As AI systems get better at holding a natural conversation, more of the interactions a business has with its customers are handled by software that sounds convincingly human. A chat window, a phone line, an email reply — increasingly the thing on the other end is a machine, and it is good enough that the person often cannot tell. That raises a question that used to be theoretical and is now practical and daily: should a business tell people when they are talking to an AI rather than a person? The answer is moving quickly from a matter of taste to a matter of expectation, and in a growing number of places, of law.
The case for disclosure rests on something simple: people feel differently about a conversation depending on who they think is in it, and discovering after the fact that they were misled tends to breach trust badly. Someone who believes they are speaking to a human may share things, or place a kind of confidence, that they would not with a machine; someone who suspects a machine but is not told is left uneasy and guarded. The harm of non-disclosure is less about the AI itself than about the deception — the sense, once the truth comes out, that the business let them believe something false to make the interaction go more smoothly. Regulators have begun to treat clear disclosure as a baseline requirement precisely because the deception, not the automation, is the problem.
What is striking is that disclosure, framed by many businesses as a cost or an awkward admission, often turns out to be an advantage. People are increasingly comfortable dealing with a capable automated system for routine matters — it is fast, available at any hour, and does not keep them on hold — and they mind far less that it is a machine than that they were not told. Being upfront lets a business set the right expectation, offer an easy path to a human when the machine reaches its limits, and build the kind of trust that comes from not pretending. The businesses that treat AI as something to hide tend to erode confidence when it is discovered; the ones that treat it as something to name tend to keep it.
The practical posture, then, is to make disclosure clear and early rather than buried or grudging: tell people plainly when they are interacting with an automated system, make it easy to reach a human when they want one, and do not design the AI to actively pretend to be a person. This is partly about staying on the right side of rules that increasingly require it, but it is more than compliance. It reflects a straightforward reality about the technology's place in a business: an AI that handles a conversation well is genuinely useful, and the way to make that usefulness durable is to be honest about what it is. The courtesy of saying so is quietly becoming the condition of being trusted at all.