Probation Management Agent

Three out of ten new hires leave the company during the probation period. In most cases, not because the selection was wrong - but because for six months no one was watching. No structured feedback, no documented milestones, no deliberate decision at the end. The probation window passes, and with it the only phase in which separation with two weeks of notice and without a written reason is possible.

The 183-day window

This agent follows the Decision Layer principle: each decision is either rule-based, AI-assisted, or explicitly assigned to a human.

The probation period is not a soft get-to-know-you programme. It is a precisely timed decision window under employment law. Across most European jurisdictions, notice periods during probation are reduced to two weeks. From day 183 in many systems, full employment protection kicks in - longer notice, social criteria for selection, reasoned justification. Missing the transition is not paid for in weeks but in months and severance.

The time architecture looks like this:

Day 1         Day 30        Day 90        Day 150       Day 180
|-------------|-------------|-------------|-------------|
Onboarding    1st feedback  2nd feedback  Decision      Period
              due           due           due           ends

Between day 150 and day 180 sits a 30-day window in which the continued-employment decision has to happen. Not should happen - has to. A termination on day 179 with two weeks of notice ends on day 193, after the probation period. Too late.

Why HR loses the overview

A company of 2,000 employees and an 8 percent annual hiring rate has around 13 active probation periods at any time. Over a year, that is more than 150 individual deadlines, feedback dates, and decision points. No shared service centre tracks that reliably in spreadsheets.

The real problem is the information asymmetry. HR knows the deadlines but not the performance. The manager knows the performance but forgets the deadlines. Between them sits a communication gap that only becomes visible when it is too late: on day 178, when HR asks whether employee X is being confirmed, and the manager thinks about it for the first time.

Studies on early attrition show the cost of a failed hire ranges between 33 and 213 percent of annual salary, depending on seniority. A substantial share of this cost does not come from the separation itself, but from the delayed separation: extended ramp-up without prospects, productivity losses in the team, employment-law effort after the probation period has expired.

What the agent changes

The Probation Management Agent solves no motivation problem and no culture-fit issue. It solves a steering problem. Its contribution is structural: it ensures that at every point of the probation period, the right information sits with the right person and the right decision is demanded on time.

Concretely: as soon as an employment contract is recorded in the HR system, the agent calculates the milestones from the start date. It differentiates by contract type and probation duration. By default, it sets checkpoints at 30, 60, 90, and 150 days. The first wave is purely informative - a reminder to the manager to hold the first feedback conversation. From the 90-day checkpoint onwards, the tone turns serious: is documented feedback on record? Are warning signals present - absences, missed training completions, open onboarding tasks?

The critical moment is day 150. Here the agent forces a decision. Not as a recommendation, but as an escalation: if no documented assessment exists by then, parallel notifications go to the manager and HR. The reason is arithmetic: 30 days remain, two weeks of notice must be observed, and in co-determination jurisdictions the works council hearing needs a week - leaving no room for delay.

The difference between a reminder and a process

Calendar reminders already exist. They fail on three points: they have no escalation chain, they do not document, and they do not connect manager and HR in a shared decision process.

The agent, by contrast, builds a documented decision point at every milestone. Feedback is not only scheduled but tracked. The manager rates performance, integration, and development potential in a structured format. HR sees in real time which probation periods need escalation and which are running smoothly. At the end of the probation period, a complete decision trail exists - not as a compliance exercise, but as legally defensible documentation.

For separations: dismissal during probation does not require a reason but does require worker-representative hearing where it applies. Without it, the termination is invalid. The agent ensures the hearing is initiated in time and the required documents are in place.

For confirmations: the agent formally closes the onboarding process and hands over to the performance review cycle. No open milestones, no catching up on missed conversations in month seven.

When this pattern works strongest

Probation management is not a complex AI problem. It is a timing problem with high consequences. Exactly that makes it the ideal entry point for rule-based automation: clear deadlines, defined escalation levels, human decision at the end. The agent takes over the scaffolding - the manager makes the decision. Once this pattern is implemented, it recognises itself across the organisation: in certification monitoring, in the performance review cycle, in every time-driven HR decision. (US: at-will employment changes the legal framing, but the operational challenge of structured 30/60/90-day check-ins is identical; the agent’s milestone tracking applies without change.)