An AI HR operating model is the way an HR function organizes its people, processes, data and technology once AI handles a large share of HR service delivery: routine questions and transactions move to AI-assisted self-service, while HR professionals concentrate on advice, judgment, governance and the design of work. It is a change of structure, not a new tool, and it affects headcount, roles, skills, budgets and pay inside HR itself.
This guide focuses on that structural question. It explains how AI changes the classic three-pillar model, which HR tasks are being automated and which are augmented according to official and peer-reviewed research, which HR roles employers are hiring for according to TalentUp’s global job-ad data, and what the EU AI Act, GDPR and worker consultation rules require. For use cases by HR function, see our guide to AI in HR; for the individual skills HR people need, see AI fluency in HR.
Key takeaways
- AI does not delete the three-pillar HR model, it rebalances it: shared services shrink and become AI-first, while data, technology and governance roles grow.
- The ILO places payroll clerks and personnel clerks in its highest generative AI exposure group and HR managers in the lowest.
- In TalentUp’s data, people analytics and HR systems roles gained share of HR job ads between 2024 and 2026 in almost every major market.
- 13.5% of people analytics ads ask for AI skills in 2026, against 2.5% of HR generalist and admin ads.
- In the EU, AI used for hiring, promotion, termination and worker evaluation is high-risk; those obligations apply from 2 December 2027.
From the Three-Pillar Model to an AI HR Operating Model
Most large HR functions still follow the three-pillar model popularized in the late 1990s: HR business partners (HRBPs) who advise business leaders, centres of expertise (CoEs) for topics such as reward, talent and learning, and shared services that run transactions and answer employee questions, often supported by a tiered help desk.
That structure was designed before AI was part of everyday business software. According to Eurostat, about 20% of EU enterprises with 10 or more employees used AI technologies in 2025, and 55% of large enterprises did. AI puts pressure on each pillar differently:
| Element |
Classic three-pillar model |
AI HR operating model |
| Service delivery |
Tier 0 portal, tier 1 help desk, tier 2 specialists |
AI assistants and agents answer and execute most routine requests; people handle exceptions and sensitive cases |
| Shared services |
Large teams for data entry, contract changes, payroll queries |
Smaller teams that supervise workflows, check AI outputs and fix data quality |
| Centres of expertise |
Design policies and programmes, often with limited data |
Design policies with analytics and market data; own the rules that AI applies |
| HR business partners |
Mix of advice and operational firefighting |
More time on workforce planning, organization design and change; less on transactions |
| Data and technology |
Often an IT or HRIS side function |
A core capability: HR systems, people analytics, data governance |
| Governance |
Policies, audits, data protection |
Adds AI inventories, risk classification, human oversight, worker information |
| Main KPI |
Cost per employee served, ticket resolution |
Quality and fairness of decisions, adoption, time freed for advisory work |
Two practical consequences follow. First, the boundaries between pillars blur: an AI agent that answers a pay question uses rules written by the reward CoE, data maintained by HR systems and escalation paths owned by business partners. Second, ownership of data and rules becomes the real organizing principle, because AI only performs as well as the job architecture, policies and records it reads. Our guide to the digital transformation of HR in Europe covers that data foundation in detail.
Which HR Tasks AI Automates and Which It Augments
The strongest evidence on exposure comes from occupation-level studies rather than vendor surveys. The ILO’s refined global index of generative AI exposure (2025), which scores the tasks of each occupation, ranks HR occupations very differently:
| HR occupation (ISCO-08) |
ILO exposure group |
Mean task exposure score |
| Payroll clerks (4313) |
Gradient 4, the highest exposure group |
0.61 |
| Personnel clerks (4416) |
Gradient 4, the highest exposure group |
0.60 |
| Personnel and careers professionals (2423) |
Gradient 2 |
0.45 |
| Human resource managers (1212) |
Minimal exposure |
0.36 |
The ILO stresses that exposure is not job loss: because most occupations consist of tasks that still require human input, transformation of jobs is the most likely impact. Globally, it estimates that one in four workers is in an occupation with some generative AI exposure, and clerical occupations remain the most exposed.
Early labour-market evidence points in the same direction. The revised “Canaries in the Coal Mine?” paper by Brynjolfsson, Chandar and Chen at the Stanford Digital Economy Lab (August 2026), based on US payroll records, finds that employment of workers aged 22 to 25 in AI-exposed occupations is 19% below where it would have been had it kept pace with less-exposed peers, with no comparable gap for experienced workers. Declines concentrate in occupations where AI mainly substitutes for tasks; where it mainly complements workers, employment is flat or rising.
