Using Automation vs. When to Use an AI Agent

This week I’m going solo to make a case that some of the most important infrastructure in your business isn’t software, it’s your people. Your systems can move data, but they can’t understand the work, so people like a customer success rep named Mark have become the intelligent integration layer between your CRM, billing platform, Slack, and email. I walk through Mark’s customer request step by step, show where traditional APIs and automations stop, and explain how AI agents can move work with the right context when the movement isn’t deterministic. The goal isn’t to remove people from the business but to get them out of the infrastructure layer, and I close with an exercise to find where that’s happening in your own company. If you’re an investor or operator designing a workforce of people and AI agents, this one is for you. Episode Timeline & Highlights \[0:00] – Why some of the most important infrastructure in your business is your people, not your software \[1:00] – Moving data is not the same as understanding the work, and the customer request that shows the gap \[1:31] – Mark opens the CRM, billing platform, Slack, product platform, and email to assemble the full picture \[2:56] – What Mark really did: connected systems, interpreted information, and acted as the integration layer \[5:09] – The limits of traditional integration, and a failed payment that an API can move but not understand \[6:45] – Where deterministic movement is perfect: if X happens, do Y \[7:50] – Systems could never answer who should receive the work, so people provided the intelligence \[8:28] – How an AI agent handles the failed payment using customer context, policies, and its authority \[10:21] – Moving the work with the relevant context, not a task that just says please handle \[12:20] – Not every movement needs an agent: if X reliably means Y, use the API, automation, or rule \[13:19] – Scale: a person monitors 50 situations while an agent can stay aware across thousands \[14:52] – The goal is to get people out of the infrastructure layer, not out of the business \[16:32] – A person who is the right worker for the job versus a person required to connect the work \[17:36] – Receive, execute, hand off, and who provides the intelligence between those stages \[19:59] – The exercise: find the people who seem to know how everything connects and watch what they do \[21:29] – If the movement needs intelligence, could an AI agent provide it instead of a person? Key Takeaways 1. People Became the Infrastructure: Your systems could move data but couldn’t understand the work, so people filled that gap. 2. Moving Data Is Not Understanding Work: An API can tell your CRM a payment failed, but it can’t tell you what that failure means. 3. Keep Deterministic Work Deterministic: If X reliably means Y every time, use an API, automation, or business rule instead of an agent. 4. Agents Provide Intelligent Continuity: An agent can decide where work should go and send it along with the context the next worker needs. 5. Get People Out of the Layer, Not the Business: People should be involved because they’re the right worker for the responsibility, not because the infrastructure can’t understand the work. Links & Resources - smrtPhone: https://www.smrtphone.io - That Real Estate Tech Guy: https://thatrealestatetechguy.com If this made you think of the Mark in your own business, spend a week watching what he actually does between your systems. You may find work that was never his job to begin with. More high-signal conversations coming next.