𝐒𝐨𝐦𝐞 𝐒𝐭𝐫𝐮𝐠𝐠𝐥𝐞 𝐈𝐬 𝐋𝐨𝐚𝐝-𝐁𝐞𝐚𝐫𝐢𝐧𝐠.

Last quarter you handed the monthly board pack to AI.

Years of Sunday evenings chasing numbers across six systems, back in an afternoon.

You felt nothing. No loss, no nostalgia. Almost none of it had taught you anything you still use.

Then you did the same with the first draft of your strategy note.

That one felt different, and you could not say why.

That gap is the whole point.

Not “should we automate this?”

But “which of these two is which?”

And the cost of guessing wrong may show up years late.

In April, researchers from Carnegie Mellon, Oxford, MIT and UCLA ran three trials with 1,222 people. After just 15 minutes of AI assistance on maths and reading tasks, participants performed worse when the help disappeared.
They also gave up more readily.

So yes, cognitive offloading has a cost.

But that does not make all effort sacred.

We romanticise effort surprisingly easily.

Doing something the hard way can feel virtuous long after the difficulty has stopped teaching us anything.

What matters is what the struggle was against:

A bad system, or the problem itself?

Fighting six disconnected data sources teaches you about your plumbing.

Debating a strategy question teaches you about your business.

Put differently:

Productive friction teaches you the problem. Administrative friction teaches you the system.

Automate the second aggressively.

Be much more careful with the first.

The trouble is that we often discover which was which only after the capability has gone.

In 2009, Pisano and Shih warned that offshoring supposedly “low-value” manufacturing could quietly erode the industrial commons: the shared pool of suppliers, engineers and know-how that future innovation depends on.

It sounded like nostalgia.

But once layers of manufacturing capability had moved elsewhere, bringing the final product back was no longer simply a question of labour cost.

The capability had moved too.

Nobody had to be careless for this to happen.

The feedback loop was simply too long.

That is what worries me about AI.

We can see the hours it gives back this quarter.

We may not see the judgment, intuition or understanding we stopped building until years later.

We can still act on that lesson.

We can decide which capabilities we need humans to retain, then design the work around keeping them.

The firms that do that will still understand their own business in 2036.

“How do I know which is which?”, you might ask.

Personally, I use one question before automating away anything to AI:

"If I stop doing this myself, what will I stop learning?"

That question has started to matter as much as how many hours the tool can save.

What part of your job looks automatable but might actually be load-bearing?

#ArtificialIntelligence #CoreCompetence #FutureOfWork #DecisionMaking #CriticalThinking

Contact

bruno.gentil@sherpaconsultingasia.com

© 2026. All rights reserved.