For HR leaders, that translates into a simple map:
| Type of HR work |
Likely effect of AI |
Examples |
| Transactional and clerical |
Largely automated or AI-executed, with human checks |
Data entry, contract letters, payroll queries, absence records, standard policy answers |
| Analytical |
Strongly augmented |
Pay outlier detection, attrition analysis, market matching, workforce scenarios |
| Advisory and relational |
Augmented, still human-led |
Organization design, employee relations, negotiations, change management |
| Decisions about individuals |
Must stay human, by law and by trust |
Hiring, promotion, pay, dismissal |
The risk sits at the bottom of the ladder: if shared services shrink and entry-level HR administration disappears, organizations lose the roles where HR professionals traditionally learned the job. The operating model needs a deliberate answer to that, such as rotations, apprenticeships in HR operations or junior analyst roles.
Which HR Roles Employers Are Hiring For: TalentUp Data
+37%
relative growth in the share of HR job ads for people analytics roles (1.3% to 1.7%)
11 of 11
major markets where people analytics gained share
13.5%
of people analytics ads ask for AI skills, vs 2.5% of HR generalist and admin ads
+77%
higher median advertised salary for people analytics vs HR generalist and admin roles in the US
Source: TalentUp job postings database, HR job ads worldwide, January–September 2024 and 2026.
Job ads show where employers are actually putting HR headcount. We classified HR job ads by title into seven role families and compared their share of all HR ads in the same 11 large markets (the US, India, Canada, Mexico, Italy, Germany, the UK, Australia, France, the Netherlands and Spain), giving each country equal weight so that changes in source mix do not drive the result.
| HR role family |
Share of HR ads 2024 |
Share 2025 |
Share 2026 |
Markets where share rose, 2024 to 2026 |
| Recruiter and talent acquisition |
47.9% |
48.2% |
45.7% |
5 of 11 |
| HR generalist, admin and people operations |
22.4% |
22.2% |
21.3% |
7 of 11 |
| Payroll |
16.8% |
15.6% |
18.0% |
6 of 11 |
| HR business partner |
8.6% |
9.6% |
9.7% |
8 of 11 |
| HR systems and HRIS |
2.5% |
2.7% |
3.1% |
10 of 11 |
| Compensation, benefits and reward |
2.3% |
2.3% |
2.7% |
9 of 11 |
| People analytics and workforce planning |
1.3% |
1.2% |
1.7% |
11 of 11 |
What the data shows:
- Data and technology roles are the clearest winners. People analytics and HR systems are still small, but they gained share in practically every market, consistent with an operating model where data ownership becomes central.
- HR business partner and reward roles are holding up or growing, which fits the shift of HR towards advisory and pay-design work. Pay transparency rules add to the demand for reward specialists in Europe.
- There is no global collapse of HR administration yet. Generalist and admin roles lost about one point of share on average, with the sharpest falls in the US, Canada, the UK and Australia, while payroll moved in different directions by country. Exposure, as the ILO notes, is not the same as replacement.
- Recruiting is cyclical. Its share depends on overall hiring volumes as much as on AI, so we would not read the dip as an AI effect.
Which HR Roles Ask for AI Skills
| HR role family |
Ads asking for AI skills, 2024 |
Ads asking for AI skills, 2026 |
Ads analysed, 2026 |
| People analytics and workforce planning |
3.2% |
13.5% |
~1,160 |
| HR systems and HRIS |
2.8% |
11.1% |
~1,280 |
| Recruiter and talent acquisition |
2.6% |
6.8% |
~1,400 |
| Compensation, benefits and reward |
0.3% |
5.4% |
~1,310 |
| HR business partner |
0.3% |
3.8% |
~1,410 |
| HR generalist, admin and people operations |
0.1% |
2.5% |
~1,220 |
| Payroll |
0.2% |
2.2% |
~1,080 |
Typical requirements read like “demonstrated use of AI tools to increase HR efficiency”, “experience leveraging AI-powered tools in HR operations is a plus” or “applied AI literacy relevant to compensation and benefits practices”. AI skills are becoming part of the job description in the roles that build and run the new model, while they are still rare in the transactional roles most exposed to automation.
