Time For A Career Pivot: Technical Workers Are In For A Tough 2025

The running joke is that before you replace software engineers, users must first figure out how to clearly articulate their needs and write their own specifications. It's funny because, for the 30 years that I've been in tech, it hasn’t happened. Challenges with AI agents for application development extend beyond the technology.

Still, even I am being forced to reevaluate how soon AI’s impacts on technical roles will be felt. It’s impossible to ignore the change in tone and how widespread it has become. We expect it from Founders working to hype up their products and CEOs pandering to investors. Now they’re bringing receipts.

No one with any credibility thinks software and data engineers are disappearing, but many are discussing a sea change in demand. Technical skills are still important, but technical skills alone are insufficient. New capability combinations are seeing growing demand, while pure technologist roles are moving in the opposite direction.

I was floored when Andrew Ng came out with this post earlier today. I've been saying the same thing for years, but seeing somebody like Andrew Ng join in is unexpected. The research and technical crowds rarely acknowledge the need to figure out what we should be working on before we start building or researching.

There are twice as many job openings for data and AI product managers (12K) on LinkedIn as data engineer job openings (5.7K). Last year, there were nearly twice as many data engineering job openings and only about 8K data and AI product management jobs.

Many CEOs and thought leaders in our industry have changed their perspectives about the implications of AI in the last couple of months. We may not be at the point where we can replace people, but AI will definitely augment. A growing consensus is that augmentation will significantly impact demand for technical roles.

It doesn't matter what technical role you have; this reality is coming soon to your company and every other business in the world.

What’s Changing?

Andrew's thesis is Economics 101. He explains that demand will rise as the cost of building applications (or anything that relies on code) decreases. The emerging challenge isn't writing code but articulating what the code should do. Increasingly, the most important decisions are shifting from how to build to what we should be building and why.

Andrew's economics case says that increasing the number of apps and decreasing the effort and expertise required to build them will increase demand for people who can decide what should be built. For most businesses, that's a value-centric conversation, so he's now firmly in the camp that believes product management will be one of the hottest jobs for the next decade.

An increasing number of credible people with engineering backgrounds are explaining the implications of automating much of the software development life cycle, and it is difficult to ignore.

Mark Zuckerberg has been on an incendiary tour for the last month. However, his comments about automating software engineers at some point this year aren't bluster. Based on what he's seen, Zuckerberg believes that a lot of mid-level software engineering work can be automated. I suspect that also extends to entry-level software development and engineering. Mark Zuckerberg outlines a future where only senior-level engineering capabilities are required, even in the most advanced companies like Meta.

It’s impossible to ignore the inefficiencies many of the biggest tech companies operate with and the over-hiring between 2014 and 2022. It’s also hard to ignore the feedback from frontline software engineers on AI tools. In a nutshell, ‘They’re cool and useful, but only get you so far.’ The numbers back software engineers up. I’m definitely team Ng, but I’m not buying Zuckerberg’s take. Putting AI into the hands of senior developers to make them 10X more productive might happen someday, but not this year.

I believe the CEO of Replit has a more accurate view of how AI enters the software engineering world.

Moving AI Engineering Tools From Technical To Nontechnical Users

Replit’s CEO chimed in as well this week. The company recently went through a massive downsizing. Unlike many of its peers, Replit’s revenue is growing. The company is less of a downsizer and more of an early adopter, using its own AI tools to change the way work gets done.

Replit’s CEO paints the picture of a future where nontechnical domain experts can use natural language to develop many of their own apps and APIs. He says the company isn’t targeting professional coders and software engineers with AI coding tools. Replit designed its tools for nontechnical users, and it’s a smart strategy.

Non-developers will get a much bigger productivity increase when it comes to writing code than professional developers. They’ll go from writing 0 lines of code to delivering simple apps, APIs, dashboards, and more. I have seen this happen firsthand and know how it plays out. Replit’s CEO is close, but his view isn’t 100% accurate either.

The first problem is adoption. Getting nontechnical users to use AI coding tools will take training, culture change, and new incentivization structures. Getting semi-technical users to run early pilots with these tools is easy, but wide-scale adoption is more complex. Replit’s in the honeymoon phase with semi-technical early adopters. If they don’t support users with their nontechnical challenges, Replit will see significant churn.

As adoption rises, the number of boondoggles does too. Nontechnical users will waste time building things just to do it. It’s empowering when they realize their dependency on technical teams has been broken. However, just like early career software engineers and students start a ton of cool-sounding projects with questionable utility, AI tool users do, too. This is one demand driver for product managers who will step in to prevent users from spending more time building tools, reports, and even models than doing their actual jobs.

Nontechnical user-built tools must also align with a larger product roadmap. Otherwise, the cost of managing the infrastructure sprawl becomes massive. Again, product managers will step in to help the business maintain control.

Andrew Ng’s analysis is excellent because AI tools like Replit’s will significantly increase the number of technology builders. They require a corresponding increase in the number of technology-value aligners. The trend also increases the demand for enterprise-wide data and AI strategy that breaks out of the technology organizational silos.

CEOs Will Overreact And Prematurely Reduce Technical Staffing Levels

An overreaction is inevitable and has happened with every technology cycle. Most CEOs will follow, ‘We can do the same with less.’ They’ll freeze technical hiring and cut back their technical teams. Why build AI tools when we can buy them from all these startups and vendors? Why keep our current technical staffing levels when we can use those tools to build most of what’s on the roadmap today?

Marc Benioff said last month at the Agentforce 2.0 launch that CEOs must take an abundance approach to AI. Augmenting staff means the business can do more with the same. Delivering more leads to growth. Delivering the same leads to stagnation in the face of competitors who embrace abundance.

When the pendulum swings back and hiring picks up, technical teams will look for new capabilities. Technical ICs will take on co-development projects and frequently partner with nontechnical domain experts who write code using AI tools. In my view, the days of technical teams being siloed in the technical organization and insulated from engaging with business users are over.

Communication and business acumen are as important as the ability to write code. Collaborating with the business is on an equal footing with collaborating with the rest of the technical organization. Technical leads will lead nontechnical teams as often as they lead teams of software and data engineers. Data scientists will be embedded into business units and product teams.

Data and AI product managers will work with truly cross-functional, cross-domain teams. Data and AI strategists will be required to get the business AI-ready, which demands culture change and aligning transformation.

These changes won’t happen over several years. The more people who come out and say the same things I have been, the more convinced I am that change is already happening and gaining momentum.