All Posts

A human approving every agent action isn't a control, it's a signature, and it doesn't scale. The test for which of your safeguards actually enforce anything, and where a human still earns their seat.

An agent that can act can't be measured with metrics built for chat. Here's what to actually watch, where the honest numbers come from, and the instruments that lie to you on the way.

Air Canada told a tribunal its chatbot was a separate legal entity responsible for its own actions. It lost. When an agent ships, someone still has to own what it's allowed to do, and that's a product job, not a code one. An agent with no product owner is a liability with a friendly voice.

For weeks, a lender's support agent quietly handed out free credit. Nobody caught it, because nothing looked broken. It was a buy-now-pay-later lender I worked with, for a fintech. The agent had real authority: it could pause a plan, waive a fee, restructure a schedule, on its own. Useful. Also quietly expensive. Every time it paused a plan sitting on a 0% promo, the pause broke the promo's conditions, and a broken promo defaulted to interest-free, so the customer kept the 0%