Artificial intelligence isn't going to replace your HRIS. It will build on it.
It's that subtle difference that makes all the difference, and that's where we need to start, because that's exactly where many HR AI projects fail.
An agent named Claude “knows” nothing about your company. He knows nothing about your headcount, your collective bargaining agreements, an employee’s history, or your leave policy. Everything he produces depends on the information you give him. Give it incorrect data, scattered across twelve Excel files and three email inboxes, and it will produce an incorrect answer with confidence. Give it a structured, up-to-date, and reliable database, and it becomes an incredibly accurate tool.
This structured, up-to-date, and reliable database is your HRIS. It's the foundation.
The founding principle of AI in HR: an agent is only as good as the quality of the data it works with. HRIS is not software that AI renders obsolete; rather, it is the prerequisite for AI to be useful, accurate, and compliant.
Throughout this guide, you'll see this approach come up again and again.
For each category of staff, we will specify what the platform provides. Because a recruitment assistant without its own employment data, a payroll specialist without reliable variables, or a compliance officer without an up-to-date reference framework—that’s not AI: it’s a risk.
That is exactly what quarksUp provides: a single, governed source of truth where agents can retrieve accurate information and write back clean data . AI on top, the foundation below.
You don't have to be a technical expert to understand this, but you need to go in with the right expectations. Too many AI projects fall short because people either expected too much or expected the wrong thing.
First, what is an agent? It's simply Claude, whom you've configured once for a specific HR task, and who then always responds in the same way, using the same framework.
Instead of explaining the context all over again each time, you give it to them right from the start:
- a role ( “You are an HR assistant responsible for…”),
- a framework ( the tone, what is prohibited, the rules to follow)
- and a subject ( data from the HRIS).
Expertise packaged and available on demand. In Claude, this takes the form of a Project: a dedicated space with its own instructions and reference documents (we’ll see how to create one in the next post).
Next, what you can reasonably expect from it.
Handle large volumes of information. Read , sort, summarize, and compare dozens of documents in just a few minutes—where humans would become overwhelmed and lose focus.
Ensure consistency. Always produce content according to the same guidelines, regardless of the day, location, or writer. No more responses that vary depending on who handled them.
Make the first draft count. Start with a solid outline rather than a blank page: letter, summary, outline, or template.
Make the data easy to understand. Turn a social statistics table into an easy-to-read summary; explain a pay stub in plain language.
Be available at all times, without fatigue or lapses in attention, starting at 6 p.m. on Friday.
Let it decide for you. An agent prepares and makes suggestions; judgment and responsibility remain with you. That is the golden rule of this entire guide.
It knows things that weren't given to it. It doesn't make up your internal rules. Without reliable input data, it fills in the gaps, and that's when it "hallucinates."
It may be reliable, but it’s built on shaky ground. Incorrect or scattered input data = an incorrect answer, stated with confidence.
Let it complement expertise. It enhances it and saves time; it does not replace it, especially when it comes to sensitive topics (legal matters, payroll, personal data).
The right way to view an agent: not as an independent decision-maker, but as a tireless, lightning-fast junior collaborator who is supervised and whose work is reviewed. Its value depends entirely on two things: the framework you provide, and the quality of the data set it works with.
Let's take a look at the catalog.
Here are the six key areas where a Claude agent adds immediate value to HR. For each one, we’ll explore the central concept and the role of the data foundation.
Drafting job postings, sorting and summarizing resumes, preparing interview rubrics, writing structured reports, and responding to candidates. The recruiter saves you hours on the administrative side of recruitment so you can focus on the people.
The foundation: job descriptions , position profiles, and performance evaluation criteria are stored in the HRIS. Employees refer to these to ensure they remain aligned with your actual needs, not with a generic idea of the job.
Personalized onboarding programs, welcome kits, answers to new hires’ questions, and 30-, 60-, and 90-day check-ins. The agent ensures a consistent experience regardless of the manager or location.
The basics—position , team, start date, equipment, access—are all already in the HRIS. The HR specialist uses this information to create a workflow that’s truly tailored to each new hire, without having to re-enter any data.
Preparing and verifying payroll variables, explaining pay stubs to employees, drafting amendments and letters, and tracking deadlines. The agent performs the initial review and ensures accuracy, while the expert approves the work.
