What Changes When the Answer Is Already Yes

The AI decision is the easy part.

Most organisations have already decided to use AI somewhere. The harder question is what has to move once that decision is made.

 

 

 

 

1. Someone's decision rights change, and usually nobody says so out loud

If a system now flags exceptions or drafts recommendations, the person who used to make that call is either rubber-stamping it or being quietly bypassed. Ask directly: who used to decide this, and what do they do now?

2. Approval steps built for a slower process stay in place out of habit

A three-day sign-off cycle designed around a person checking work by hand doesn't need to exist once the check happens in seconds. Most organisations keep it anyway, because removing it isn't anybody's job.

3. A manager's job changes from doing the check to trusting the check

That's a different skill, and most managers were never told the job had changed under them. It shows up as rubber-stamping everything, or double-checking everything. Both defeat the point.

4. The escalation path needs rebuilding

When a person did the work, escalation meant handing it to someone more senior. When a system does it, that definition has to be rebuilt: what triggers it, who receives it, and what they're expected to do that the system couldn't.

5. Reporting was built to catch human error, and some of it now catches nothing

A control designed to spot a person's mistake doesn't automatically make sense once a system is doing the work. It's worth an audit of which reports are still checking for the failure mode they were built for.

 

Buying the tool is the yes. Redesigning the work is the project.