Evidence-based HR
On 1 August 2026 the transparency obligations of the European AI Act came into force. Most organisations cannot explain what their AI decides. HR can close that gap, and finally give itself the purpose it always lacked: no more instruments that claim to explain people, but evidence that holds up. That is called evidence-based HR, and it is also the business case for a CFO.
A law that came into force on 1 August, not next year
On 1 August 2026 the transparency obligations of the European AI Act came into force for employers. Not next year, not “soon”. On that day. Every organisation that uses AI in how it manages people must now be able to explain that use clearly to the people it concerns, and to the works council that represents them.
Most organisations cannot do that. Not out of carelessness, but because nobody was ever given the task of translating what AI does into something that someone on the work floor, or a union delegate in a meeting, can actually follow. That translation gap has existed for years. Since 1 August 2026 it is a legal problem.
The gap HR never got to close
HR has been trying to prove for twenty years that it belongs at the table: better technology, bigger titles, a “seat at the table” whose terms nobody ever set out. What HR never got was a clear purpose of its own.
In that vacuum others did write the job description: IT saw it as a ticketing system, Finance as a cost item, and employees, study after study, as the department to be afraid of, not the department that helps.
What evidence-based HR actually means
HR has never been short of instruments that claim to explain people: personality types, colour codes, a training day and a certificate. What was mostly missing was evidence. Not intuition packaged as insight, but instruments that really held up under scrutiny: decades of published research, real reliability, genuine predictive value.
This is the moment that changes, if HR chooses to claim it for itself instead of waiting for someone else to fill it in again.
What the CFO actually sees in this
Because a CFO needs this, not as a favour to HR, but as real financial and legal exposure: an AI decision that nobody in the organisation can explain is not an ethical blemish. It is an unbudgeted liability. In the United States, one of the largest collective lawsuits ever against an HR software vendor is under way right now, precisely on this point: judges are finding that an employer cannot hide behind “the vendor did it.” Waiting until a works council, a regulator or a court asks the one question nobody prepared an answer for is not a strategy.
The financial argument for automating for the sake of automating is also weaker than most CFOs assume. A Gartner survey of 350 companies with more than a billion euros in revenue found that those that cut the most staff after AI implementation performed, on average, no better than those that cut the least, and sometimes worse. Keeping human judgement and explanation in the loop was never a soft cost. It is the price of the alternative: paid up front, instead of discovered afterwards.
What is left standing
Beneath the compliance question sits the question that really matters, the one that outlives this particular law: what does this organisation leave behind?
A company that used this moment to automate without explaining, to redesign without building, and to cut without reinvesting will be efficient and forgettable. A company that used this exact moment to build real explainability, redesign and legacy into how it works with AI becomes the company people choose to stay with, and the company others will eventually have to compete against.
The law made 1 August 2026 the deadline. Legacy is decided by what happens now.
From surviving to flourishing with AI: strengthening agile, ethical leadership, removing psychological suffering at work, filtering abuse out early, and lifting engagement to real flow. That is not one idea next to the rest of this programme. It is what the whole programme, from diagnosis to Drive, is trying to build together.
This website was prepared with the support of AI tools for drafting and structuring text. Every text, every instrument and every claim has been personally reviewed, checked for content and approved for publication by me. AI was used as an aid in the process, not as a source of the content or the method itself.
© 2026 Astrid Van Tornout / WorkDesAIgn · All rights reserved