1. News
  2. Everyday tech
  3. Work IT · AI risk
Work IT · AI risk2 min read

Companies use AI to write IT code fast, but can't undo mistakes fast

A survey of Apple IT admins finds most use AI to write work that reaches live systems, but 69% cannot undo a bad settings change within an hour.

By the Dashline deskSource: 9to5Mac
Companies use AI to write IT code fast, but can't undo mistakes fast
Photo: Sebastiaan ter Burg from Utrecht, The Netherlands, CC BY 2.0

IT teams that manage Apple devices are using AI to move faster, but many cannot fix a mistake quickly. A new survey from device management company Fleet found that 69% of admins cannot roll back a bad configuration change within an hour. And 43% need more than a day.

What the survey found

Fleet surveyed more than 250 enterprise IT workers who manage Apple devices. All had admin access to their device management systems. They oversee an average of 2,000 devices each, and nearly half have 6 to 10 years of Apple admin experience.

AI use is already routine. In the past 30 days, 68% used AI for troubleshooting or reading system logs. Another 62% used it to write scripts, and 61% used it to write device management profiles or configurations. A profile is a set of rules that tells company computers how to behave.

  • 86% allow AI-written work into live systems after some form of review
  • 14% keep AI-written work out of live systems entirely
  • 26% can recover the same day, but only by doing it by hand

Why speed cuts both ways

The column on 9to5Mac, written by longtime Apple IT admin Bradley Chambers, calls the recovery numbers the biggest red flag in the report. AI can ship bugs just like people can. The difference is how fast a bug can spread.

One example: an admin asks an AI assistant to write a configuration profile, gives it a quick look, and sends it to a few thousand Macs in seconds. If it contains a bad network setting, or a restriction that locks people out of an app they need, undoing it could take a while.

A heavier workload

The survey points to wider pressure on IT. 31% named keeping up with Apple's yearly release schedule as their biggest challenge, far ahead of anything else. On the positive side, 53% already use declarative device management, Apple's newer approach to managing devices, in production.

Chambers argues that Apple is not the problem. Security has become harder, and IT teams are expected to take on AI while still owning everything else, often without extra staff. In his view, AI is how many admins keep up.

What to do

The practical lesson for any business using AI to help run its computers is to plan the undo before the rollout. A quick review of AI-written work is not the same as being able to reverse it. Fleet's full report is available on its website.

The takeaway

Before pushing any AI-written change to company computers, make sure you can reverse it quickly.

More articles

You should really read these.

Next step

Let’s talk