Trust at work is fraying. Edelman’s 2026 Trust at Work report found trust in employers fell five points since 2024, and trust in direct managers and co-workers each hit record lows. That same research describes a broader shift in which people are growing reluctant to trust anyone different from them, which makes the workplace one of the few places left where people across differences still have to rely on each other.
AI is adding fuel. Challenger, Gray & Christmas counted 101,743 job cuts attributed to AI in the first half of 2026, nearly double the 2025 total, and in a July survey of 900 U.S. workers, 40% said they expect leadership to downplay AI’s impact on jobs and another 31% expect leadership to say nothing until employees feel the effects, while only 9% expect full transparency.
When we’re trying to solve problems like these, many of us head to frameworks. There are several that address the issue of trust. Deloitte’s Four Factors of Trust (humanity, transparency, capability, and reliability), Charles Feltman’s care, sincerity, reliability, and competence, David Maister’s Trust Equation, and Frances Frei and Anne Morriss’s trust triangle are all solid options, but we think they’re lacking.
Each of these models measures how trustworthy the organization seems, but almost none of them ask what it costs each person to extend that trust in the first place.
Trust means accepting risk, and risk is unevenly distributed
When you trust someone at work, you’re taking a chance that they’ll look out for you. Think about what that looks like day to day, from flagging a mistake your manager made, to telling HR about a colleague’s behavior, to admitting you don’t understand the new AI tool everyone else seems to be using. Each of these moments asks you to put something on the line, whether that’s your reputation, your relationships, or your job, which is a big part of why so many employees stay quiet.
When an employee raises a concern about a new AI tool, they’re taking a bigger risk than the VP sitting next to them, and women of color are often penalized for speaking up in ways their white colleagues aren’t. Class shapes the experience too, with 97% of people from working-class backgrounds saying their class affected their work experience. Edelman’s data shows the same pattern at a societal scale, with the U.S. now showing the largest income-based trust gap at 29 points.
A single organizational trust score averages all of these experiences together, which means the people carrying the most risk are often the least visible in the results.
Why this matters for change and AI
Every change initiative asks people to extend trust before the results are in, and that request can impact people on your team differently depending on who has the most to lose. AI adoption makes this especially sharp, since 54% of low-income employees fear being left behind as AI becomes more integrated into their roles. If your readiness assessment shows strong trust overall, you may want to consider asking whose trust is holding up that number and whose is missing from it.
When trust breaks, the impact is also uneven. Psychologist Jennifer Freyd describes institutional betrayal as what happens when an institution you trust or depend on mistreats you, and her proposed antidote, institutional courage, focuses on transparency, acknowledging wrongdoing, and apologizing, which turns out to be a solid change management practice too.
How power impacts trust
Adding a power-aware approach by asking who is being expected to trust, what happens to them if that trust is misplaced, and whether the organization trusts its people as much as it asks to be trusted in return, can deepen your connection to your team and create a more supportive, healthy, and productive environment.
Takeaways and action items
- For individual contributors, notice where your own trust is easy and where it feels costly, since that difference is useful information to bring to your manager or an employee resource group.
- For team leads, before asking your team to embrace a change, name what each person is risking and what support you’ll offer if things go sideways.
- For executives and HR, disaggregate trust and engagement data by role, level, and identity, because an averaged score hides the people with the most at stake.
- For survey design, separate questions about kindness from questions about fairness, so people who experience a warm manager and an unfair promotion process can tell you both.
- For change and AI rollouts, measure trust in both directions by asking whether leadership is extending autonomy, information, and decision-making power to the people doing the work.
- For repair, practice institutional courage by acknowledging harm openly and quickly, especially when it falls on people with less power to absorb it.