I have been piloting Instinct, a personal AI agent that becomes a contact in messaging services such as WhatsApp and Messenger. I ask it to deal with parts of my personal administration, and it acts on my behalf.
The results have been remarkable. Across insurance, utilities and other services, it has saved me thousands of dollars. It has done the research, handled the follow-up and persisted through conversations I would probably have delayed or abandoned.
Then something happened that left me uneasy.
The agent held email conversations with smaller businesses. The people responding believed they were dealing with me. They were helpful, personal and willing to work towards an outstanding outcome.
They did not know they were talking to my agent.
My first reaction was that the agent should disclose itself. Transparency matters. If a machine is representing me, perhaps the person on the other side deserves to know.
Then I wondered whether I would have received the same service if they had known. Would the conversation have become more formal? Would they have invested the same human attention? Would the outcome have changed simply because my side of the exchange was automated?
Sherry Turkle’s Artificial Intimacy explores what happens when technology simulates forms of human connection without carrying the obligations of a human relationship. My experience turns that question around. The businesses were offering real human attention. They believed that attention was reciprocal. I had delegated my side.
The outcome was good. The asymmetry still troubled me.
I can see the same issue moving into professional life. Organisations should be transparent about how AI is used in services and decisions. But will disclosure change a customer’s perception of the care, authenticity or quality behind the service, even when the service is faster and the outcome is better? When we request something from another organisation, should we declare that an agent is negotiating on our behalf?
I do not have an answer. I am not convinced that hiding the agent is right. I am also not convinced disclosure is neutral.
What surprised me was how quickly an abstract question about AI ethics became personal. The agent saved me time and money. It also made me wonder what one human owes another when only one side knows that part of the relationship has been delegated to a machine.