In many organisations, digital assistants now draft replies, reports and follow-up questions, often without the other side knowing. Recent analyses from software development reveal the scale: a large share of machine-written contributions is no longer reviewed by humans at all, and most review comments are themselves produced by automated systems. Organisational research adds an important finding. Recipients discount machine-generated messages, not because of lower quality, but because perceived authenticity is missing. From this we derive a clear practice: declare openly where automation is involved, delegate routine tasks deliberately, and keep sensitive conversations with people. In addition, defined checkpoints are needed where someone intervenes when systems negotiate with each other. Companies that follow this approach gain the time savings without putting years of hard-earned customer trust at risk.