IMD - International Institute for Management Development, Switzerland
Developing the skill of conducting impactful conversations is crucial for leading oneself and others. However, not all executives have access to leadership coaches, resulting in untapped potential. Given the capability of large language models (LLMs) to engage in "conversations" with humans, we explored whether leaders could utilize LLM-powered tools to enhance their ability to lead conversations. Working with executives from Africa, Asia and Europe, we explored two avenues to observe whether LLM-powered tools could generate insights that are novel and/ or useful for users by analysing their conversations. Firstly, we programmed an app, based on a GPT-4 version, including voice fingerprint recognition. For more nuanced analysis, we fine-tuned it with a well-established coaching framework. Secondly, we tested the openly available version GPT 4o. It lacked voice fingerprint recognition, which we mitigated by an analytical voice identification protocol. In both set ups, human coaches observed and documented conversational effectiveness, dynamics and tone. As a result, both approaches recognized different coaching styles correctly and gave useful recommendations to improving conversational effectiveness. To our surprise, our programmed tool also identified alternative coaching strategies. We created a framework measuring how Novel and how Useful executives found the experience. About 30% found the observations of the tool useful but not novel About 50% found the feedback novel and useful About 10% found the AI tool’s insights novel but unhelpful For 5%, it was neither helpful nor novel. Our findings suggest AI-tools are useful conversational sparring partners.
