The first generation of enterprise AI answered questions. The second generation executes tasks. That shift — from assistant to agent — is the point at which process design stops being an IT concern and becomes a management one.
Most enterprise processes encode a compromise: a sequence of handoffs designed around human attention spans, working hours, and error rates. Agents have none of those constraints. When you give an agent the ability to plan, call tools, and act, you inherit the option to delete the handoffs rather than automate them.
“Automating a step saves a percentage. Deleting a handoff saves a role.”
In practice we see three postures. Automators bolt agents onto existing steps and measure cycle-time savings. Redesigners re-derive the process from the outcome and keep only the human checkpoints that carry accountability. Reinventers go further and change the product itself — a claims process becomes a real-time settlement service; a maintenance schedule becomes a continuously re-optimised contract with the machine.
Automators capture 10 to 20 per cent. Redesigners routinely capture 40 per cent or more. Reinventers occasionally create revenue lines that did not exist, but they also carry the highest change-management load, because reinvention invalidates job definitions rather than merely resizing them.
Whichever posture you choose, four controls must exist before agents touch production: a bounded action space, an audit trail that reconstructs why an action was taken, a human escalation path with a service level, and a kill switch that a business owner — not an engineer — can pull.
The organisations moving fastest are not the most permissive. They are the most precise about which decisions they are willing to delegate and which they never will.
