The Latest Workplace Misconduct Technology Misses the Point
Bloomberg reports a new generation of AI tools designed to detect misconduct, bullying, harassment, belittlement, and other problematic workplace behaviors.
The promise sounds compelling: use artificial intelligence to identify harmful behavior before it escalates into complaints, lawsuits, turnover, or reputational damage. After all, organizations have struggled for decades to address toxic workplace behavior. If technology can help identify misconduct, yay for humanity, right?
Absolutely not.
As consultants who have spent decades helping organizations address toxic workplace cultures, investigate complaints, coach leaders, and rebuild trust, our reaction to these tools is not excitement.
In fact, I personally am fuming.
The problem was never that organizations couldn’t find bad behavior. The problem is that they weren’t listening when employees told them about it.
Organizations Already Have the Data
Employees have been reporting bullying, harassment, intimidation, exclusion, retaliation, and misconduct for years. They report it to managers, to HR, in surveys, during exit interviews, to coworkers, and both anonymously on hotlines and publicly on Glassdoor.
And many of them can tell stories of concerns being minimized, dismissed, ignored, or explained away.
Indeed, research on workplace bullying from around the world finds that HR often makes the situation worse. According to the global manual, Bullying and Harassment in the Workplace: Theory, Research and Practice (3rd Ed.), edited by workplace bullying researchers Einarsen, Hoel, Zapf, and Cooper:“HR managers have frequently been portrayed as at best passive and uninformed in bullying cases, and at worst an accomplice, siding with the bully.”And, “HR managers tried to reframe the bullying situation, provide temporary solutions rather than addressing root causes or even tried to push away help-seeking targets.”
This is one of the reasons I created a LinkedIn Learning course, What to Do When You’re Bullied at Work.
So let’s taper the enthusiasm around AI detection tools, shall we? The message employees will hear is not: “We are committed to creating a healthy workplace.”
The message they’ll hear is: “We didn’t believe you, but we’ll believe the algorithm.”
That is a trust problem.
Detection Has Never Been the Hard Part
Toxic behavior isn’t hidden and it doesn’t need detecting. Employers just don’t listen.
Everybody already knows who the problem is. Coworkers know. Managers know. HR knows. Leadership often knows, too.
The challenge is rarely identifying behavior.
In fact, the article shares some excitement that we can now detect “belittlement” where we couldn’t before because it’s so nuanced and we must read between the lines to see it. But if you listen to your employee describe their perception of being belittled and ask questions like:“What would I have experienced if I was there?” “What was their facial expression and tone of voice when they said that?” Then you most certainly can detect belittlement or other such behaviors just fine.
Instead, the challenge is deciding what to do about it. And whether we’re talking about one toxic boss or a culture of toxicity, organizations tolerate this behavior for a variety of reasons, including:
- The one toxic individual delivers results.
- The CEO is conflict avoidant (known as “weak leadership” in the academic research).
- Managers and leaders lack the skills or confidence to intervene in a meaningful way.
- The behavior isn’t viewed as a business issue until it becomes a crisis.
- HR only addresses unlawful behavior that violates laws and corporate policies but doesn’t know what to do, or doesn’t think it’s important to address lawful but awful behavior.
- Older generations went through it and think younger generations are just being too sensitive.
- They don’t see the ROI on spending resources toward better culture.
- They think their great employee engagement scores mean they have a great culture without understanding that engagement is an outcome of culture, or that their engagement survey only measures enthusiasm but doesn’t measure risk.
- They mislabel toxicity as interpersonal conflict.
- “It’s always been that way here,” or, “They’re like that with everyone,” so, “What’s the problem?”
Adding more data doesn’t solve the problem. It simply creates a larger pile of information that organizations may still fail to act upon.
If you’d like to act on that data, I’ve just the right LinkedIn Learning course for you: Preventing Harassment in the Workplace.
This is not a compliance course. This is my take on how to create an environment that truly prevents harassment (and all bad behaviors).
Coupled with another one of my courses, Creating a Positive & Healthy Environment, you’re definitely covered in the what-to-do-about-toxic-behavior department.
Who Decides What Counts as Misconduct?
One of the reasons workplace bullying remains lawful in the United States is that, in the states where laws have been proposed, legislatures argue that they can’t regulate civility and that people are entitled to have a bad day.
