Blog

AI in HR? Without a clear data foundation, digitalisation becomes a risk

Published in: personal manager – the Austrian trade journal for human resources, Summer 2026 issue (print and online).

Digitalisation and artificial intelligence promise more efficient HR processes. But their value stands or falls with the quality of the data and how well systems work together. Incomplete information, unclear responsibilities and poorly functioning interfaces lead to duplicated work, errors and delays. Anyone who wants to digitalise HR sustainably must therefore first get clarity on data flows, responsibilities and system breaks.

System breaks slow down digital processes

In many organisations, HR processes are digitally supported but not connected end to end. Once an offer has been accepted, the recruiting system already holds application, contact and qualification data. HR now has to decide which details are carried over into the authoritative employee master record, which system maintains it and who is accountable for its accuracy.

Often, HR administration transfers the data manually and adds contract details, working hours and organisational assignment. The information then goes on to payroll, IT, time recording or learning and development. On paper, this process looks simple. In practice, manual entries and hand-offs by email or Excel lists cause errors or contradictory information. This becomes a real challenge, for example, when statutory deletion cycles have to be applied across systems. The problem lies less in individual tools than in the interfaces and in how the systems interact.

When data moves, responsibility moves too

Across the employee lifecycle, data does not just move between systems but also between areas of responsibility. Recruiting, HR administration, payroll, learning and development and line managers all need different data and access rights. A responsibility matrix can clarify, for every step, who captures, checks, changes and corrects data. What matters is that HR, IT, data protection and process owners put these rules into practice together, day to day.

Process analysis comes before digitalisation and AI

Before digitalising or introducing AI, organisations should analyse how the process actually runs: which data is needed, where does it originate and in which systems is it held? Clear data flows and responsibilities are the foundation of any HR digitalisation and any use of AI. A simple process map shows where data is created, checked and passed on. Hand-offs and manual intermediate steps are particularly revealing, because that is where waiting times, queries and misunderstandings tend to arise. The analysis should question every process step: Is this field really needed? What happens to the information captured? Why is a piece of information recorded twice? Which decisions can be automated, and where is a professional or legal review required? This makes entrenched routines and gaps in automation know-how visible – and therefore changeable.

Data quality is more than a completed field

Data quality has several dimensions. Data must be up to date, consistent, fit for purpose and unambiguously attributable. In addition, its source and most recent changes should be traceable.

Even records that are formally complete can be unusable. Differing definitions of “headcount”, for instance, produce different results. Are people on parental leave, marginal part-time employees, freelance contractors or people in phased retirement included? Data management needs shared terminology and binding rules.

For AI, data must also suit the use case. Historical data sets may contain past selection patterns or discrimination that an AI system adopts or amplifies. AI-ready data quality therefore also includes fitness for purpose, representativeness and traceability.

GDPR: accuracy, purpose and traceability

HR works with personal data. The General Data Protection Regulation therefore does not only become relevant when a new AI system is introduced. It applies to the entire data flow, from capture in recruiting through to later use in HR processes and analytics.

The GDPR principles include lawfulness, transparency, purpose limitation, data minimisation, accuracy, storage limitation, and integrity and confidentiality. In practice, this means HR must document why a piece of information is needed, who may access it and when it will be deleted. If a system does not merely prepare HR decisions but takes them solely by automated means, with significant effects on the person concerned, the GDPR requirements for automated decision-making must also be assessed.

EU AI Act: understand the use case first

The AI Act also classifies certain AI systems in the area of employment and workforce management as high-risk. These include, in particular, systems that play a significant role in selecting and hiring people, and in decisions on working conditions, promotions or the termination of employment. Solutions that allocate tasks or evaluate or monitor people may also be covered.

Organisations may only use high-risk systems under strict conditions. The systems must not take decisions on their own without a human reviewing them. The data used must be relevant and representative for the purpose – and organisations must monitor and document the system’s operation. Before such a system is deployed in the workplace, the affected employees or their representatives must be informed. A fundamental assessment of the risks and impacts is also required. Not every digital HR application falls into the high-risk category. The classification depends on the system’s purpose, function and actual impact. A legally sound assessment must therefore look at the process, not just the tool.

AI changes data structures and ways of working

In this assessment, organisations must not forget one thing: AI systems do not just use data. They also generate new data from existing information, such as structured notes or machine-readable files. These can be used to document knowledge from conversations, capture skills and experience in a clear overview, analyse processes, identify training needs or automate recurring tasks.

This is also changing the work of human resource management. Some routine data capture and preparation may disappear. That creates more room for professional judgement, conversations and more personal support for employees. It does, however, require people to check automatically generated content, interpret it correctly and use it responsibly. New data needs quality rules just as clear as those that apply to manually captured data.

AI can make existing quality problems worse: it processes incorrect assignments, inconsistent terminology and outdated information faster and on a larger scale than a manual process. As soon as AI prepares or significantly influences decisions about applicants or employees, a particularly careful legal and professional review is required. This is why the new AI skills that HR staff need to acquire specifically for their day-to-day work are so important. They include an understanding of the new data structures that AI systems will systematically draw on.

How to get an HR data project started

In practice, this means digitalisation projects should begin by creating a clean data structure. That takes time. Overhauling data management is not a “side project”. The task is to map the current state step by step along the employee lifecycle: Who captures which information, when, in which system and for what purpose? And who passes it on, and when?

Informal intermediate steps such as emails, personal notes or Excel lists must also be made visible, because that is often where delays and errors arise. For every key piece of information, the project leads should define where it originates, which system serves as the leading source, who is allowed to check the data and who is responsible for its quality, currency and correction. This way, data is not simply stored but actively managed.

After every adjustment, the project team measures whether errors, queries and duplicate entries are decreasing. Data management is then understood not as a one-off project but as an ongoing leadership and improvement task within HR management.

Conclusion

Data management is the foundation of successful HR digitalisation and responsible use of AI. Those who understand their data flows, system boundaries and responsibilities can spot technical gaps, organisational ambiguities and unnecessary steps.

Digitalisation therefore does not begin with a new tool, but with binding rules for data quality, leading systems and responsibilities. For HR to shape these changes actively, it also needs targeted data, automation and AI skills.