luismori.

AI and Engineering Leadership

AI Engineering Product Future of Work Developer Productivity

How AI Is Reshaping Engineering: The Shift From Code to Product

AI isn't just making engineers faster — it's changing what engineering means. Here's what that shift looks like, and why the best days for software teams are still ahead.

7 min read

Something fundamental is changing in software engineering. Not the tools — the role itself.

For decades, engineering was largely about turning requirements into working code as efficiently as possible. Product managers brought the ideas, designers shaped the experience, and engineers built the thing. It was a conveyor belt. Each role had its lane.

AI is blowing up the conveyor belt.

The Speed Shift Is Real

Let’s start with the numbers, because they’re striking. Developers using AI coding tools complete certain tasks 51% faster. In a study across 4,800 developers at Accenture, JavaScript work finished 55% faster with GitHub Copilot, pull request cycle times dropped from 9.6 days to 2.4 days — a 75% reduction — and successful builds increased by 84%.

92% of developers now use AI coding tools. About 41% of all code written today is AI-generated. This isn’t a niche experiment anymore. It’s the new default.

And the implication isn’t subtle: if it takes a fraction of the time to write code, the thing that’s now scarce isn’t engineering capacity. It’s good ideas about what to build.

The Attention Is Shifting to Product

When implementation gets cheaper, product thinking gets more valuable. This is already showing up in how teams are organized.

Etsy’s CPO noted their PM-to-engineer ratio shifted from 1:10 to 1:6 — not because engineers disappeared, but because discovery, experimentation, and product definition suddenly required more attention. LinkedIn’s chief economic opportunity officer described how “the full stack builder takes what would’ve been days or weeks as a conveyor belt between design, product, engineering… and gives it to an individual with these tools.”

Anthropic’s own internal research on how their team uses Claude found that 27% of AI-assisted tasks were work “that would not have been done otherwise” — engineers using AI to revive abandoned ideas, build internal tools that previously weren’t worth the time, and scale in-progress projects in directions they couldn’t before. The bottleneck wasn’t talent. It was time. AI is dissolving that bottleneck.

This is the real unlock: when engineers aren’t heads-down on implementation details, they can think about what actually matters. Better user experience. Tighter product loops. Faster experimentation. And faster experimentation means learning what actually generates revenue — sooner.

Engineering Is Evolving, Not Disappearing

The natural fear here is obvious: if AI writes the code, do you still need engineers?

The evidence says yes — emphatically — just differently.

Klarna shrunk its workforce significantly while deploying an AI assistant it says is equivalent to 700 full-time employees. Shopify’s CEO told teams they must demonstrate why a task can’t be done with AI before requesting more headcount. In 2025, 55,000 job cuts were directly attributed to AI efficiency gains. These are real shifts.

But zoom out, and the picture gets more interesting. The World Economic Forum’s 2025 Future of Jobs Report projects 92 million jobs displaced but 170 million new ones created by 2030 — a net gain of 78 million. AI engineer roles grew 143.2% year-over-year. 35,445 AI-related positions were open in just Q1 2025, up 25% from the year before.

What’s actually happening isn’t replacement. It’s recomposition. One engineer can now do what three could before. But when that engineer — or that small team — builds a better product faster, gets it to market, and starts generating revenue, then the demand for engineering scales up again. Now you need more people building features. More people maintaining systems. More people experimenting. Except now those engineers are working with AI, so they’re dramatically more capable than the engineers you’d have hired five years ago.

The Problems Engineers Get to Tackle Are Getting More Interesting

Here’s the part that doesn’t get talked about enough: AI isn’t just making existing work faster. It’s opening up work that was previously too expensive to attempt.

Anthropic’s research found that engineers are increasingly becoming “full-stack” in a new way — backend engineers confidently shipping frontend interfaces, security engineers building data visualizations, solo developers tackling architecture decisions that previously required a whole team. Infrastructure that was “too complex to fix” is now approachable. Technical debt that was “too risky to touch” is now manageable. That long-running idea for an internal tool that nobody had bandwidth for? Built in an afternoon.

This is a meaningful expansion of what engineering teams can actually accomplish — not just speed on existing tasks, but scope expansion into problems previously off-limits.

What This Means in Practice

The engineers who are thriving right now share a few traits:

They spend more time on product judgment. What should we build? What’s the simplest version that tests the hypothesis? What’s the riskiest assumption here? These questions matter more when execution is cheap.

They’ve gotten better at directing AI. Writing a vague prompt and accepting the first output isn’t engineering. Knowing how to decompose a problem, validate AI output, catch what it missed, and iterate toward something solid — that’s the skill.

They’re comfortable with a wider surface area. The specialist who only speaks backend or only cares about infrastructure is at a disadvantage. AI makes it feasible to be useful across the full stack, and teams that leverage that are moving faster.

They think in systems, not lines of code. Architecture, reliability, observability, scalability — these decisions compound. AI can generate code; it can’t make good architectural choices on your behalf.

The Optimistic View (And Why It’s Justified)

Here’s the honest case for optimism: software is not a solved problem. The world is still full of industries with terrible tooling, broken workflows, and unmet needs that could be addressed with software. The bottleneck has always been engineering capacity. If AI meaningfully expands that capacity, the number of valuable products that can be built — and therefore the number of engineers needed to build and maintain them — goes up.

The companies that figure this out fastest are going to grow their engineering teams, not shrink them. They’ll have engineers focused on the things AI can’t do well: customer empathy, product intuition, judgment calls under uncertainty, and the messy human work of figuring out what actually matters.

That’s a more interesting version of engineering than “turn this Jira ticket into code.” And it’s the version that’s arriving.

Sources