Anthropomorphism of AI Agents: Yes? No?

There has been a lot of buzz this year surrounding AI Agents - how they code, will replace humans, how humans will manage teams of AI agents, and more. Entrepreneurs are boasting of scaling their workforce with far fewer people. They are setting new benchmarks for revenue per employee by deploying agentic systems and thinking of AI agents as replacements for human staff. Now I will park of all that - too much to debate.

"Dawn of the Agent Economy" was the headliner that drew me into Intercom + Apify's event last week.

Consider a very broad definition of "AI Agent" - the chatbots that so many envision, the OpenAI/DeepSeek/Google LLMs that reason, and the billion bits of code that Salesforce will deliver by September 2025.

Here are the questions these entrepreneurs raised and were working to answer (and solve):

• Should AI agents have identities - unique identities?

• If agents (think little ones more akin to APIs) are going to interact with other agents or entities, do they need the ability to pay? if so, how?

• If agents misbehave or make mistakes, do we scold them or make them feel bad?

If we start to treat technology like a person (and I am not condoning this), there are a lot more questions - ethical, practical, safety - that we need to answer.

Somewhat related, Intercom has an outcome-based business model. They have 5,000+ customers (not necessarily all paying). They charge $.99/resolved question or outcome. It's working for them. Reminds me of the early days of mobile and the Internet. Will too many incumbents miss opportunities with new technology while protecting their old tech stacks and business models? We'll see.

Background: On April 23rd, 2025, I attended an event (Dawn of the Agent Economy) hosted by Intercom and sponsored by Apify. Great event for so many reasons - well organized, fast paced, hard-hitting, thought-provoking, and authentic. I appreciate when entrepreneurs code on stage or demo not quite ready application - keeps it real.