Few things in a manager’s job are more difficult than telling an employee that they’re underperforming, that they’re not getting the raise they expected, or that they’re being fired from their job. It seems like unpleasant conversations at work will never go away. But new AI coaching tools can help managers deliver tough messages, especially in the run-up to those awkward conversations.
Indeed, delivering tough news is an area where many managers, especially those early in their careers, need help. A July survey by The Predictive Index, the company behind role-playing, coaching, and Obi, a text-based chatbot that helps managers prepare for difficult conversations, found that while 74% of CEOs and business leaders said they were “very confident” that their managers could handle difficult conversations with people on their own, 42% of managers said they knew what they wanted to say but struggled with how to say it.
Furthermore, only 20% of managers say they prefer to prepare for difficult conversations on their own, without using a framework, coaching guide, or input from HR. According to the study, managers’ biggest blind spots include providing constructive criticism, anticipating employee reactions, and remaining objective.
“Spontaneity is not the goal of high-stakes work conversations,” says Emily DeJew, professor of business communication at Carnegie Mellon University’s Tepper School of Business. “We want to know if executives are really thoughtfully preparing for important conversations.”
There are a variety of free and paid tools on the market, and their functionality may vary. For example, some offer text-only interaction, while others are voice-enabled. Some incorporate specific employee data, such as personality profiles, to better prepare the AI, while others are more general. These tools can provide managers with instructions to improve their tone, delivery, and messages. Regardless of the tool, the goal is not to automate HR processes, but to allow managers to practice delivering tough messages and simulate how people would react in certain scenarios. It’s better than talking in front of a mirror, Dejou said, because you can actually get feedback.
Role-play tense office conversations before the real thing
Managers can run through multiple scenarios to prepare for difficult conversations. In the first round, you might ask the AI to simulate an employee who is angry at the manager’s guidance. Next, you might ask the AI to simulate an emotional employee. “You never know what will happen in real life, so it’s beneficial to practice multiple scenarios,” says Karan Kashyap, co-founder and CEO of Posh, which provides AI role-playing training tools for businesses.
Managers often find that their overall communication improves with these drills. “It’s like batting practice,” said John Morgan, president of LHH, a human resources consulting firm that provides AI coaching tools through its subsidiary. The only way to get better is to practice, but in the real world opportunities are limited.
Another benefit is that you have 24/7 access to the tool, so if you want to practice at midnight on a Sunday, you can, but a human coach can’t always accommodate your every whim. “Most companies don’t hire 500 coaches for every 500 managers,” says Casey Schaefer, founder of Definitions Coaching and Consulting.
AI works best when it has certain parameters and context. For example, you can tell an AI tool to role-play a scenario in which you need to give feedback to a sensitive employee who is chronically late. Managers can also instruct the AI to score employees on specific scales, provide critical feedback on messages, and flag when topics move into risky areas from a legal perspective, said Megan Verbier, Xactly’s chief human resources officer. Penny, a text and voice-based AI tool, provides managers with customized scripts and real-time coaching guidance, said Megan Barbier, chief human resources officer at Xactly. Managers can also ask for feedback on tone, tell the AI tool that they’re trying to be more direct, and ask for constructive criticism about whether their language is concise, Barbier said.
Role-playing is great, but the real lessons come from feedback, says Anthony Belluccia, senior product scientist at The Predictive Index. Chatbots provide feedback in a constructive and customized way, helping managers understand what they can do better, Belluccia said.
Without this feedback, many managers have no idea how others perceive their tone and word choices. Dejou gave the example of a manager who heard bad news and instead of hammering home the message, started a difficult conversation. He didn’t realize that his outspokenness could make employees uncomfortable until the AI pointed it out to him. AI advised managers to start a conversation by discussing the importance of employees’ contributions and their value to the company before settling on negative news. “It taught him how to reframe the conversation,” Dejeu said.
AI can also point out when managers talk too much, repeat themselves, or frequently use filler words like “um” and “so” that can interrupt the flow of the conversation. “If you can speak clearly and effectively, it’s a gift to others. It takes a lot of effort to get there,” Dejeu said.
Of course, it is also possible for two people to play the same game. Some employees may not realize they’re facing a difficult conversation with their manager, but if you know things are about to get tense, preparing in advance using free tools like Claude or ChatGPT can help you get a more thoughtful, clear, and effective response.
The risks of using AI in the wrong way at work
It’s important to avoid letting the AI tell you what you want to hear. AI tools are designed to give consent, but that can be a problem when constructive criticism is needed, Dejou said. If you’re using it for role-playing and the AI is too friendly, tell it to stop being so friendly and be more critical and direct, she said.
Additionally, HR experts say it’s generally recommended not to enter personal information into AI tools due to potential legal and privacy implications. Instead, managers need to explain actions and anecdotes so that the AI can better adjust the scenario and perform a more realistic role. “You want to avoid anything that includes personally identifiable information,” Schaefer said.
Managers should be especially careful when entering employee names, performance details, and other sensitive information into public AI tools. Employer-approved enterprise systems may have different privacy controls and permitted uses, so administrators should be familiar with those controls and how to be fully compliant. Administrators also need to be aware of how organizational and employee behavioral data is used within the enterprise, and whether personal chats are visible to other users within the organization. Additionally, you should consider the vendor’s data retention, access, and AI model training policies.
Even when using AI, it’s important to include a human coach. According to Morgan, managers sometimes need a completely different perspective that only they can provide, especially when the situation is very complex.
Schaefer recommends that managers first practice with an AI coach and then debrief with their boss or a member of their company’s human resources team to ensure they haven’t missed any key points, messages, or laws. It’s also another opportunity to practice communicating your message to someone, she said, because looking into someone’s eyes is a different experience than looking at a computer screen. “The emotional weight of a termination conversation is nearly impossible to recreate.”