The foundation: variables (hours, bonuses, absences, time off) are the raw material of payroll. If they are accurate and properly structured in the HRIS, the payroll specialist can verify and explain them without error. Otherwise, the error is propagated.
Initial review of labor law issues, drafting of initial templates (disciplinary letters, termination notices, memos), monitoring of regulatory requirements, and preparation of case files. The agent lays the groundwork to save time for human experts.
The foundation: applicable collective bargaining agreement , seniority, status, disciplinary history—the employee can only make sound judgments based on accurate facts drawn from the HRIS. And the employee never makes the final decision: they prepare the case, and a human makes the call.
Compilation of social data, analysis of key performance indicators (turnover, absenteeism, total payroll, workplace equality), preparation of the BDESE and financial statements, and decision support for the Executive Committee.
The foundation: here more than anywhere else, “garbage in, garbage out.” Reliable analyses require consolidated and consistent data—exactly what an HRIS centralizes. The analyst transforms clean data into strategic insights.
Identifying needs, developing development plans, recommending career paths, providing educational materials, and summarizing annual and performance reviews.
The foundation: skills possessed, training history, career goals, past interviews—the HR representative cross-references this data from the HRIS to make well-informed recommendations, not random ones.
Field: Recruitment
Mission: Review a large number of resumes, extract a comparable and structured summary from them, and identify the profiles that are a good fit for the position—without ever making the decision for you.
What a difference it makes. Whereas sorting through 80 resumes used to take you half a day, the tool now generates a standardized profile for each candidate in just a few minutes (background, key skills, areas to watch, fit with the role). You retain the final judgment; it takes the tedious reading off your hands.
A concrete example.
Submission: the text of a resume + the job description for “Experienced Payroll Manager.”
Output: “ Fit: Strong. 6 years of experience in multi-collective-agreement payroll (vs. 3 years required). Proficient in Silae and SAP. Point to watch: No experience in an industrial environment. Question to ask in the interview: Managing variable components for large volumes.”
The role of the foundation. The job description, evaluation criteria, and competency matrix are sourced from your HRIS (Recruitment module). The recruiter uses these to evaluate each candidate based on your actual needs, not on market standards. And the summary generated can be fed back into the candidate’s file: the data remains centralized and isn’t lost in an email inbox.
The safeguard. The system makes a recommendation; the human makes the decision. There is never any automatic screening: the risk of discrimination (even unintentional) and the GDPR framework require a human decision. We’ll revisit this in the section on compliance.
Area: Onboarding
Mission: To create a personalized onboarding plan for each new hire and answer their questions during their first few weeks, ensuring a consistent experience regardless of who their manager is.
What a difference this makes. No more onboarding that depends on the manager’s availability. Every new hire receives the same level of attention: a plan for the first 90 days, key contacts, and immediate answers to common questions (time off, equipment, health insurance, remote work).
A concrete example.
Input: “New hire: Léa Martin, Communications Specialist, Marketing Team, starting September 2, Lyon office.” Output: a structured week-by-week onboarding plan, access permissions to arrange, meetings to schedule (with manager, HR, mentor), and resources to provide—all tailored to the position and location.
The role of the platform. Position , team, manager, location, start date, equipment: this information already exists in the HRIS (People & Onboarding modules). The agent reviews it to customize the onboarding process without having to re-enter any data and can check off completed steps directly in the file; the HRIS serves as the onboarding dashboard.
The safeguard. The agent only shares verified information. This is based on your official documents (onboarding handbook, internal policies) rather than on the agent’s general knowledge, to prevent the agent from inventing a leave policy that doesn’t exist at your company.
Field: Payroll & Administration
Responsibilities: Prepare and verify variables before the payroll run, identify discrepancies, and explain pay stubs to employees in plain language.
What a difference it makes. The agent reviews the month’s data and flags anything out of the ordinary—an unusual bonus, an inconsistent number of hours, or unused vacation time— before the error appears on the pay stub. And when an employee doesn’t understand their pay stub, the agent explains it to them line by line, without using jargon.
A concrete example.
