When People and AI Break Trust the Same Way

Here’s the story I keep coming back to when I talk about trust.

Not too long ago, I got glowing feedback. I was told—bright lights, all smiles—that I was a “rock star.” Two days later, I heard from someone else that the same person said something very different in another meeting.
Not cruel. Just… inconsistent.

That tiny crack changed everything. I found myself checking my tone, documenting everything, asking for “clarity.”
It hit me: trust isn’t lost in a blow-up — it leaks out through avoidance and silence.

Then came Zoom.

After ten years of muscle memory, Zoom quietly swapped my email for a “Zmail,” tried to slurp my contacts, and—icing on the cake—left me negotiating with a chatbot that canceled my existing invitations as “help.” I didn’t authorize any of it. I was beyond frustrated. And yet… I stayed. Not because it earned me back, but because the switching cost felt higher than my anger that day. Zoom is now officially “on probation”: one more breach, and I’m out.

These two moments—one human, one software—are the same story with different characters: trust corrodes when words shift with the room, and when systems act without accountability. If you’re adding AI into your product, this matters more than your next model upgrade.

Users don’t need AI to be perfect. They need it to be verifiable and repairable. If I can see what happened, understand why, and fix it without begging a bot, I’ll forgive a lot. If I can’t, even “helpful” becomes hostile.

Behind the scenes: how this changed my audit

As a trust strategist, I used to focus my GenAI Trust Audit heavily on the front-stage: explanations, consent, friction in the right places. Those still matter. But after the “rock star” whiplash and my Zoom saga, I shifted weight to three backstage gaps:

• Transparency debt
  Your system technically tracks everything, but no one can actually understand it.
  Audit logs, update notes, and AI explanations are written for engineers, not humans — so users can’t tell what changed, why it changed, or how to verify it.
• Accountability theater
  Teams talk about “responsible AI” but rarely assign owners, timelines, or rollback rules.
  Without clear responsibility, even the best intentions become slogans.
• Repair deserts
  When something goes wrong, users can’t fix it themselves.
  They’re forced to submit tickets, wait days, or plead with a chatbot that makes things worse.
  If users can’t recover easily, trust dries up fast.

Now, when I run the audit, we pressure-test the product’s verification path (What can the user see and check?), the repair path (How fast can they recover without starting over?), and the ownership path (Who’s on the hook when it breaks?).

We also test the “conflict-avoidant colleague” risk: does the system say one thing in one context and another elsewhere? If so, you’ve built a product with a double narrative — the digital equivalent of saying whatever’s easiest in the moment. And nothing erodes trust faster than inconsistency disguised as helpfulness.