What HR Roles Pay
Where the sample is large enough, advertised salaries confirm that the growing roles are also the better-paid ones. Median advertised annual salaries in employee job ads, 2026, converted to EUR:
| HR role family |
United States |
Canada |
United Kingdom |
| HR generalist, admin and people operations |
€71,600 (n=683) |
€47,600 (n=429) |
€38,500 (n=644) |
| Payroll |
€91,100 (n=438) |
€56,700 (n=265) |
€37,900 (n=489) |
| HR business partner |
€121,100 (n=384) |
€70,500 (n=142) |
€56,800 (n=208) |
| Compensation, benefits and reward |
€125,100 (n=210) |
€78,600 (n=73) |
not enough data |
| HR systems and HRIS |
€111,700 (n=380) |
€55,400 (n=130) |
€57,700 (n=32) |
| People analytics and workforce planning |
€126,800 (n=201) |
€80,000 (n=33) |
not enough data |
In the US, the median people analytics ad offers about 77% more than the median HR generalist and admin ad, and HR systems roles about 56% more. Part of the gap reflects seniority, because generalist and admin roles include many assistant positions, but the direction is clear: moving HR budget from transactions to data, technology and advisory roles raises the average cost per HR head even as total HR headcount may fall. For the wider market picture, see our analysis of the AI pay premium.
New Roles and Capabilities in an AI-First HR Function
The roles that grow in the data map onto the capabilities an AI HR operating model needs:
| Capability |
Typical roles |
What they own |
| People analytics |
People analytics lead, HR data analyst, workforce planning analyst |
Metrics, models, workforce scenarios, analytical quality of AI outputs |
| HR systems and data |
HRIS manager, HR systems analyst, HR data steward |
Core records, integrations, data quality, configuration of AI features |
| AI governance in HR |
HR AI lead, often with legal, data protection and IT |
AI inventory, risk classification, human oversight, worker information, vendor checks |
| Employee experience and service design |
HR service owner, HR product manager |
Journeys such as onboarding or leave, which parts AI handles and when it escalates |
| Reward and pay equity |
Compensation analyst, reward manager |
Job architecture, salary bands, pay equity analysis, pay transparency compliance |
| Organization and work design |
HR business partner, organization development |
Redesigning roles, workflows and decision rights as AI changes the work |
The common thread is ownership: someone must own each data domain, each rule set the AI applies and each decision that stays human. HR teams that build on data analytics in talent management and a skills view of work, as in skills-based pay, find these roles easier to define.
EU AI Act, GDPR and Worker Consultation
The operating model has to be designed around European rules from the start, because HR is one of the areas the EU AI Act treats as high-risk.
The EU AI Act
Annex III of the AI Act (Regulation (EU) 2024/1689) lists as high-risk AI systems intended to be used for:
- recruitment or selection, in particular placing targeted job ads, analysing and filtering applications and evaluating candidates;
- decisions affecting terms of work relationships, promotion or termination;
- allocating tasks based on individual behaviour or personal traits;
- monitoring and evaluating the performance and behaviour of workers.
Timeline. The ban on emotion recognition in the workplace (except for medical or safety reasons) has applied since 2 February 2025. The obligations for stand-alone high-risk systems, including employment, were originally due on 2 August 2026; the Digital Omnibus on AI, Regulation (EU) 2026/1744, published in the Official Journal on 24 July 2026, moved them to 2 December 2027. Transparency duties, such as telling people they are interacting with an AI system, were not postponed.
Deployer duties for employers using high-risk HR systems include using them in line with the provider’s instructions, assigning human oversight to people with the necessary competence, training and authority, monitoring their operation, keeping the logs under their control and, before putting a system into service at the workplace, informing workers’ representatives and the affected workers. Affected people also gain a right to an explanation of individual decisions taken on the basis of a high-risk system’s output.
GDPR
The GDPR applies alongside the AI Act. Article 22 restricts decisions based solely on automated processing that significantly affect people, a data protection impact assessment is usually needed for large-scale people analytics or monitoring, and Article 88 allows member states to adopt more specific rules for employment data. An AI HR operating model therefore needs data protection built into service design, not added after go-live.
