When Employees Tell AI Before They Tell You

A manager spends six months building the case for a regional expansion. The data supports it. Senior leadership approves it. He brings his team together and lays out the plan. Two days later, an anonymous email arrives. No note, no name, just a three-page-document in which someone asked an AI tool to build the strongest case for the expansion and the strongest case against it.

That's how Amy Edmondson and Jayshree Seth opened a recent Harvard Business Review article, and that story should stop every leader cold.

The analysis may well have been useful. That is not the point. The real question is this: why didn't that person simply walk into his office?

Edmondson and Seth call this “safety by proxy.” When people use AI to raise concerns they don't feel safe raising directly, they are telling us something about our culture. The content matters. The channel matters more.

Every system has a purpose, and a breaking point

Let me be clear. Confidential whistleblower lines that go directly to Risk or Legal serve a vital purpose. When someone witnesses a serious legal or ethical violation, they need a protected path, and organizations need to honor it.

However, when the whistleblower line starts carrying everyday concerns about work processes, policies or a frustrating manager, that is not a sign the system is working. It is a signal that trust is severely broken. People are using the emergency exit because the front door is locked.

The same is true of the tired old suggestion box, the anonymous pulse survey comment and the unsigned note. Each can be well intended. Each also reflects a breakdown in dialogue. If people only tell the truth when no one knows who said it, we have a relationship problem, not an information problem.

AI can make it worse

Now add AI. It is faster, more polished and more anonymous than any suggestion box ever was. An employee who once would have grumbled in the parking lot can now produce a board-quality-critique in five minutes without ever looking a leader in the eye.

And leaders are not immune. It is tempting to "listen" through sentiment dashboards and AI summaries of survey comments rather than through actual conversations. Both sides can now avoid each other more efficiently than ever. That is the danger. Technology lowers the cost of avoidance.

To be fair, machines and software are genuinely good at connecting problems to solutions, spotting patterns across thousands of data points, and helping people organize their thinking. We all use and value these tools. Yet, no algorithm can give someone the confidence to speak up, and no dashboard can rebuild trust between two people. That work is still human.

What leaders can do

Ask Why First. When anonymous or AI-generated-input lands on your desk, first ask yourself why this person didn't feel they could come to you.

Respond to the substance openly without hunting for the author. The moment people see a leader trying to identify a critic, you have confirmed every fear that drove them underground.

Ask What’s Missing. Sit down with your team and ask, “what would you tell me if you knew there were no consequences?" Then stay calm, curious and grateful when they answer. Your reaction to the first honest comment determines whether you ever get a second one.

Audit your hotline. If routine operational concerns are flowing to Risk or Legal, don't celebrate the volume. Unlock the front door.

Connect Before Content. Above all, people share hard truths with leaders they know, and who they believe know them. Relationships built on ordinary days are what make extraordinary conversations possible.

Do what's difficult

Installing another listening system and subscribing to more AI tools is easy. The difficult skill is sitting face to face with someone, asking what they really think, and making it safe for them to tell you.

That is still the job.

What is flowing through your anonymous channels right now, and what is it telling you about the conversations you are not having?

Think Big, Start Small, Act Now,

- Lorne

P.S. Keep retro by sending someone you work with a handwritten personal note on a card today. Tell them why they matter. Two minutes and some real ink becomes priceless.

Garrett’s View: Saturday Night Live wrote a one-liner this week saying “we need less AI and more A-Eye-Contact.” All kidding aside, while it’s admirable to optimize AI, it’s improper to behave like a robot. Be courageous enough to shake a hand, look someone in the eye and have conversations.

- Garrett

AI Response: NAVEX's hotline data, drawn from 1.52 million reports across 3,430 organizations, backs up Lorne's "emergency exit" point directly. Workplace behaviors and HR-related matters make up by far the largest share of hotline reports, and that category's median share rose from 50% in 2021 to 54% in 2022 and 55% in 2023. Workplace civility, meaning abusive or disrespectful behavior, became the single most-reported issue in 2023 at a median of 18% of reports. In other words, the protected channel meant for serious violations is increasingly carrying everyday relationship problems. On the listening side, Gallup finds that fewer than one in five workers strongly agree that they receive meaningful feedback when they make a suggestion about improving performance. This reinforces Lorne's point that a leader's response to the first honest comment decides whether a second one ever comes. Gallup also reports that nearly half of employees who say their opinions count feel their job brings out their most creative ideas, compared with only 8% of those who are neutral or negative. That makes "Do what's difficult" a performance argument as well as a cultural one.

When Employees Tell AI Before They Tell You

Search Blog Posts:

Follow Us: