Integration Depth Ladder

AI agents ship as finished workers, so the agent itself prices like a commodity. What does not commoditize is how many of a client's systems the agent touches and how tightly it is bound to their review cycle. The Integration Depth Ladder ranks deployments by surface area: a single-channel agent (say, an AI receptionist answering calls) sits at rung one and is trivially replaceable; an agent wired into CRM, calendar, billing, and a human approval step sits at rung four and is expensive to rip out. Agencies should price and pitch by rung, not by agent count. Vendasta's white-label AI workforce and Lindy's inbox-and-calendar assistant both illustrate the low rungs, where resale margins compress fast. Relevance AI's sales and follow-up agents, deployed against a client's existing pipeline stages, show the higher rungs where the wiring, not the model, justifies a monthly retainer. Audit each client deployment and ask which rung it actually occupies.