Only 46 per cent of people globally are willing to trust AI systems, according to a KPMG study. For HR leaders rolling out AI tools across hiring, performance, and development, that trust deficit directly threatens adoption and return on investment. The question is whether algorithmic trust comes from system design or leadership behaviour - and the answer determines who owns AI outcomes.
Employees do not trust an algorithm because management tells them to. They trust it when they can see how decisions are made, when they can question them, and when leaders visibly act on the results. That requires both thoughtful system design and consistent leadership behaviour. The two are not interchangeable.
"People don't trust algorithms. They trust the culture and leadership behind those algorithms," said Prasad Menon, Chief People Officer at Amagi. "Technology can make trust scalable, but it cannot manufacture trust."
Amagi introduced Amber, an AI chief listening officer, to collect employee feedback. The recurring question was whether anyone would tell an AI system how they truly felt. The concern had little to do with AI itself - employees wanted to know whether leaders would listen and whether anything would change. The company has 25 nationalities across multiple continents and nearly every time zone.
Trust is not built when AI makes a recommendation, but when leaders consistently demonstrate what they do with those recommendations. At Amagi, every insight creates accountability, every trend triggers discussion, and every concern has an owner.
"Amber never owns the response. Leaders do," said Prasad. "The AI helps us hear. Leadership demonstrates that we listened."
System design defines who actually decides
Satyajit M. Menon, President and Global Head of People Experience at Innovaccer, said people do not resist AI because it is imperfect. They resist it because they cannot tell where the system's judgment stops and their own begins. Fix that line and trust improves.
Design matters in practical ways. "The override button should be as easy to use as the approve button," said Satyajit. "If overriding takes five extra clicks, people would stop bothering. That's when people realise they've quietly built a rubber stamp."
At Innovaccer, the system's only job is to provide context for decisions. The decision stays with the person it affects. The algorithm never acts alone. The firm maintains a pushback scorecard tracking how often decisions change when employees challenge AI, a "why clock" measuring how long it takes to get a real explanation, and consistency checks across teams.
Prasad backs similar metrics: action after insight, employee confidence in AI fairness, and how often leaders thoughtfully challenge a recommendation.
Who owns algorithmic trust?
HR leaders uniformly rejected the idea that technology teams alone should own algorithmic trust. Distributed ownership is the alternative, but only when decision rights are also distributed.
"If it's tech-only ownership, you get a model that's technically clean but organisationally dead on arrival," said Satyajit. "Splitting ownership without splitting actual decision rights just creates three teams that can each say 'Not my department' when something goes wrong. Distribute the authority, not just the meeting invite."
Pankaaj Phatak, Vice President - HR at Biotech Healthcare, described algorithmic trust as the new boardroom currency. In one organisation, a 40% attrition spike was traced back to target setting. AI's real value in that case was not making the decision but acting as a referee, and leadership behaviour determined whether flagged issues actually got addressed.
DHl Express India established a Digitalisation (Digi) Council, where cross-functional Digi Champions promote AI adoption and innovation. Its senior leadership intends for success to be measured by whether employees use AI confidently and responsibly, not by how advanced the algorithm is.
"One of the biggest mistakes organisations can make is assuming AI decisions are purely technical. They aren't. Every algorithm that influences hiring, promotions, development, or performance, is ultimately someone's career. Human judgment must remain central," said Prasad.
Why this matters for HR
HR leaders do not need to wait for perfect explainability before pressing forward. Start with the observable signals of broken trust. When employees stop asking questions about why a decision was made, or revert to manual workarounds, the system has failed. Your manoeuvre being visible. Run the tests: whether an AI recommendation stays valid when employees are given a one-click challenge option, track override rates over several weeks, and publish consequences for decisions changes. Directional patterns matter more than perfect metrics, and they will be defined by how you as leadership respond, not by how you deployed the algorithm.
For HR the takeaway is straightforward. The next generation of leadership is not asking whether algorithms work. It is asking who owns their decisions, who can challenge them, and who answers for outcomes. That answer lives in behaviour, not just design.
Explore the AI Learning Path for VPs of Strategy for frameworks on leading AI adoption across functions.
Your membership also unlocks:
