Managers are using AI for one of work's toughest tasks: Difficult employee conversations

Few managerial duties are harder than telling an employee their performance is falling short, they won’t get the raise they expected or their job is being eliminated. Uncomfortable workplace conversations are unlikely to go away. But new AI coaching tools can help managers deliver hard messages, especially in the preparation phase before these sticky conversations take place.

To be sure, delivering hard news is an area where many managers need help — especially those early on in their careers. A July survey fielded by The Predictive Index, which offers a text-based chatbot, Obi, that assists with role-playing, coaching and preparing managers for difficult conversations, found that while 74% of CEOs and business leaders say their managers are “very confident” handling tough people conversations on their own, 42% of managers say they know what they want to say but struggle with how to say it.

What’s more, only 20% of managers say they prefer to prepare for a difficult conversation on their own, without a framework, a coaching guide or input from HR. The survey found that top blind spots for managers include delivering criticism constructively, anticipating an employee’s reaction and staying objective.

“Spontaneity is not a goal of a high-stakes work conversation,” said Emily DeJeu, a professor of business communication at The Tepper School of Business at Carnegie Mellon University. “You want to know the executive has really thoughtfully prepared for an important conversation.”

There are a variety of free and paid tools on the market, and features may vary. For instance, some offer text-only interactions, while others are voice-enabled. Some incorporate certain employee data such as personality profiles to help better prepare the AI, while others are more generic. These tools can offer instruction to improve the manager’s tone, delivery or message. Regardless of the tool, the goal is not to automate the people process but to allow the manager to practice delivering hard messages and simulate how a person might react in a specific scenario. It’s better than talking in front of the mirror, because you can actually get feedback, DeJeu said.

Role-playing tense office chats before real life

Managers can run multiple scenarios to prepare for hard conversations. In a first round, they might ask the AI to simulate an employee who reacts angrily to the manager’s guidance. They might then ask the AI to simulate an emotional employee. Practicing multiple scenarios can be useful “because you don’t know what real life will throw at you,” said Karan Kashyap, co-founder and chief executive of Posh, which offers an AI role-play training tool for businesses.

Managers often find their overall communication improves by doing these drills. “It’s like batting practice,” said John Morgan, president at LHH, a human resources consulting firm that offers an AI coaching tool through a subsidiary. The only way you’re going to get better is practice, but you only get so many chances in the real world. 

Another advantage is that the tools are accessible 24/7, so if you want to practice at midnight on a Sunday, you can, whereas a human coach isn’t always available on a whim. “Most companies aren’t hiring 500 coaches for 500 managers,” said Casey Schaffer, founder of Definitions Coaching and Consulting.

AI works best when it has specific parameters and context. For example, you can tell the AI tool you want to role-play a scenario in which you have to give feedback to a sensitive employee who is habitually late to work. Managers could also tell the AI to grade them on specific measures, offer critical feedback on their message or flag if they move into risky topics from a legal perspective, said Megan Barbier, chief human resources officer at Xactly, whose text- and voice-based AI tool Penny gives managers tailored scripts and real-time coaching guidance. Managers could also ask for feedback about tone, or tell the AI tool they are trying to be more direct and ask for constructive criticism on whether the language is concise, Barbier said. 

The role-playing alone is good, but the real lessons come from the feedback, said Anthony Belluccia, senior product scientist at The Predictive Index. Chatbots can deliver feedback in a constructive, tailored way to help managers understand what they can do better, Belluccia said.

Without this feedback, many managers don’t realize how others may perceive their tone or word choice. DeJeu offered the example of a manager who began difficult conversations with the bad news rather than easing into the message. Before AI pointed it out, he didn’t realize his blunt delivery could be off-putting to employees. AI advised the manager to kick off conversations by discussing the importance of the employee’s contributions and his value to the company before easing into the negative news. “It guided him on how to reframe the conversation,” DeJeu said.

AI can also point out when managers talk too much, repeat themselves or frequently use filler words such as “um” and “so” that can interrupt conversation flow. “It’s a gift to other people when you can speak clearly and effectively. It just takes a lot of work to get there,” DeJeu said.

Of course, two can play the same game. While some employees may not realize they’re headed into a difficult conversation with a manager, if they have an inkling that things are going to be tense, preparing in advance with a free tool such as Claude or ChatGPT can result in more thoughtful, articulated and effective responses.

Risks of using AI at work in the wrong way

It is critical to make sure managers do not let AI tell them what they want to hear. AI tools are designed to be agreeable, but that can be an issue when you want constructive criticism, DeJeu said. If you’re using it for role-playing and the AI is being too agreeable, tell it to stop being so agreeable and ask it to be more critical and direct, she said.

Also, because of the potential legal and privacy implications, it’s generally good practice not to enter personal details into AI tools, according to human resources professionals. Instead, managers should describe behaviors or anecdotes so AI can better size up the scenario and perform a more realistic role. “I would avoid anything that has any identifiable information in it,” said Schaffer.

Managers should be particularly cautious about entering employee names, performance details or other confidential information into public AI tools. Employer-approved enterprise systems may have different privacy controls and authorized uses, so managers should be familiar with them and how to be in full compliance. Managers should also be aware of how organizational and employee behavioral data is used inside the corporate enterprise and whether individual chats are visible to others in the organization. Additionally, a vendor’s data retention, access, and AI model training policy should be considered.

Even when using AI, it’s important to have a human coach in the mix. Sometimes managers need a totally different viewpoint that only a person can offer, especially if a situation is really complicated, according to Morgan.

Schaffer recommends managers start by practicing with an AI coach, and then work with a supervisor or member of the company’s human resources team to debrief and ensure no important points, messaging or legalities were missed. It’s also another opportunity to practice the message with a person since looking into someone’s eyes is a different experience than looking into a computer screen, she said. “The emotional weight of something like a termination conversation is nearly impossible to replicate.”