an employee on a performance improvement plan can amount to adverse action, and can found a general protections claim even where the employer had a legitimate performance concern. In Pezzimenti v Rotary International, for example, the implementation of a performance improvement plan was treated as adverse action on the facts.

The practical consequence is that the reason for a PIP, and its timing, carry legal weight. If a plan lands on an employee shortly after they raised a complaint, took protected leave or disclosed a health condition, the causal question arrives whether or not anyone intended a connection. The Fair Work Commission's general protections jurisdiction exists to test exactly that link.

Who is the decision-maker when a rule pulls the trigger?

This is where an automatically launched PIP becomes dangerous, and it is a point of evidence rather than intention.

Section 361 of the Fair Work Act reverses the ordinary onus of proof. Once an employee establishes that they had a workplace right and that adverse action was taken, and alleges the action was taken for a prohibited reason, it is presumed that the action was taken for that reason unless the employer proves otherwise. The employer has to disprove the prohibited reason on the balance of probabilities.

How does an employer discharge that? By leading direct evidence from the decision-maker about the actual reasons for the decision. In Board of Bendigo Regional Institute of Technical and Further Education v Barclay, the High Court put the decision-maker's real reason, their genuine state of mind, at the centre of the inquiry. Practitioners draw the obvious lesson: it is extremely difficult to displace the section 361 presumption if no direct testimony is given by the person who actually made the decision.

Now ask the question the feature forces. If a business rule launched the PIP the instant a rating was saved, who is the decision-maker? No person turned their mind to this employee and decided to start a formal process. A threshold was crossed and a form opened. You cannot call a rule engine to the witness box to explain its reasons, and the person who configured the rule months earlier decided about a category, not about this individual. The employer is left trying to rebut a presumption with no decision-maker evidence to lead, which is the weakest position an adverse action respondent can be in.

Said plainly, automating the trigger manufactures an evidentiary hole in precisely the place the law puts the burden on the employer. The phrase the system did it automatically is not a defence. It is close to an admission that nobody weighed the individual circumstances at all.

Does an auto-launched PIP still deliver procedural fairness?

Procedural fairness is the second exposure, and it runs through the unfair dismissal jurisdiction rather than the general protections one. Section 387 directs the Commission, when deciding whether a dismissal was harsh, unjust or unreasonable, to consider whether there was a valid reason related to the employee's capacity or conduct, whether the employee was notified of that reason, and whether they were given a real opportunity to respond. A well run PIP is often the mechanism that delivers those things before any capability dismissal, which is why the discipline of AI-drafted performance documents and procedural fairness matters so much.

But procedural fairness is about genuine consideration, not about a form appearing on time. A valid reason has to be sound, defensible and well founded on facts a person has actually verified. A rating crossing a threshold is not a verified reason. If the rating itself was shaped by calibration, a forced distribution or an AI-assisted scoring step, and the formal process then launched itself off that number, the chain from evidence to adverse action can run end to end with no human judgement in it. A process that looks automated is a process an applicant will characterise as run for appearance, and the Commission is alert to the difference.

The rule of thumb: a system can flag, only a person can trigger

The line to draw is not between using the system and not using it. It is between flagging and triggering.

Systems are genuinely good at surfacing patterns a manager might miss: a rating that has slipped two cycles running, a mismatch between strong goal results and weak competency scores, a run of missed check-ins. Surfacing those is useful and entirely legitimate. Let the system flag.

The trigger is different. Starting a formal process that can alter someone's position is a deliberate act, and it should be performed by a named manager who has looked at the individual's context, can articulate the reason in their own words, and could give evidence of it later if asked. Between the flag and the trigger sits a mandatory human checkpoint. In configuration terms, that means the rule notifies a manager and HR that the threshold was crossed, and stops there. The manager decides whether to open a plan, and records why.

That is the same human-in-the-loop principle the site applies to AI worker monitoring and automated decisions, and it is the direction Australian regulation is travelling on decisions that significantly affect people. Keeping a person accountable for the decision is good practice today and a safe bet for what is coming.

How do you configure this safely before enabling it?

Treat enabling the rule as a governance decision, not a switch a system administrator flips. Work through this the Monday before anyone turns it on.