Input: This month's variable table. Output: " 3 points to check. 1 ) Employee #4471: 182 hours reported vs. 151 usual hours; overtime to be confirmed. 2) One-time bonus of €1,200 with no reason provided. 3) Absence from May 12 to 14 not linked to any absence type."
The Role of the Foundation. This is the most telling example of the importance of the foundation. Payroll does not tolerate approximation: the payroll specialist can only verify specific, structured variables. If hours, absences, and bonuses are entered consistently into the HRIS (Absences, Compensation, and Hello-Payroll modules), the payroll specialist acts as a safety net. If this data is scattered across various files, there is nothing reliable for the payroll specialist to verify.
The safeguard. The agent prepares and notifies; he or she never approves a payroll run alone. Responsibility remains with a human. And no sensitive payroll data should ever leave your secure environment; this topic is covered in the section on system integration.
Field: Legal & Social Compliance
Mission: To provide an initial review of a labor law issue and draft initial document outlines, thereby saving time for your legal team (in-house or outside counsel).
What a difference it makes. Instead of starting from a blank page, in just a few seconds you get a draft of a letter (notice of meeting, warning, mutual termination agreement) or a focused analysis of a recurring issue. The expert reviews and refines it, rather than writing everything from scratch.
A concrete example.
Input: “Employee on a permanent contract, 4 years of service, Syntec collective bargaining agreement. Three unexcused tardies this month. What are the options, and what should the initial letter template look like?” Output: a summary of the applicable framework, the graduated disciplinary measures (meeting, formal reprimand, written warning), and a warning letter template to be customized and approved.
The Role of the Foundation. The applicable collective bargaining agreement, seniority, status, and employment history of the employee are sourced from the HRIS. The system can only make accurate determinations based on correct facts: if the initial collective bargaining agreement is incorrect, the entire reasoning is flawed. The foundation ensures that the system starts with the correct data.
The safeguard. The agent is not a lawyer and does not issue legal decisions. The agent lays the groundwork and prepares the case; any decision with significant implications (such as a penalty or termination) must be approved by a qualified human reviewer. The compliance of the usage itself is covered in detail in this series.
Field: People Analytics & Management
Mission: To transform your HR data into clear insights and decision-making tools for management.
What a difference it makes. You query your data using natural language, just as you would when speaking to an analyst. The agent generates a summary, highlights trends, and prepares the materials (BDESE,
social report (COMEX agenda item), without having to manually create a pivot table.
A concrete example.
Entrée : « Donne-moi l'évolution du turnover par département sur 12 mois et les points d'alerte. » Sortie : « Turnover global stable à 11 %. Alerte : département Support à 23 % (+9 pts vs an dernier), concentré sur les profils <2 ans d'ancienneté. Recommandation : analyser les entretiens de départ de cette population. »
The role of the data repository. Here , the “garbage in, garbage out” principle is unforgiving. An HR analysis is only valuable if the data is consolidated, consistent, and complete: a single data repository, not ten out-of-sync exports. That is exactly the role of the HRIS: to centralize data so that the user sees a reliable picture of reality. Without a clean data foundation, people analytics produces false precision, which is worse than having no figures at all.
Safeguards. Social data is sensitive and is often aggregated for legal reasons (anonymity thresholds). The user must work within a secure environment and comply with these thresholds, which is a key aspect of the integration with your HRIS.
AI in HR isn't about some kind of magic software, but rather about having a clean data foundation on which to build well-defined agents.
One agent = HR expertise packaged into a system, configured once, and available around the clock. The HRIS is not threatened by AI—it is a prerequisite for it. It is the single source of truth that the agent reads and updates.
The golden rule, which applies to all six families: the agent prepares, the human decides.
quarksUp serves as this foundation: a structured, up-to-date, and well-managed HR system designed to ensure that your AI agents work accurately, quickly, and in compliance.
This guide is the first in a series designed to walk you through the process step by step:
1. You are here: The Catalog of AI Tools for HR.
2. Setup: How to configure your first Claude agent (or another LLM) from scratch.
3. 10 Use Cases: Claude in Everyday Life, Connected to Your HRIS.
4. Regulation & Compliance: The AI Act and GDPR—What an HR Director Needs to Ensure.
5. Integration: Connect the AI to your HR software—simply and securely.
