Logo Contxto

AI workplace advice raises questions about employee trust and the future of team collaboration

Contxto

Por Contxto

November 17, 2026

Workers already ask chatbots what to do about a difficult boss or a teammate who keeps dropping the ball. Companies, meanwhile, are weaving AI into workflows, employee support and decision-making. The two trends meet in a place few leaders are watching: the advice itself, and what it does to relationships.

Cloverleaf Labs put five leading large language models (LLMs) through a series of workplace conflicts. Across 75 conversations, the models produced 638 distinct pieces of advice. Just three encouraged the employee to genuinely invest in repairing the relationship.

The number comes with a caveat. Each model heard only one side of the story, and plenty of human advisers would also take the complainer's side. Even so, the uniformity stands out.

Kirsten Moorefield

For Kirsten Moorefield, co-founder and Chief Strategy Officer of Cloverleaf, it was the real shock.

"I really couldn't believe that every single LLM said you should build up your exit ramp, start thinking about quitting, find a new team," Moorefield said.

That doesn't mean a resignation wave is coming.

"I don't think that's the accurate headline," she explained.

AI may reinforce workplace disengagement

The risk is slower and harder to see. McKinsey research cited by Moorefield indicates that employees who are power users of AI are 10% more likely to leave their jobs within the next three to six months. Stronger AI skills also mean more marketable skills, which may explain part of that.

The bigger danger lies in how people feel about change and about each other.

"I think actually, what you're going to see is a continued drop in all the things that everyone says we need right now. Like, really leaning in and having agility and embracing new tools. You're going to see an increase in disengagement and fear," Moorefield said.

Managers look worse still. By Moorefield's count, one in three was engaged a few years ago; now four in five are disengaged.

That's the part executives should take seriously. A chatbot that nudges one frustrated employee toward the door is a small problem. A habit of asking software to validate every grievance is a much bigger one, because it trains people to look for an exit before they look for a conversation.

Fear of replacement adds to the unease. As AI moves deeper into workplace decisions, some employees will wonder: "Am I training this tool to take my job?"

None of it shows up quickly in standard performance metrics. Moorefield calls it a "silent disease" that erodes collaboration from the inside.

How personality data could improve AI workplace advice

Moorefield's example makes the case. Picture a highly analytical manager who processes information externally and thinks out loud, paired with an employee who is less detail-oriented and more focused on the future and the big picture.

With no context, an AI might read the employee's complaint as proof of a toxic relationship and suggest finding a place where they feel more appreciated. With personality information, it could read the manager's behavior differently.

"If the AI has the right data, it responds in a way that actually builds your curiosity about the other person," Moorefield explained.

That's the approach Cloverleaf Labs is exploring, combining personality science with AI-powered workplace coaching. Instead of validating frustration right away, the AI explains communication styles, nudges the employee to consider the other person's view and suggests ways to raise the disagreement.

When Cloverleaf analyzed the 638 recommendations, it set four conditions for relational repair. Advice had to encourage genuine curiosity about the other person, consideration of one's own behavior, communication of personal needs and openness to feedback.

"That only happened three times out of 638 pieces of advice," Moorefield said.

According to the findings, all three came up when a manager was the one raising the complaint about a direct report.

Two things deserve weight. Cloverleaf sells coaching built on personality data, and its study concludes that AI improves with personality data. That doesn't make the findings wrong, but independent replication would carry more. And 75 conversations with five models is a small sample, so the study works better as a warning flag than as a verdict.

The takeaway isn't that employees should stop asking AI for advice. A general-purpose model hearing one side tends to side with the person typing. Companies that deploy AI coaching should test for that before their people rely on it, and employees should treat the answer as one perspective, not a ruling.

Moorefield's conclusion is shorter.

"AI needs to be informed with the right data," Moorefield said.

Keep up to Date with Latin American VC and Startups News!