Trust in AI is low. That makes it a management skill.

In short: fewer than half of people trust AI, and many rely on its output without checking it. Trust at work follows what managers do: using AI openly, checking it visibly, being honest about jobs and making it safe to report mistakes. That makes building calibrated trust, neither blind nor dismissive, a management skill.^1,2,3

The 2025 Edelman Trust Barometer found that 49 percent of people worldwide trusted AI. The figure was 32 percent in the United States and 72 percent in China.¹ Stanford’s 2026 AI Index found a 50-point gap between experts and the public on whether AI will improve how people do their jobs: 73 percent of experts expected a positive effect, against 23 percent of the public.⁴ Pew found in 2025 that 52 percent of US workers were worried about AI’s future impact on their workplace.⁵

The second row is the uncomfortable one. Low trust and uncritical reliance exist side by side: people distrust AI in general and over-trust the answer in front of them. The goal is not more trust. It is calibrated trust, where people know when to rely on a tool and when to check it.

Your colleagues, employees and customers are that public. For anyone leading AI adoption, low trust is not a communications problem to solve once. It is a condition to manage every day.

What builds trust at work

Microsoft’s 2026 Work Trend Index reported that where managers actively modeled AI use, employees reported a 30-point lift in trust in agentic AI and a 17-point lift in the value they got from it. Those figures come from a separate survey of 1,800 people and show an association, not a cause, but they fit what most managers observe.² People trust tools their managers use openly, check visibly and talk about honestly.

• Use the tools yourself, in the open. Show how you check the output.
• Be honest about jobs. Say what you know and what you do not. Silence is heard as bad news.
• Make it safe to report errors. If admitting an AI mistake is embarrassing, mistakes go unreported until a customer finds them.
• Keep a human route. For customers and employees alike, a way to reach a person is part of a trustworthy design.
• Say what is not automated. Clear lines about what AI will not decide are as reassuring as any promise about what it will do.

The checking habit is the same one a golden set builds into a system: test before relying, and keep testing after changes. At the level of a team, it is what our briefing on hiring for AI judgment called owning the result.

What erodes it fastest

Three things undo trust quickly. Announcing AI with a headcount message first, as Duolingo learned, makes every later tool look like a threat. Hiding AI from customers who later find out turns a reasonable choice into a deception. And punishing people for errors they report teaches everyone to stop reporting. Each is a management decision, which is why trust sits with managers rather than with the technology team.

Why this is a career skill

Technical skill with AI tools is spreading quickly. The rarer skill is getting a team to use them well: willingly, carefully and with honest reporting of what does not work. Managers who can do that will be the ones organizations trust with larger AI programs. The practice starts small: at your next team meeting, show one thing an AI tool did well for you this month, and one thing you caught it getting wrong.

Culture, trust and communication are the first lesson of Module 12 of the AIMP program. The 2026 Work Trend Index suggests the return on that lesson is larger than most technical training: the managers who model careful use are the ones whose teams get value from it.