AI ethics in the workplace has moved from a theoretical boardroom debate to an urgent operational reality. Companies have poured an estimated $30–40 billion into AI, yet 95% of those investments have not delivered returns. The problem, as multiple HR leaders argued at the Horizon Summit 2025, is not the technology. It is the absence of ethical frameworks, cultural readiness, and governance structures to guide how that technology gets used — particularly when it touches people’s careers, data, and livelihoods.
What emerged from the conference was a clear and uncomfortable truth: most organisations are deploying AI they do not fully understand, in processes that directly affect employees, without adequate safeguards. For HR leaders, this is not someone else’s problem. It is theirs.
The Culture Problem Behind the Ethics Problem
Dan Strode, former head of culture at Santander, founder of Daniel Strode Consulting and a keynote speaker at the conference, framed the challenge directly. Organisations are stuck in what he called an “admin economy” — a cycle of endless meetings, emails, and process-heavy workflows. Without fundamentally changing how people work, AI will simply accelerate bad habits.
This is where AI ethics in the workplace becomes inseparable from organisational culture. When AI is layered onto broken processes, it does not just fail to deliver value — it actively amplifies inefficiencies and embeds biases at scale. An algorithm that screens CVs based on historically biased hiring patterns does not replicate one manager’s blind spots. It institutionalises them across every open role, every geography, every hiring cycle.
Strode’s argument is that culture is the true competitive advantage. The technology is democratised and accessible to everyone. What separates organisations that use AI responsibly from those that stumble into risk is whether their people are empowered to experiment, question, and course-correct. A growth mindset — where failure during experimentation is learning rather than punishable — is the precondition for ethical AI adoption.
Four Ethical Principles Every HR Team Should Adopt
Strode proposed a practical framework built on four ethical principles that HR teams can implement without waiting for regulation to catch up. These are worth examining in detail, because they offer something most ethical AI discussions lack: specificity.
The first principle is fairness, which in practice means bias auditing. Every AI tool used in HR — from recruitment screening to performance analytics — needs systematic checks for discriminatory patterns. This is not a one-time exercise. It requires ongoing monitoring as models learn from new data and as workforce demographics shift. The urgency is real: an estimated 87% of companies now use AI-driven tools in their hiring processes, and high-profile cases like the Workday class action lawsuit filed in 2024 — alleging systematic discrimination based on race, age, and disability — show what happens when bias auditing is absent.
The second is transparency. Employees and candidates deserve to know when AI is being used in decisions that affect them. A panel discussion at the conference reinforced this with a striking example. When a virtual HR assistant named “Eva” became temporarily unavailable, the team chose to tell users openly that they had been interacting with AI rather than concealing it. The result was not backlash — it was acceptance. People trusted the system more once they understood what it was.
The third principle is human oversight. Someone — a real person with accountability — must make the final call on decisions that significantly affect employees’ careers, compensation, or employment status. AI can inform, recommend, and surface patterns. But the moment organisations remove human judgement from consequential decisions, they cross an ethical line that no governance framework can walk back easily.
The fourth is privacy. AI systems in HR inevitably consume vast amounts of personal employee data. Who has access, how long data is retained, what it can be used for, and how employees can challenge decisions made about them — these are not IT questions. They are ethical ones that HR must own.
Strode’s core recommendation was that companies should lead with these principles proactively rather than waiting for regulation. The regulatory landscape will arrive unevenly across jurisdictions — though the EU AI Act, which entered into force in August 2024, has already classified AI used in recruitment and employment decisions as “high-risk,” with full compliance requirements enforceable from August 2026. Organisations that build ethical guardrails now will be ahead of compliance requirements, not scrambling to meet them.
When AI Replaces 30% of Your HR Team
If Strode’s keynote provided the ethical framework, Carlo Steenvoorden, EVP of HR at Dutch telecommunications company KPN, delivered a case study in what happens when AI ethics in the workplace meets the sharp edge of workforce transformation.
KPN was processing over 50,000 employee and manager queries annually, at roughly €15 per query. Sixty percent of that volume was repetitive questions. The company decided to build an in-house conversational AI assistant, rejecting off-the-shelf solutions to maintain flexibility and control.
The results were dramatic. Query costs dropped from €15 to as low as €0.15 per interaction. Ticket volumes fell noticeably, and HR business partners finally had capacity for strategic workforce planning.
But here is the part that makes this a story about ethics, not just efficiency. KPN replaced 30% of their HR staff — 15 to 16 people out of a team of more than 40 — within nine months. They brought in technical talent with AI, data flow, and system integration skills to replace case-handling roles that the technology had made redundant.
This raises the central ethical tension every HR leader adopting AI must confront. Efficiency gains are real and substantial, but they come from somewhere — frequently from displacing people whose roles have been automated. AI ethics in the workplace cannot be limited to how algorithms treat candidates or analyse performance data. It must encompass how organisations treat the people whose jobs are eliminated by the technology itself.
KPN’s approach included comprehensive AI literacy training for top managers — intensive two-day technical sessions, not superficial overviews. They were transparent about the transformation and invested in helping remaining staff transition into new roles. Whether that goes far enough is a question each organisation must answer, but the transparency and investment in upskilling represent a baseline many companies deploying AI have not yet met.
The Governance Gap Most Organisations Are Ignoring
Martin Barner, Head of Sandoz EX and HR Transformation, reinforced this with a warning. He observed that technological change is outpacing human ability to understand it — what he called the “control paradox.” Unlike individual human biases that tend to balance out across decision-makers, a biased AI model propagates errors instantly and uniformly across an entire organisation. One flawed recruitment algorithm does not produce one bad hiring decision. It produces thousands, all carrying the same systematic bias.
Barner’s recommendation was cross-functional governance involving HR, IT, legal, and data privacy teams. This means specific policies detailing which AI tools are permitted for which use cases, who has access, and where human review is mandatory. It also means auditing how many AI tools are actually operating in your organisation — including ones employees may have adopted without formal approval.
The panel discussion added another dimension: regional and cultural variation. Seventy to eighty percent of HR roles are expected to be affected by AI within three to seven years, but the impact will differ across geographies. Ethical frameworks need local adaptation while maintaining consistent core principles.
What Responsible AI Adoption Actually Looks Like
Across these sessions, a consistent set of practical priorities emerged for HR leaders who want to get AI ethics in the workplace right.
Start with a culture audit before a technology rollout. If your organisation punishes failure and discourages experimentation, no AI ethics policy will survive contact with reality. People need psychological safety to flag problems with AI outputs and challenge recommendations without fear of blame.
Build your ethics framework before you scale. Fairness, transparency, human oversight, and privacy are not abstract values — they are operational requirements that need written policies, assigned accountability, and regular review cycles. Share them openly with all stakeholders and train HR teams on implementation before expanding AI into higher-stakes decision areas.
Invest in AI literacy at every level. KPN’s two-day technical training for senior managers and Sandoz’s focus on organisation-wide transparency both point to the same conclusion: superficial awareness sessions are insufficient. Employees who do not understand technology cannot use it responsibly.
The Window Is Closing
With major technology companies investing a combined $380 billion in AI in 2025 alone, and HR operations already reaching what Gartner describes as the “plateau of productivity” for AI applications, the question is no longer whether AI will transform HR. It is whether HR teams will shape that transformation ethically or be shaped by it reactively.
AI ethics in the workplace is not a topic for next year’s strategy offsite. It is a capability that HR leaders need to build now — in their policies, their governance structures, and their willingness to have honest conversations about what this technology means for the people it is supposed to serve. In a world where global employee engagement has dropped to just 21% and trust in institutions is eroding, the organisations that get AI ethics right will not just avoid risk. They will earn the trust of their workforce at exactly the moment when trust matters most.
