While few managerial duties are as challenging as delivering negative performance feedback, denying a raise, or announcing layoffs, new artificial intelligence coaching tools are emerging to assist leaders in preparing for these high-stakes interactions. According to experts, discomforting workplace dialogues are not disappearing, but technology can now help managers refine their approach before sitting down with an employee.
A July survey conducted by The Predictive Index, which provides a text-based chatbot named Obi for role-playing and coaching, highlighted a significant confidence gap. The study found that 74% of CEOs and business leaders believe their managers are “very confident” in handling tough conversations independently. However, only 42% of managers agreed they could effectively deliver their message, despite knowing what they wanted to say. Furthermore, just 20% of managers prefer to prepare for such discussions without a structured framework, coaching guide, or HR input. Common blind spots identified include delivering constructive criticism, anticipating employee reactions, and maintaining objectivity.
Emily DeJeu, a professor of business communication at Carnegie Mellon University’s Tepper School of Business, emphasized that spontaneity is not the goal in critical work discussions. “You want to know the executive has really thoughtfully prepared for an important conversation,” DeJeu said. She noted that while practicing in front of a mirror is common, AI tools offer superior feedback by simulating realistic employee reactions.
The market now offers a variety of free and paid AI solutions, ranging from text-only interfaces to voice-enabled platforms. Some tools integrate personality profiles to tailor scenarios, while others provide generic coaching. Karan Kashyap, co-founder and CEO of Posh, which offers AI role-play training, explained that managers can run multiple simulations, such as an angry or emotional employee response, to better handle unpredictable real-life outcomes.
John Morgan, president at LHH, a human resources consulting firm, compared the practice to batting practice. “The only way you’re going to get better is practice, but you only get so many chances in the real world,” Morgan said. He also highlighted the accessibility of these tools, noting that AI coaches are available 24/7, unlike human coaches who cannot be summoned on short notice.
Megan Barbier, chief human resources officer at Xactly, whose tool Penny offers tailored scripts and real-time guidance, stated that AI performs best when given specific parameters. Managers can request feedback on tone, conciseness, or legal risk flags. Anthony Belluccia, a senior product scientist at The Predictive Index, added that the primary value lies in the feedback, which helps managers understand how their word choice or tone might be perceived. DeJeu cited an example where a manager was unaware his blunt delivery was off-putting until AI suggested he begin by affirming the employee’s value before introducing negative news.
Despite the benefits, experts warn of potential risks. DeJeu cautioned that AI tools are designed to be agreeable, which may lead to insufficiently critical feedback. She advised users to explicitly instruct the AI to be more direct and critical if the responses feel too soft. Additionally, privacy concerns remain paramount. HR professionals recommend that managers avoid entering identifiable personal details, names, or confidential performance data into public AI platforms. Instead, they should describe behaviors or anonymized anecdotes to maintain compliance with corporate policies and data retention rules.
Morgan stressed the importance of retaining a human element, suggesting that AI practice should be followed by a debrief with a supervisor or HR representative. This ensures that no legal or messaging nuances are missed and allows managers to practice interpersonal dynamics that screens cannot replicate. “The emotional weight of something like a termination conversation is nearly impossible to replicate” digitally, Schaffer said.
Interesting how only 42% feel confident despite being told they are. This gap in self-perception is probably the real problem here.
The privacy warnings are valid though. Don’t feed real employee data into these tools without checking your company policy first.
Finally, a practical solution for managers who dread these talks. Accessible 24/7 practice beats awkward mirror rehearsals every time.
I love the idea, but can we really simulate empathy with an algorithm? I worry we are losing essential human connection.