The anthropomorphization phase of AI
Every AI agent apparently needs a first name, a personality, and a job title. Your agent is not your co-worker.
Steve Jobs made skeuomorphism famous. He designed software to look like real-world objects — leather calendars, wood bookshelves. Anthropomorphism is the AI version. Except we are not making software look like objects anymore. We are making it look like people. Meet Carly, your new "team member." You can "hire" her on a landing page.
It is clever. It makes the technology less threatening and adoption easier. I even considered it for our own onboarding at Causa Prima. But Carly is not your co-worker. She is a piece of software with a memory, a set of skills, and an LLM for a brain.
Marketing it like a human is risky for two reasons. First, it is a business-model play: it lets SaaS companies sell per-seat licenses. One agent, one license. But an agent is not a user. It is software with connections everywhere, tapping into dozens of systems. Charging per seat for that makes no sense.
Second, when you treat an agent like a colleague, you stop questioning it. An LLM does not think — it predicts. But if you believe Carly is your co-worker, you are not auditing her output line by line.
For Apple, skeuomorphism was a phase. The same will happen with AI agents, and the faster it does, the better. Your agent is not your co-worker. It is software that works 100x faster than any co-worker you have ever had.