The future of engineering: what's left for humans to do
By Albert Nahas, 9 min read, updated
The future of software engineering is less about typing code and more about deciding what to build, checking that it works, and owning what happens after it ships. AI will not replace software engineers who do that work, but it is already shrinking the jobs that were mostly about turning a clear ticket into code.
I have been shipping software for 13 years, and I have never seen the job change this fast. At a European health-tech scale-up I led an AI coding rollout to 90% adoption across engineering. On the side I run a product, OpenWhen, that I built alone and that more than 70,000 people have used. Agents help me run it. So this guide comes from someone who uses these tools every day, and who also mentors the people most worried about them.
What changed in the last two years
The short version: AI coding went from an autocomplete trick to the default way most developers work.
- In the Stack Overflow Developer Survey 2025, 84% of developers said they use or plan to use AI tools, up from 76% the year before. About half of professional developers used them daily.
- Google's DORA 2025 report put AI adoption among software professionals at 90%.
- The Stack Overflow 2026 survey results show 73% of people who use AI coding tools use them daily, and Claude Code and GitHub Copilot as the most-cited agents.
What didn't keep up is trust. In the 2025 survey, 46% of developers said they distrust the accuracy of AI output and only 33% trust it. The top frustration, named by 66%, was "AI solutions that are almost right, but not quite."
That gap between how much we use these tools and how much we trust them is where the future of the job sits.
Will AI replace software engineers?
Not as a profession, based on the evidence so far. It is changing who gets hired and for what.
The most careful study I know is Stanford's "Canaries in the Coal Mine?" by Erik Brynjolfsson, Bharat Chandar and Ruyu Chen, built on ADP payroll data for millions of US workers. The August 2026 revision finds no evidence of economy-wide job displacement. It also finds that employment for workers aged 22 to 25 in AI-exposed occupations is 19% below where it would be if it had kept pace with less exposed peers, while experienced workers show no comparable gap.
Two details matter more than the headline:
- The gap comes mainly from fewer hires of young workers, not from people being fired.
- Declines cluster where AI substitutes for human tasks. Where it complements people, employment is flat or rising.
So the question isn't "will AI replace engineers?" It is "which parts of engineering does AI substitute for, and which parts does it amplify?" Your career depends on standing in the second group.
One more honest caveat: the authors' own February 2026 follow-up says interest rates and timing also play a part. AI is one cause, not the only one.
The parts AI substitutes for
From my own work, these are the tasks that now take minutes instead of days:
- Writing a component, endpoint or migration from a clear description
- Boilerplate, glue code, test scaffolding, type definitions
- Translating between languages, frameworks and API versions
- First drafts of documentation
- Routine bug fixes where the error message points at the cause
If your job is mostly the list above, you are competing with a tool that costs a few dollars a day and never sleeps. I don't say that to scare you. I say it because pretending otherwise would be bad mentoring.
The parts that got more valuable
Here is what I spend my time on now, and what I see senior engineers being paid for:
| Still human (for now) | Why AI doesn't own it |
|---|---|
| Deciding what to build | Needs users, business context and taste the model doesn't have |
| Writing the spec and constraints | Output quality is capped by input quality |
| Reviewing and verifying | "Almost right" code is the most expensive kind |
| Owning production | Someone has to answer for the outage, the bill and the data |
| Measuring outcomes | Shipping isn't the goal; changing a number is |
| Designing guardrails for agents | Agents need limits, kill switches and escalation rules |
Verification deserves a closer look. A 2025 METR study gave 16 experienced open-source developers real tasks, with and without early-2025 AI tools. With AI they took 19% longer, yet they believed AI had made them about 20% faster. The authors warn against generalizing, and the tools have improved since. The lesson still holds: feeling productive and being productive are different, and only measurement tells you which one you are.
DORA describes AI as an amplifier. It makes strong teams stronger and messy teams messier. I have seen exactly that. The engineers who got better with AI were the ones who already wrote clear tickets, tested their assumptions and cared about the user.
What the engineer of 2030 looks like
My bet, based on what I do every day: the default engineer becomes a product builder who directs agents.
On OpenWhen I am the only human. More than 70,000 people have used it, and scheduled Claude Code agents do most of the routine work. Each agent works from a written charter with a kill switch and rules for when to stop and ask me.
What do I actually do? I decide what matters this week. I read the numbers. I write charters and skills. I review the pull requests that touch money, identity or user data. I talk to customers. Very little of my time is typing code.
That isn't a fantasy future. The Stack Overflow 2026 survey found the share of developers working as an "organization of one" rose from 4% to 10%. More people are running whole products alone because the tools now allow it.
This shift changes the skill stack:
| Old core skill | New core skill |
|---|---|
| Knowing syntax and APIs | Knowing what good looks like |
| Writing code fast | Specifying and checking work fast |
| Following the ticket | Questioning the ticket |
| Owning a component | Owning an outcome |
| Working alone at a keyboard | Running a small team of agents |
How to bet your career on it
If you are early in your career, here is what I would do. If you are senior, most of it still applies.
Get good at the boring fundamentals
You can't review what you don't understand. Learn how HTTP, databases, auth and browsers actually work. When an agent writes a SQL query, you should spot the missing index or the data leak. Fundamentals are what make AI output safe to ship.
Use AI every day, on purpose
Pick one agent (Claude Code, Cursor, Copilot) and use it for real work. Keep notes on where it fails. Write your own prompting rules. In my experience, the teams that get good with AI fastest are the ones that write their guidelines down and share them. Tools matter less than people think.
Own something end to end
Build one small product that real people use. Pick the problem, ship it, measure it, support it. One live product with ten real users teaches more than ten tutorial apps.
Learn to read numbers
Learn basic analytics: funnels, retention, conversion. Every feature decision I make on OpenWhen starts with a number and ends with a keep-or-revert rule I set before the test.
Practice writing
Specs, charters, pull request descriptions, user emails. Writing is how you direct both humans and agents. If your instructions are vague, your agent's output will be vague.
Learn to run agents safely
Start with Claude Code skills, then hooks and a written charter. The engineers who can run agents unattended and safely will be scarce for years.
What I'd do next
If the product-builder idea makes sense to you, read what a product engineer is and how to become one. If you are a junior worried about getting hired, read my honest take on whether AI will replace junior developers, which includes a 90-day plan.
Questions people ask
Will AI replace software engineers?
Not as a profession, based on current evidence. Stanford's 2026 payroll research finds no economy-wide displacement, but it does find fewer hires of young workers in AI-exposed jobs, so the entry path is narrower and the work is shifting toward judgment, verification and ownership.
What is the future of software engineering with AI?
Engineers will spend less time writing code and more time deciding what to build, specifying work for AI agents, reviewing output and owning results in production. The role is moving toward the product engineer: one person who owns a problem from user to metric.
Is software engineering still a good career in 2026?
Yes, if you aim at the parts AI amplifies rather than the parts it substitutes. Engineers who understand fundamentals, own outcomes and can run AI agents safely are in demand, while pure ticket-to-code work is shrinking.
What skills will software engineers need in the AI era?
Solid fundamentals so you can review AI code, product sense, basic analytics, clear writing, and the ability to set up and supervise AI agents with guardrails. Typing speed and syntax recall matter much less than before.
Do developers trust AI coding tools?
Mostly not yet. In the Stack Overflow 2025 survey, 46% of developers distrusted the accuracy of AI tools and only 33% trusted it, even though 84% were using or planning to use them.