Hallucinating Partners: When Human Error Becomes an AI Problem

In this episode, we discuss the growing convenience of blaming artificial intelligence for things senior partners have done themselves. Joining us is Felicity Ranscombe, Director of Partner Reputation and Narrative Management at Pemberton Rook, following the firm's retreat scandal in which Felix Dacre's AI assistant, Nearly, accurately transcribed a conversation he believed he was having with the balance sheet.

We explore how an embarrassing human incident became an AI-governance emergency, why a correct transcript can be described as a hallucination, and what a crisis-communications team does when the technology's logs are rather more reliable than the partner's recollection. Felicity explains the firm's commitment to human oversight, although the question of who was overseeing the human remains under review.

Guest: Felicity Ranscombe, Director of Partner Reputation and Narrative Management, Pemberton Rook