Do Not Create an AI Elevator Operator

Definition

When elevators got buttons, the right response was not to hire a new person to press them. It was to redesign the system. AI automation without workflow redesign makes the same mistake — preserving outdated processes with a new layer of technology instead of rethinking them.

The Metaphor

Early elevators required an operator to control the mechanism, announce floors, and open the doors. When automatic buttons arrived, the task of “pressing the button” could have been assigned to a new role. Instead, the system was redesigned: passengers pressed their own buttons, operators were no longer needed for that function, and the entire workflow changed.

The failure mode in AI adoption is the equivalent of assigning the button-pressing to a new role: using AI to automate one step in a process without questioning whether the process itself still makes sense.

What This Looks Like in Practice

• AI generates a meeting summary → but the meeting format, frequency, and attendee list are unchanged
• AI drafts a status report → but the report still goes to eight people who barely read it
• AI processes an approval form → but the approval chain has twelve steps designed for a paper-based world
• AI summarizes customer feedback → but the feedback loop never reaches the team that could act on it

In each case, the workflow is decorated with AI rather than redesigned around it. The result is usually marginal improvement — or additional complexity — rather than real value.

The Redesign Question

The right question after deploying AI on a task is not “is it faster?” It is: “Does this workflow still need to exist in its current form?”

Three redesign prompts:

• Delete. What steps no longer create value and can be removed entirely?
• Simplify. What steps exist because of previous constraints that AI has now removed?
• Reimagine. What becomes possible at the system level that was not possible when humans had to do every step manually?

The AWS example from Innovate is the redesign at scale: Amazon did not automate its existing infrastructure provisioning process. It redesigned the entire model into a reusable service — which became a different business.

What to Pay Attention To

• Where AI has been added to a process without removing anything
• Where processes designed for paper, email, or previous-generation tools are still running unchanged
• Where the biggest gains would come from deleting workflow steps rather than automating them
• Where “we’ve always done it this way” is the reason for a process that predates the current technology

Connections

Delegate Tasks vs Jobs Innovate Adoption Gap

Sources

• [inferred from workshop teaching — elevator operator metaphor is a classic workflow redesign analogy]

Tags: workflow redesign, operating model, automation, system thinking