HR · AI assistant for working time questions

Consolidating questions about working time models in HR

A service company with almost 700 employees receives many requests about working time models every day, and HR administration loses time.

  • Services
  • Human resources

Key metric

Processing time per standard request

2514min

Average processing time per HR request

At a glance

Initial problem

Unstructured questions about working time models

Solution type

Agentic workflow for HR requests

Complexity

Expansion

Key metric

Processing time per HR request

User group

HR administrators and team leads

Implementation time

3-6 weeks

Operating model

Cloud-based HR assistance system

Starting point

The company offers a wide range of working time models, from full-time and part-time to fixed-term project contracts. Changes are mostly requested by email or through a self-service portal. HR administration handles 80 to 120 requests a day, around half of which concern working time models, extensions or changes. Information on the rules is spread across works agreements, policies and templates on several drives and in the HR system. Many employees ask similar questions, for example about part-time work during parental leave or switching from a fixed-term to a permanent contract. Each time, HR administrators have to find and check documents and write individual replies.

80-120
HR requests per day
700
employees

The challenge

HR administration is stretched thin because standard requests about working time models take up a lot of time. Each request first has to be understood, matched to the right set of rules and checked against the individual contract data. The relevant information is spread across policies, works agreements and notes, often in different versions. Questions are frequently asked several times because answers are not clear enough or employees describe their situation differently. Depending on complexity, handling a request takes between 15 and 45 minutes, and urgent cases such as expiring deadlines compete with routine questions. At the same time, HR is expected to give reliable, consistent information and comply with legal requirements.

Typical symptoms in the organization

“I spend half the day answering the same questions about part-time work.”

“For every request I have to open several documents just to confirm a simple rule.”

“We only notice late that deadlines for fixed-term contracts or parental leave changes are about to expire.”

“Employees get different answers depending on who happens to be on duty.”

Why this remained unsolved

Until now, the problem was seen as an unavoidable side effect of a complex HR landscape. The requests seemed too small-scale for a dedicated IT project, yet too numerous to ignore. Classic FAQ pages were hardly used because employees prefer to describe their situation in their own words. Rule-based forms would not have covered many special cases and would have been costly to maintain. HR also had no development capacity of its own to build a flexible system that understands free text and takes contract data into account.

Our solution

An agentic workflow supports HR administration in handling requests about working time models. Incoming emails and portal messages are read in automatically, understood and assigned to suitable categories such as a switch to part-time, an extension of a fixed-term contract or a reduction in working hours. The system pulls in relevant contract data from the HR system, checks standard situations against stored policies and works agreements, and drafts structured replies. HR administrators see all information and suggestions in one interface, can make adjustments and approve the reply. Complex cases are flagged so that experienced colleagues can handle them specifically.

Where generative AI comes in

Generative AI understands freely worded requests, identifies the underlying concerns and assigns them to suitable categories. It summarizes the relevant rules from policies and works agreements, drafts replies based on contract data and turns differently worded or incomplete information into a structured HR proposal.

Building blocks
  • Read-only integration with the HR system to retrieve contract data, working time models and fixed-term contract information.
  • Classification of incoming emails and portal messages by concern, urgency and required HR review steps.
  • Generation of draft replies that combine individual contract data with the matching rules from policies and works agreements.
  • Interface for HR administrators to review, adjust and approve replies and to flag complex special cases.
  • Analysis of request volumes, processing times and frequent topics to continuously improve policies and templates.
How it works

How the HR workflow works

  1. Collect requests

    Incoming emails and portal messages are captured automatically, matched to the employee and created as an HR request in the workflow.

  2. Understand the concern

    Generative analysis identifies the concern, assigns it to categories such as a switch to part-time or a fixed-term extension, and checks urgency based on deadlines.

  3. Link the rules

    The workflow pulls relevant contract data from the HR system and links it to the matching rules from policies and works agreements.

  4. Draft a reply

    Based on the identified situation, the system generates a structured draft reply with clear options and guidance for the employee.

  5. Approval and refinement

    HR reviews the draft, makes adjustments, approves the reply and flags special cases that should be better supported by automation in future.

Impact

Measured: after 3 months in production

Period: after 3 months in production.

MetricValueBasis
Processing time per standard requestKey metric8-15 minutes instead of 20-30 minutesTime tracking before/aftermeasured
Share of automatically prepared replies65-75 percent of requestsSystem measurementmeasured
Corrections needed after sendingQuality metric5-8 percent need follow-upSampleestimated

Cost-effectiveness

Effort class

Compact (25-60 person-days)

Payback period

6-12 months

Main lever

Time savings

Ongoing costs

Ongoing costs arise from hosting, maintenance, model usage and occasional adjustments to the rule sets.

Limitation

The solution mainly covers standard situations. Complex special cases under employment law still have to be reviewed by experienced HR professionals. Changes to works agreements require careful maintenance of the stored content.

Integration and operation

Connected systems

The workflow is connected to the central HR system and the email gateway, reads contract data and requests, and writes status information back.

Operation

The system runs in a European cloud environment with separate tenants, encrypted data transfer and role-based access.

Compliance

Personal data is processed in line with applicable data protection rules, access is logged, and retention periods follow internal policies.

Labeling

Replies remain labeled as HR communication; the system does not act as a contact in its own right and refers to HR when in doubt.

Accountability

Automatically generated draft replies only become binding after review by HR; critical decisions on contract changes remain with the department.

Escalation path

Cases with an unclear legal situation or conflicting data are flagged automatically and forwarded to designated HR experts for clarification.

Traceability

For every request, the data sources used, the rules applied and any manual changes are logged so that decisions can be traced and reviewed later.

Operational roles

HR administrators handle suggestions, team leads monitor metrics, IT is responsible for technical support and the HR specialist team maintains the rule sets.

Also works for

Consolidating questions about study regulations and exam deadlines in university administrations.

Structuring requests about shift models and allowances in municipal utilities.

Pre-sorting standard requests about duty rosters and on-call shifts in hospitals.

A similar case in your organization? Let's talk.

Briefly describe your process. We'll get back to you with an initial assessment of whether the pattern can be transferred and what it would realistically take.

Apply to izz.ai

PDF, DOC or DOCX · up to 3 files, max. 10 MB each

I accept the privacy policy

Get in touch

A friendly pink robot opens a door behind which lies a futuristic world