1. Inventory every automated rule in your HRIS that can change an employee's status, launch a form, or start a process: PIP launches, development plan launches, disciplinary routing, and any status flags feeding them.
2. Disable any rule that launches a performance improvement plan, development plan or disciplinary form directly off a rating, and leave it off until step 4 is done.
3. Reconfigure the rule to notify only. It should alert the responsible manager and an HR contact that a rating threshold was crossed, and take no action on the record itself.
4. Insert a named human decision step. A specific manager decides whether to open a plan, and records a short rationale that names the performance facts, not the rating number.
5. Check whether enabling workforce-wide performance automation is a change that triggers a consultation obligation under an award or agreement, using the test in our guide on when an AI rollout triggers your consultation clause.
6. Brief managers in writing that a system flag is not a decision, and that a rating crossing a threshold does not oblige them to start a formal process.
7. Turn on logging so that for every plan you can show who decided, when, and on what stated reason, and keep that record with the plan.

To prepare the ground work the system used to skip, without letting the system make the call, give a manager a structured brief rather than a launched form. This prompt keeps the analysis de-identified and stops short of any decision:

You are helping a manager prepare to consider whether a formal performance
conversation is warranted. Do not decide, recommend, or draft a plan.

Using only the notes I paste below, produce a neutral chronology for
[EMPLOYEE_ROLE] in [TEAM]:
- dated performance observations, each tied to the specific expectation it relates to
- any check-ins, feedback or prior conversations on the same issue, with dates
- gaps in the record where evidence is thin or missing
- any recent events unrelated to performance (a complaint, leave, a disclosure)
  that the manager should be aware of before acting, flagged neutrally

Do not include the employee's name. Do not score, rank or conclude. End with
the sentence: The decision to open a plan, and the reasons for it, are the
manager's to make.

Notes: [PASTE DE-IDENTIFIED NOTES]

Second, audit the automation itself. Run this against your HRIS configuration export or with your system administrator:

Act as an HR systems governance reviewer. From the configuration details I
provide, list every automated rule that can:
- launch a form (PIP, development plan, disciplinary, review)
- change an employee's status or record
- route a case to a manager or HR

For each rule, return: the trigger condition, what action it takes, whether a
human approves before the action lands, and whether the action is logged with a
named decision maker. Flag any rule that takes an action affecting an employee
with no human approval step as [HUMAN GATE MISSING].

Configuration details: [PASTE CONFIG EXPORT]

Before you enable anything, confirm you can answer yes to each of these:

• The rule notifies a person; it does not launch the plan itself.
• A named manager makes and records the decision to open any plan.
• The recorded reason names verified performance facts, not a rating score.
• Managers have been told, in writing, that a flag is not a decision.
• You have checked whether enabling the automation triggers consultation.
• Every launched plan has an audit trail of who decided and why.
• No plan can be opened, escalated or closed with no human in the loop.

A worked example

Assume [EMPLOYEE_NAME], a [ROLE] in [TEAM], has an overall rating saved as Below Expectations at the mid-year check-in. Two weeks earlier they lodged a bullying complaint against their team leader, which is still being investigated.

With the rule set to launch, the PIP form opens automatically the moment the rating is saved. There is no manager decision, no consideration of the open complaint, and no record of anyone weighing whether a formal process is appropriate right now. If the employee brings a general protections claim, the employer must disprove that the complaint was a reason for the plan, and has no decision-maker to put in the box, because there was no decision, only a threshold. That is a claim an employer should expect to lose or settle.

With the rule set to notify, the manager and HR receive a flag. HR sees the live complaint, recognises the timing risk, and a named manager decides to hold the formal process, continue informal coaching, and document that the complaint, not any reluctance to manage performance, is the reason for waiting. The performance concern is still addressed, on a defensible timetable, by a person who can explain the call. Same rating, same software, entirely different legal position.

Bottom line

SAP SuccessFactors can now launch a performance improvement plan by itself when a rating drops, and other HR systems will follow. The feature is not the problem; letting it make the decision is. A PIP can be adverse action, the reason and timing of one carry legal weight, and the Fair Work Act makes the employer prove the decision-maker's reasons through direct evidence. An automatically launched plan leaves no decision-maker to give that evidence and no genuine consideration to point to. Keep the flag if it helps managers see what is happening. Move the trigger back to a person who decides, records why, and could stand behind it later.