The AI Productivity Reckoning: Why Software Engineering Expectations Are Outpacing Reality

The software/digital industry has been undergoing a grand transformation for over 3 years now.

Post-ZIRP, LLMs, market shifts in SaaS, funding withdrawal, pandemic WFH, bubble burst – plenty of reasons piling up simultaneously.

The transformation isn’t over yet. This is how disjointed the market has been between 2012 and 2022.

Right now, AI “appears” to be doing a solid job – designs, MVPs, mockups. This increases the gap between a one-shot prototype and a production-grade system, or upgrading a platform that’s been developed for a decade or more, especially in finance or healthcare or aviation.

Expectations rising, salaries compressing, timeframes 5X to 20X shorter than the 2021-2022 peak.

I would say that today’s expectations are 3X less realistic than the norm, but they were 5X+ sloppier prior to that. The truth is in between.

With millions of FTEs across large corporations putting in 4-6-8 hours of real work weekly in during 2020-2022, asking them to perform at 80-100 hours of real output today is virtually impossible.

> You can stretch them at 15-20-25 hours.
> ~35 if they realize the severity of the situation and drop down admin overhead.
> 45 if they are really motivated or see growth

80+ requires the level of mastery for level 8 through 10 ICs in FAANG, usually making $600K + stock. The million-dollar salaries in AI are no surprise if unicorn engineers with high intelligence level, business acumen, management skills (people + agents), deep understanding of complex problems can act as CTOs with a team of 10 individually (and realistically cost a third of the combined team).

So, if the average engineer is underperforming and at level 2 or 3, and the market norm is looking at the unicorns, demand and supply see a major gap.

This isn’t helping the job market in the short terms, although the computer science world will level up gradually, weeding out the freeloaders from the bubble days. But some qualified talent will also step down or switch industries if pay is mostly constrained and requirements are 5X, and timelines for what used to be a 2-week sprint are now 2 days for 3 different initiatives.

AI will help some of that, sure.

And break a portion of the industry on the go.