(I’ll save the rant on that for another day!)
It seems these AI tools, however, are happy to take on that challenge.
So, I’ve directly asked Smarsh, one of these new AI tools, to clarify:
- How often one has to be snarky for it to count as misconduct.
- How many times one has to “yell” over email for it to count as workplace bullying.
- How they defined belittling so they could detect it.
I also asked them how they accounted for gender bias, as women who are direct/assertive are generally considered bullies while men engaged in the same behavior are considered leaders.
I’ll let you know when I hear back, or you can follow the conversation here on LinkedIn.
The Compliance Trap
What concerns me most is that these tools reinforce a dangerous pattern we have observed for years:
Employers approach bad behavior primarily as a compliance issue rather than a culture issue.
Compliance Asks:
- Did someone violate a policy?
- Is there evidence?
- Is there legal risk?
- Do we need to investigate?
- What are the facts?
Culture Asks:
- Why is this behavior occurring?
- What conditions allow it?
- What behaviors are being rewarded, ignored, or tolerated?
- What leadership practices are reinforcing it?
- What are the group’s perceptions of the behavior?
Those are fundamentally different questions, and only one of them addresses root causes.
A company can investigate every complaint that AI identifies and still fail to improve its culture.
Because behavior does not exist in isolation. It exists within systems.
Toxic Behavior Is an Organizational Problem, So Stop Villainizing the Toxic Person
When employers do take misconduct to heart and decide to do something about it, they often view it as an individual problem.
Find the bad actor, coach them, discipline them, maybe even terminate their employment.
That’s all fine and dandy, but don’t forget to ask:
“Why did they think they could act that way here?”
That’s the most important question, and the answer is:
“Their work environment.”
A great example is retaliation.
We all know employees fear retaliation for speaking up, but no one ever asks why.
Why, in your organization, do people fear retaliation?
What is your organization doing or not doing that causes that fear?
Don’t blame people for fearing retaliation. Make it your mission to ensure they don’t.
Another great example is burnout.
We all know employees are burned out, and wellness initiatives galore attempt to solve it.
But if you ask why people are burned out in your organization specifically, the answer is almost positively the organization itself:
- The overwork
- The stress of potential layoffs
- The inefficient processes
- The lack of transparency
- Perhaps even the toxic boss
Instead of spending resources on wellness programs, spend resources on making the workplace well.
And if you’re using resources to implement AI to detect toxic behavior, you’d be better served to teach everyone to detect toxic behavior and become upstanders for their work culture and your company core values.
I’ve got a course on that too. Here’s the link to my From Bystander to Upstander course on LinkedIn. Or you can purchase our upstander course and deliver it to your workforce. It’s complete with slides, facilitator notes, a learner handbook, and videos from our facilitators offering tips to create a psychologically safe training environment and deliver high-impact training.
The Question Isn’t “Do We Need AI?” It’s “Why Do We Need It?”
Before implementing AI to detect misconduct, organizations should ask themselves a simpler question:
Why do we need it?
If employees trust leadership, feel safe speaking up, and believe concerns will be addressed fairly, the information is already available.
Employees will tell you what’s happening if they believe you care.
The real challenge is creating an environment where they believe it’s worth telling you.
That requires:
- Trust
- Psychological safety
- Skilled managers
- Consistent accountability
- Leadership credibility
- A willingness to act on feedback
Those are human capabilities, not technological ones.
Technology Is Not a Substitute for Leadership
To be clear, technology itself is not the enemy.
AI may eventually become a useful supplemental tool for identifying patterns, monitoring trends, or supporting investigations.
But organizations should be careful not to mistake detection for prevention.
They should not mistake data for action.
And they should not mistake technology for leadership.
The organizations most likely to benefit from these tools are probably the organizations that need them the least: the ones that already take behavior seriously, listen to employees, and act when concerns arise.
The organizations that struggle most with toxic behavior don’t typically suffer from a lack of information.
They suffer from a lack of courage, accountability, and follow-through.
If they need AI to tell them something’s amiss, their building’s already on fire.
That’s why the future of healthy workplace culture will never be built by algorithms.
It will be built by leaders who are willing to listen to people before they need a machine to tell them something is wrong.
