DEV Community

Anton Loginov
Anton Loginov

Posted on

Deploying agents is harder than deploying software — and that is the point

#ai

A take from the delivery side, after two years shipping agentic workflows into professional-services firms.

Software vs agents

Software generally works the same way every time. For decades it has mostly been updated versions of an existing concept, so the enterprise knows how to slot it into a workflow.

Agents are different. You are deploying the equivalent of work output inside the business. The customer effectively uses you as a professional-services provider for a task they expect solved nearly end to end.

The harder gap

Everyone frames this as a vendor-skill gap — you need forward-deployed engineers who understand the business. True. But the harder gap is on the customer side: most enterprise processes were never designed to be handed to anyone, human or agent.

The first job on the ground is usually not deploying the agent. It is making the operating model legible enough that an agent can even be pointed at it.

The math nobody likes

Deploy is a week. The redesign is the quarter. Teams that budget only for the deploy keep shipping capable agents into processes that quietly reject them, then blame the model.

The model is the easy part. It always was.

Top comments (0)