8 Comments
User's avatar
Aaron Laam's avatar

The idea of shrunken jobs is the part I'll keep thinking about. If AI takes the first drafts and routine analysis, the people who used to do that work can become the reviewers of it, which keeps them building the judgment you describe instead of losing the reps. Does the PwC program teach that kind of review, or does it stay more general?

Fard Johnmar's avatar

I haven’t seen their full curriculum.

But sometimes the first draft is the most important, so it’s a balancing act.

Aaron Laam's avatar

Fard, that’s the part I’d miss. Writing the first draft is usually where I work out what I think. Maybe the answer is letting juniors write some drafts from scratch on purpose, even when AI could do it faster.

Fard Johnmar's avatar

Yes. Mainly so they can go through that process.

Another tactic is brainstorming an idea entirely without AI and then working with an LLM to bring that idea to life. Coming to it with a clear sense of purpose and using it solely as a tool to bring ideas to life is also a useful skill.

Lisa Kostova's avatar

Fard, I appreciate your perspective here. Something you said, paired with what I'm going through right now, made me think deeply. You said that people can keep their advantage by specializing and honing the parts of a job that others outsource to AI. I think that may be valid for certain tasks and not valid for others.

I suspect that there are two variables that influence that choice - CONSEQUENCE and OUTPUT TYPE (Deterministic or qualitative). Being probabilistic, AI is going to make mistakes when it's asked to provide a deterministic output, and when the consequences of that mistake are high, you won't be able to safely outsource this task to AI. Examples are regulated industries - for example, there's a reason why only licensed people are allowed to sell insurance or prescribe medication.

On the other hand, tasks where the output is qualitative and the consequences of a mistake are low are a great use case to outsource to AI - I'm talking about low-consequence research, summaries, recommendations, non-creative content generation (like product descriptions) and low-consequence information retrieval. I don't think it would be wise for any person to specialize in doing those tasks anymore, even if they really enjoy them.

The key is to be discerning and discriminating about what you choose to continue specializing in, and not just go after everything other people are outsourcing to AI.

Fard Johnmar's avatar

Thanks for reading/watching and engaging Lisa! I’m glad my piece got you thinking. :).

I was referring to the phenomenon where people are outsourcing a lot of their critical thinking, analysis, strategic insights, and ‘gut instincts’ to AI.

Those core skills, regardless of the task, are high-value. If others are willing to give up these advantages, that’s fine, but I don’t think that’s the right choice. And, there’s an opportunity here as well.

Lisa Kostova's avatar

I should be thanking you. Your perspective gave me the angle that I’m drafting my next essay around - so thanks for acting as the spark! I think the outsourcing of skills question still comes down to the variables of consequence and accuracy. When either one is high, the outsourcing risk is bigger than any efficiency gained IMO.

High-consequence activities include visible writing whose quality reflects on the reputation of the writer - such as Substack essays, a book, a talk. Product descriptions, meeting summaries or drafting a usage manual are not such type of activities, so I would argue that even if they were written by writers who used discernment and critical thinking in putting those together in the past, those activities should be outsourced because AI will do a better job much faster than any human on those.

So unfortunately, the test is not whether you used to employ critical thinking in certain domains - the test is whether the domain is worth keeping a human in the loop in, even if it means parting with things you enjoyed using your brain on.

Fard Johnmar's avatar

Thank you!

There's a saying in healthcare: Let clinicians practice up to their license.

Meaning take away the repetitive, low-value tasks and allow them to concentrate on the highest impact activities.

In the United States, clinicians spend a lot of time writing up notes, filling out paperwork, contesting decisions of insurance companies, etc.

There are tools and solutions that take that off a clinician's plate so they can focus their attention toward the highest-value activities. That's where their crucial thinking, judgment and strategic thinking are best used.

So there's value in outsourcing that activity.

But what to outsource really depends on the context. For example, getting a product description right might have legal and regulatory implications. In those cases, that's a high-value activity that humans should pay a lot of attention to. In other contexts, it's not that important and can be outsourced. We used to outsource those types of activities to other humans, now it's being given to AI.

So it's context dependent. If it's a high-stakes decision or strategy, both the inputs (routine stuff) and the outputs need attention.