Agency, Autonomy, and the AI That Speaks for You
A woman in her seventies was trying to cancel a subscription. I was sitting two seats away in a coffee shop, not eavesdropping — it was impossible not to hear. She'd been on hold for twenty minutes. When she finally got through, the agent asked her to verify her identity. She gave her name. Then the account number — reading it slowly off a piece of paper she'd brought with her, the way you'd read a phone number to someone writing it down. Then a postcode. Then the last four digits of a card she wasn't sure she still had.
She got through. She explained what she wanted. The agent said she'd need to be transferred. Hold music again.
She hung up.
I sat there thinking: this is what the absence of agency looks like. Not cruelty. Not incompetence. Just a system that was never designed to give the customer any power at all.
The word we keep misusing
"Agent" gets thrown around constantly now. AI agents, customer agents, autonomous agents, agentic workflows. It's become one of those terms that means everything and nothing. But the word has an older, more precise meaning — one that matters here.
Agency is the capacity to act on your own behalf. To make choices. To have those choices respected. It's not about speed or efficiency. It's about whether you're a participant in the interaction or a subject of it.
Most customer service, even the expensive kind, treats the customer as a subject. You enter their system. You follow their process. You wait when they tell you to wait. You verify yourself on their terms, repeat your problem to their schedule, and accept their resolution or escalate through their hierarchy.
The customer has almost no agency in this exchange. Never has.
Now picture the reverse
The customer has their own AI. Not a chatbot — an agent that acts on their behalf, with their preferences, their history, their authority. When they need to deal with a company, they don't call. They don't wait. They don't navigate an IVR or retype their address for the third time. Their agent contacts the company's system directly.
That woman in the coffee shop? Her AI would have had the subscription details, the account number, the card. It would have called, navigated the menu, verified, requested the cancellation, confirmed, and sent her a message: Done. Cancellation confirmed. Refund in 3–5 days.
She would have been drinking her coffee.
This is what customer agency actually looks like. Not a better chatbot on the company's side. Not a smoother IVR. The customer having their own capability — their own representative — that operates with genuine autonomy.
The social contract shifts
Here's where it gets interesting, and where most of the industry conversation goes missing.
When both sides have AI, the relationship between company and customer changes at a structural level. The social contract — the implicit deal about how we interact, what's expected, what's acceptable — has to be renegotiated.
Think about trust. Today, companies verify the customer's identity. That's a one-way gate. But when the customer's AI is calling, who verifies the company? How does the customer's agent know it's actually talking to the real service line and not a phishing operation? Trust becomes bidirectional, or it collapses.
Think about data. Today, the company holds all the interaction history. The customer starts from zero every time. With a personal AI, the customer holds their own history. They know what was promised last Tuesday. They have the transcript. The information asymmetry that underpins most service interactions — the company knows more about the relationship than the customer does — evaporates.
Think about power. Today, the company sets the terms: operating hours, channel availability, resolution authority. A customer's AI doesn't care about operating hours. It'll try at 3am. It'll try a different channel. It'll try again tomorrow. The company can no longer control the tempo of the interaction.
The real product
Most CX leaders I talk to are focused on building the company's AI. Better bots. Smarter routing. More deflection. Reasonable priorities, given where budgets sit. But the disruption isn't coming from inside the company. It's coming from the customer's side.
When your customer sends their agent to deal with your agent, the question isn't "how do we contain this?" The question is "what kind of relationship do we want?"
Because the customer now has a choice they've never had before. They can send their AI to deal with you patiently. Or they can send it to find a competitor who's easier to deal with. The switching cost just dropped to near zero — not because of price comparison, but because the friction of leaving became negligible when a machine handles the logistics.
The companies that understand this will stop thinking about AI as a cost reduction tool and start thinking about it as something else entirely: a medium for trust.
Agency — the real kind, the philosophical kind — is the product now. Not efficiency. Not deflection. Not containment.
Autonomy.
The woman in the coffee shop just wanted to cancel a subscription. She shouldn't have needed twenty minutes and a piece of paper and a card number she couldn't find. She shouldn't have needed any of it. The technology to give her that freedom exists today. The question is whether the industry will use it to give customers genuine power, or just build a shinier version of the same old cage.