How to Hire Software Engineers in the AI Era
What to change at intake, sourcing, assessments, interviews and identity checks now that engineers write code with AI tools

Recruiters hiring software engineers today are meeting candidates who may write much of their code with AI coding assistants. That changes what a take-home test can tell you. It also raises a question many hiring teams haven't settled: what a strong engineer looks like when a tool handles part of the routine work. For remote roles, face-swapping tools add one more thing to confirm.
Hiring software engineers in the AI era means running the familiar stages from intake to offer while deciding, at several points, how AI use fits in. Those decisions affect whether you can compare assessment scores and whether interviews show how a candidate thinks. They also affect how confidently you can present someone to a hiring manager or client.
Key takeaways
- Ask the hiring manager or client which AI tools the team uses and allows, because the answer shapes the posting, the assessment and the interview.
- Source for work engineers have built and shipped, since an "AI" keyword on a profile can mean anything from daily tool use to building models.
- Choose one AI rule for the technical assessment and tell candidates what's allowed before they start.
- Run the same identity checks for every remote candidate, and treat a single odd signal as something to clarify rather than proof of fraud.
Are software engineers still in demand?
Yes, demand for software engineering work is projected to keep growing, although the outlook for early-career workers in AI-exposed fields is less clear. The BLS Occupational Outlook Handbook counts 1,905,400 jobs in 2025 for software developers, quality assurance analysts and testers, and projects 10% growth through 2035. The handbook calls that much faster than average and links part of the demand to software development for AI. It puts the group's 2025 median pay at $134,040 a year.
BLS economists are more cautious about how AI itself will reshape the work. In a February 2025 Monthly Labor Review article, they described AI's effects on software developer jobs as highly uncertain. They argued that productivity gains could lower software prices and increase demand. They concluded that employment in the occupation is unlikely to decline, which is a projection rather than a measured result.
Early-career workers show a different pattern. A Stanford working paper revised in August 2026 found that employment among workers aged 22 to 25 in AI-exposed occupations was 19% below its expected trend. That trend is what it would have been had it kept pace with less-exposed peers, as measured by ADP payroll data. The authors attribute the gap mainly to reduced hiring of young workers, report no comparable gap for experienced workers, and describe their findings as descriptive rather than causal.
The paper covers AI-exposed occupations in general rather than software developers specifically, so treat it as a signal to watch rather than a forecast for engineering roles. For experienced engineers, the BLS projections point to continued demand. Whether junior openings are shrinking in your market is a fair question to raise with clients planning their seniority mix.
What's different about each stage of engineering hiring
The stages you use to hire software engineers haven't changed, but AI coding tools affect what you ask, test, and check at several of them. The graphic shows each stage, from intake to offer, what has changed, and what to adjust.
Intake carries the most weight because the team's position on AI shapes the posting, assessment, and interview loop that follow. Assessments and interviews change together, since a task completed with AI help says more once the candidate explains how they checked it.
What to ask at intake when hiring software engineers
A hiring manager or client may have a clear view of the technical stack but a vaguer one of how their engineers use AI day to day. Asking about it at intake gives you what you need to write the posting, choose an assessment approach, and brief interviewers. Useful questions include the following:
- Tools in use: Which AI coding assistants does the team use day to day, and are any restricted for security or licensing reasons?
- Shape of the work: How much of the job is writing new code, and how much is reviewing, debugging, integrating systems, or designing them?
- Review and security expectations: Who reviews AI-generated code, and what does the team expect an engineer to check before it's merged?
- AI features: Will this engineer build AI features into the product, or use AI tools only as part of daily work?
The second question is where teams differ most. A team that spends much of its time reviewing and integrating code needs evidence of careful reading and testing. A team building new features may care more about design choices, and knowing which one you're hiring for tells you what the assessment should measure.
Seniority deserves its own conversation. If the team still plans to hire early-career engineers, ask what support they'll get, such as code review, pairing, or a named mentor. That tells you what a junior candidate can reasonably be expected to know on day one. Write the answers into your ideal candidate profile so everyone screening candidates works from the same picture.
Put the same answers in the posting. A clear job description can name the team's tools, say whether AI is allowed in the assessment, and describe the mix of building and reviewing work. Agency recruiters may find that a client hasn't decided on an AI policy, and the intake call is a natural place to raise it.
Sourcing software engineers when titles and keywords shift
Recruiting software engineers now works better when you search for the work people have done, because AI-related skills show up under many different titles. An engineer building features on large language models might be listed as a backend, platform, full-stack, or machine learning engineer. A search built around one title will catch only some of them.
Shipped products, open-source contributions, internal tools described on a profile, conference talks, and technical write-ups all show what someone has built. Searching these signals overlaps with skills-based hiring, which assesses people on demonstrated skills rather than titles or credentials.
An "AI" keyword on a profile tells you less than it seems to. It might mean the person uses a coding assistant every day, fine-tunes models, or built one prototype at a hackathon, and the profile rarely says which. Treat the keyword as a topic for your first conversation, along with whether the person is open to a move at all.