The EU framework directive on information and consultation (2002/14/EC) requires employers above national thresholds to inform and consult employee representatives on decisions likely to lead to substantial changes in work organisation, which a redesign of HR service delivery can be. National rules often go further: in Germany, for example, works councils have co-determination rights over technical systems capable of monitoring employees’ behaviour or performance (Works Constitution Act, section 87(1) no. 6). If the new model leads to collective redundancies, the Collective Redundancies Directive (98/59/EC) adds consultation duties. Involving employee representatives early is both a legal requirement in many countries and the best way to build trust in AI-assisted HR.
What It Means for HR Headcount, Budgets and Pay
There is no reliable official benchmark for “the right HR-to-employee ratio” in an AI-first model, and vendor estimates vary widely, so we do not quote one. What the evidence does support:
- Headcount shifts before it shrinks. Job-ad demand is moving towards analytics, systems, reward and business partnering, while clerical HR work is the most exposed. Expect smaller shared services and a higher share of specialist roles.
- Entry-level roles need protection by design. The Stanford findings on young workers suggest that the first jobs to thin out are junior ones. Build career paths that do not depend on transactional work disappearing slowly.
- Budgets move from people to platforms and data. Spending on HR systems, data quality and AI governance rises, and the remaining HR roles cost more per head. Plan the business case on total HR cost, not on headcount alone, using real-time salary data for HR budget planning.
- Pay structures inside HR need updating. New roles such as people analytics or HR systems lead need their own job levels and salary bands, priced against the market rather than against legacy HR grades. As national laws transposing the EU Pay Transparency Directive (2023/970) apply, those levels must rest on objective, gender-neutral criteria.
How to Redesign Your HR Operating Model for AI
- Map HR work at task level. List the main activities of each HR team and classify them as automate, augment or keep human, using the categories above.
- Inventory AI already in use. Many HRIS, applicant tracking and payroll tools already contain AI features; classify each against the AI Act’s high-risk categories.
- Fix data and rules first. Clean core records, job architecture and policy content before scaling AI service delivery.
- Redesign service delivery. Decide which requests AI answers, which it executes and when it escalates to a person, especially for sensitive cases.
- Define ownership and new roles. Assign owners for data domains, rule sets, AI oversight and each human decision; create job descriptions and pay levels for the new roles.
- Inform and consult early. Bring in employee representatives, data protection and legal teams before go-live, not after.
- Reskill and protect career paths. Move people from transactional work to analytics, service design and advisory roles, and keep entry routes into HR open.
- Measure what matters. Track decision quality and fairness, employee adoption and satisfaction, time freed for advisory work and compliance readiness, not just ticket volumes.
Conclusion
AI is rewriting the HR operating model by changing where HR work happens and who owns it. Routine service delivery moves to AI-assisted self-service, while data, systems, reward and advisory roles carry more of the value. TalentUp’s job-ad data shows the first signs of that shift: people analytics and HR systems roles are gaining share almost everywhere, they ask for AI skills most often and they pay more than the transactional roles most exposed to automation.
In Europe, the redesign has a clear legal frame: high-risk AI duties for employment from 2 December 2027, GDPR and national consultation rules. HR leaders who redesign roles, data ownership and governance together, and who price the new roles against real market data, will be ready for both. To benchmark HR and other roles across countries, start with TalentUp’s salary benchmarking methodology.
Sources
- TalentUp. Job postings database, HR job ads worldwide, January–September 2024, 2025 and 2026 (see methodology above).
- Gmyrek, P., Berg, J., Kamiński, K., Konopczyński, F., Ładna, A., Nafradi, B., Rosłaniec, K. and Troszyński, M. (2025). Generative AI and Jobs: A Refined Global Index of Occupational Exposure. ILO Working Paper 140, International Labour Organization.
- Brynjolfsson, E., Chandar, B. and Chen, R. (2025, revised August 2026). Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence. Stanford Digital Economy Lab.
- Eurostat. Use of artificial intelligence in enterprises. Statistics Explained.
- European Union. Regulation (EU) 2024/1689 (Artificial Intelligence Act). EUR-Lex.
- European Parliament. Digital Omnibus on AI, Regulation (EU) 2026/1744. Legislative Train Schedule.
- European Union. Regulation (EU) 2016/679 (General Data Protection Regulation). EUR-Lex.
- European Union. Directive 2002/14/EC establishing a general framework for informing and consulting employees. EUR-Lex.
- European Union. Directive 98/59/EC on collective redundancies. EUR-Lex.
- European Union. Directive (EU) 2023/970 on pay transparency. EUR-Lex.