When a company adds AI to an outdated process, it may not achieve a more modern operation. It may simply reproduce unnecessary approvals, data gaps and organizational barriers faster. A survey of 2,850 senior executives shows a shift away from indiscriminate adoption toward more rigorous examination of value. Before automating, therefore, you need to decide whether the original process should continue to exist at all.
The first wave of AI adoption in companies often rested on the assumption that the existing process could remain intact and simply have more powerful technology inserted into it. This approach is now beginning to hit its limits.
Forbes describes a G-P survey of 2,850 executives at vice-president level or above. The share reporting aggressive use of AI for innovation fell substantially from the previous year. This does not necessarily mean leadership is losing faith in AI. Companies are beginning to examine more carefully whether the technology actually addresses the root cause of a problem.
A typical mistake occurs when an outdated workflow is automated. If, for example, a company uses a complicated recruitment or onboarding process with several unnecessary handoffs and approvals, adding an AI assistant does not remove the underlying flaw. It may only accelerate the administrative part of a system whose logic no longer fits current needs.
The management approach should be the opposite. First ask what outcome the process is supposed to achieve. Then identify the steps that genuinely create value or control a significant risk. Everything else is a candidate for removal, not automation.
Only once the process has been simplified does it make sense to decide where to use AI. Otherwise a dangerous situation emerges: the technology accelerates the flow of work while also increasing the number of wrong or low-value steps performed per unit of time.
This approach also changes how return is evaluated. It is not enough to count the number of tasks processed by the system. You need to track whether the overall process became shorter, whether handoffs declined, whether errors fell, or whether experienced people spend less time on low-value work.
For leadership, a simple rule is useful: no significant AI project should begin by asking how to automate the existing procedure. It should begin by asking whether the company would design the same procedure today if it were building it from scratch.
AI can therefore function as an uncomfortable diagnostic tool. It exposes places where the organization has tolerated inefficient operations for years because the costs were dispersed across people and departments. If the work model itself is not changed first, the technology merely creates a more sophisticated version of the same problem.
KEY TERMS
- Automating the mess: Speeding up a process whose underlying structure is itself inefficient.
- Process redesign: Rebuilding a workflow around the desired outcome rather than the current sequence of steps.
- Process handoff: The point at which work or information moves between people, departments or systems.
- Operational friction: Unnecessary delay, administration or coordination that adds no value to the outcome.