To tell a process problem from a software problem, ask two people to describe the same task

The fastest way to tell a process problem from a software problem is to ask two people who do the same task to describe it separately. If their descriptions differ, you have a process problem, and new software will automate the disagreement. If they agree but the tool cannot record what they do, you have a software problem, and the right tool or connection will fix it.

Disagreement between two people is the signal of a process problem

Owners reach for software first because software is easy to buy. The signs of a process problem are quieter. A step exists only because the previous step cannot be trusted. Someone keeps a spreadsheet beside the official tool. Two team leads run the same work in different ways, and both think theirs is the standard.

A software problem looks different. Everyone agrees on the steps, but the tool forces them to retype information, has no place for a field everyone needs, or cannot talk to the system that holds the record. Those problems are worth buying or building for.

The same tools did about twice as much inside a redesigned process

Bain’s Technology Report 2025 looked at AI coding tools. Teams that used them on their own saw productivity gains of 10–15%. Companies that paired the tools with a redesign of the end-to-end process reported 25–30%.¹ The tools were the same. The process around them was what changed.

Approach	Productivity gain reported
AI coding tools on their own	10–15%
Tools plus a redesigned end-to-end process	25–30%

Source: Bain & Company, Technology Report 2025¹

Ford cut accounts-payable staff by 75% only after redesigning the process

In 1990, Michael Hammer described how Ford set out to cut the headcount of its North American accounts-payable department, then more than 500 people, by about 20% with new computer systems. Only when it redesigned the process itself did it reach a 75% reduction.² Clerks had matched 14 data items across three documents before paying a supplier. The new process matched three items, automatically.

1. New systems, as first planned20%
2. Process redesigned75%

Redesign cut Ford's accounts-payable headcount by 75%; new systems alone were planned to cut 20%

Approach	Headcount reduction
New systems, as first planned	20
Process redesigned	75

Source: Hammer, Harvard Business Review, 1990²

Hammer’s title for the article still works as advice: “Don’t Automate, Obliterate”. Remove the steps that should not exist, then automate what is left.