Responsible AI
HR decisions affect people directly — and employers carry a legal duty of care (Swiss Code of Obligations, Art. 328). That's why avalyx is built around clear principles that keep the use of AI in HR processes controllable and traceable.
1. No judgments about individuals
avalyx agents never conduct employee conversations and never render individual judgments about people.
2. Deterministic where it matters
Analytical results are computed deterministically — no LLM-invented scores in evaluations.
3. Human in the loop
Critical actions pass through explicit approval gates. For particularly high-risk decisions, an additional second approval can be configured, tailored to the customer's governance requirements (e.g., dual approval by HR and CFO). New agents run in shadow mode before going live.
4. Full auditability
Every agent decision is logged and traceable; corrections happen only as new, versioned records — nothing is silently overwritten.
5. Data sovereignty
On-premise inference is available — data can stay entirely on the customer's infrastructure; compliant with the Swiss FADP and GDPR, with data subject rights endpoints built in.
6. Regulatory alignment
We develop avalyx in line with the requirements of the EU AI Act for high-risk HR systems (transparency, human oversight, documentation).
Scope note
- avalyx is software that supports HR processes. It does not provide or replace legal, tax, or social-insurance advice.
- Legal responsibility for filings and employer obligations always remains with the employer. avalyx supports this responsibility through full auditability (audit trail) and explicit approval gates — it does not take it over.
- avalyx acts as an intelligent review layer for compliance requirements, used alongside certified payroll systems, and does not replace certified payroll software.