Decide the AI rules before you start technical assessments
Before any candidate sees a technical assessment, agree with the hiring manager or client on whether AI is allowed. Then tell candidates the rule in writing ahead of time, because scores aren't comparable if some assumed AI was allowed and others didn't. The table compares three ways to handle AI in a technical assessment, what each shows, and its main drawback.
| Approach | What it shows you | Trade-off |
|---|---|---|
| AI not allowed | Unassisted coding and problem solving | Hard to enforce outside a live session and less like daily work on teams that use AI |
| AI allowed and discussed | How candidates use tools and explain their choices | Needs a follow-up conversation to see the reasoning |
| AI part of the task | How candidates direct AI and check its output | Needs a clear rubric for judging the review work |
The intake answers point toward one of these. A team that restricts AI tools for security reasons has a good case for an AI-free assessment, because it mirrors the job. A team that uses assistants every day will learn more from seeing how candidates work with them.
Take-home exercises are the hardest format to keep AI-free, because nobody sees how the work was produced. If the team wants to see unassisted coding, a short live exercise is more practical. If AI is allowed, schedule a follow-up where candidates walk through their solution and describe any AI suggestion they changed or threw out.
Give candidates a way to request accommodations, such as extra time or a different format. Also check that an AI rule doesn't block assistive technology a candidate relies on, and involve HR or legal when your process needs a full accessibility review. Designing a technical assessment also means settling format, length, and scoring before the first candidate starts.
Spend assessment time on the right engineers
Book a demoHow to interview engineers on AI-assisted work
Interviews are where you learn how an engineer checks and explains AI-assisted work, and self-reported speed is a weak guide to that. In a July 2025 METR study, 16 experienced open-source developers took 19% longer with AI tools but afterward believed AI had sped them up by 20%. They worked on 246 real issues from large open-source projects and had expected to finish faster, not slower.
The study has clear limits, and METR called it a snapshot of early-2025 tools in one setting. In a February 2026 update, METR said AI tools are likely speeding developers up more now. It also called its new data only very weak evidence of how large that change is.
The research supports a narrower point for interviewing: self-reporting. A candidate's impression of their own productivity may be inaccurate, so concrete accounts of recent work give interviewers more to evaluate. Questions that draw those out include:
- A recent change made with AI help: What was the change, and which parts did the tool write?
- How they checked it: What tests, reviews, or manual checks did they run before it shipped?
- What they rejected: Which AI suggestion did they decide not to use, and why?
- A bug the tool introduced: How did they find it, and what did they change in their process afterward?
Asking every candidate the same questions, as a structured interview does, makes the answers easier to compare across a shortlist. Score them against an interview scorecard that describes what a specific, well-reasoned answer looks like, so interviewers rate substance rather than confidence.
Verifying identity in remote engineering hiring
When you hire software engineers for remote roles, confirm identity at several points rather than only at the offer. In a January 2025 public service announcement, the FBI described North Korean IT workers observed using AI and face-swapping technology in video job interviews. Its recommendations apply to any organization hiring remote workers:
- Verify identity throughout: Confirm identity during interviews, onboarding, and throughout employment for any remote worker.
- Cross-check applicants: Look in HR systems for other applicants with the same resume content or contact information, including reused phone numbers and email addresses.
- Meet in person where possible: Complete as much of the hiring and onboarding process in person as the role allows.
- Train the people involved: Educate HR staff, hiring managers, and development teams about the threat.
The warning describes one specific threat, so it's a reason for consistent checks rather than suspicion of individual candidates. A frozen video feed, a preference for keeping the camera off, or a shared phone number can each have an ordinary explanation. A weak connection, discomfort on camera, or a family phone plan can all cause it. When something doesn't match, document it and ask about it directly, the same way you'd clarify a gap on a resume.
The announcement also advises employers to confirm that third-party staffing firms follow sound hiring practices. Agency recruiters can expect clients to ask how they confirm identity before presenting a candidate, so agree with the client on which checks happen at which stage.
Start with the intake before you change the interview loop
Most teams that hire software engineers won't need to rebuild their process, since these changes fit into the stages you already run. Begin by adding the AI questions to your next intake call and writing the assessment rule into the candidate instructions. Then add a few questions about AI-assisted work to the interview guide. For remote roles, agree on identity checks before booking the first interview.
Search for engineers by the work they do
Book a demoFrequently asked questions
Is it harder to hire junior software engineers now?
It can be harder to find openings for them, since the Stanford research above found reduced hiring of young workers in AI-exposed occupations. Where junior roles exist, ask whether the team expects new engineers to review and test AI-generated code as well as write their own. Then calibrate the assessment to that expectation.
What should a software engineer job description say about AI tools?
It should say which AI tools the team uses, whether candidates may use AI in the assessment, and whether the role involves building AI features. Plain descriptions of the work help candidates more than a list of AI buzzwords, and they set expectations before the first interview.
How much does it cost to hire a software engineer?
Pay is the cost that continues after the hire, and the BLS median pay figure covered above is a national starting point for the combined occupation group. Pay varies by location, seniority and specialty, so check current ranges with salary benchmarking tools before you agree on a range. Agency fees, assessment tools and interviewer time add to the total when you hire software engineers.
About the author

Founder & CEO of Rotto, building tools that help tech recruiters source better candidates, faster.





