.webp)
The Interview
What Does a Remote Interview Look Like Now?
Remote engineering interviewers will recognise a pattern. The candidate repeats the question, then gives an accurate but general answer that would suit any team at any company. It proves nothing, since candidates may simply be nervous, and Code Clan's interviewers see it too.
Code Clan's interviewers usually respond by changing the next question. Instead of a general deployment problem, they ask how it would play out on the candidate's current team and role. Someone describing their own work can usually answer from experience, while a generic answer has little to attach to.
Anthropic, which builds AI, redesigned its take-home test more than once after one model outperformed most applicants within the time limit. A later model matched the best human result in two hours (Anthropic, January 2026).
Both Sides
What Changes When Both Sides Use AI?
CVs, take-home tests, and polished interviews are now cheap to produce or improve, so each carries less weight as evidence. Most candidate AI use is ordinary, such as drafting a CV or preparing for an interview.
In Gartner's late-2024 survey of 3,290 candidates (Gartner, July 2025):
- 39% used AI in applying; of those, 54% used it for CV text and 50% for cover letter text
- 52% believe AI screens their application
- 26% trust AI to evaluate them fairly
A forecast from the same release suggests that by 2028 one in four candidate profiles worldwide will be fake. It is a forecast and not a measured rate, and it concerns profiles, so it says nothing about the many candidates whose profiles are accurate.
The sensible response is to look for evidence that doesn't rest on documents alone. The candidate's own history is a more useful guide, including what their references say about their projects and how they talk through that work live. A person who has done the work can usually do that comfortably.
AI also gives employers real benefits, and the sections below treat both sides fairly, across four stages: matching, vetting, contract and payroll, and onboarding.
Stage 1 · Matching
Stage 1: Where Does AI Help with Matching, and Where Is It Overreaching?
AI matching reads a written requirement quickly and applies it the same way to every profile, which makes it a useful first filter. Like any matching, it relies on the data it is given: the candidate's profile and CV.
Where AI Helps
A recruiter can review hundreds of profiles in the time a person reads a dozen, and the shortlist doesn't depend on who was tired on a Friday afternoon. Code Clan's matching uses a proprietary, evolving algorithm that searches and matches in several layers, one based on AI.
Where It Falls Short
- A strong engineer with thin profile data can be placed lower than their experience deserves.
- A profile polished with AI help to mirror a job description can outrank the person behind it, even when their real skills and qualifications are superior.
Both come from matching on documents, whichever tool does the matching.
What a Person Still Checks
At Code Clan, a CV is compared by hand with other historical records, such as LinkedIn, other databases, and the candidate's application history. That turns a strong match on paper into one that has been checked.
For roles open for months, Code Clan's hard-to-fill engineering roles service combines its global talent network with its hiring process, so the search doesn't rest on one filter.
Stage 2 · Vetting
Stage 2: What Can an Interview Still Prove?
A polished interview used to be reasonable evidence of ability. It now needs a second layer of checking, because several of the things that made it persuasive can be produced with help. An unsupervised take-home exercise measures what a candidate and an AI tool can produce together.
The Pragmatic Engineer surveyed 63 interviewers in 2025. Of those at large tech companies, 81% suspected candidates of using AI, and 31% said they were certain a candidate had used AI tools and presented the answers as their own.
Among Big Tech interviewers, 58% said they had changed their questions (The Pragmatic Engineer, September 2025). The interviewers are self-selected users of one mock-interview platform, so the results indicate a pattern without measuring it.
Code Clan sees something similar: in roughly one in five of its interviews, interviewers see signs consistent with AI assistance. This is Code Clan's own observation, not a research statistic, and it applies across roles with none standing out.
The signs include:
- long pauses
- a question repeated before it is answered
- eyes moving to a second screen
- code that is written and runs successfully in one go
Each has ordinary explanations, so interviewers treat them as a reason to ask more, not as a conclusion.
In a randomised trial by METR, 16 experienced developers completed 246 tasks, and tasks where AI tools were allowed took about 19% longer. Their own estimates of the effect, before and after, pointed towards a speed-up (METR, July 2025).
It covered experienced developers on their own code with early-2025 tools, so it says nothing direct about hiring, but it supports asking candidates to explain their reasoning and not only to produce a result.
Identity is a separate question from skill: Code Clan uses ID and address verification and reference checks.
One remedy is to meet in person, and Google, Cisco, and McKinsey have reportedly reinstated in-person interviews for some part of their hiring (Axios, August 2025). It is hardest when the engineer lives in another country, as in offshore hiring.
Transcripts help modestly: a second reviewer can re-read an interview and check what was said. At Code Clan, they are recorded, summarised, and used for further analysis once the interview has finished.
What a person still checks is depth:
- follow-up questions on past work
- live problem-solving
- asking the candidate to explain why they made a choice
When written signals are cheap to produce, an experienced interviewer asking follow-up questions is the check that remains.
When written signals are cheap to produce, an experienced interviewer asking follow-up questions is the check that remains.
Code Clan's vetting assesses technical capability, communication, reliability and real-world execution. For businesses without an in-house technical leader to run interviews, its service for hiring without internal technical expertise covers defining requirements and evaluating candidates.
Its failed hire replacement service helps when a hire that looked strong on paper hasn't worked out.
Stage 3 · Contract & Payroll
Stage 3: What Happens to Contracts and Payroll if the Details Don't Match the Person?
When a detail doesn't match, the employer and the engineer both take on administrative and compliance work to correct it. Contracts, payroll, tax, and employment obligations depend on knowing who the engineer is and where they work, and they vary between jurisdictions.
Details can drift for ordinary reasons too: people move, change roles, or hold out-of-date records.
AI can draft documents and process records quickly, which is useful for contract and payroll administration. Speed applies equally to correct and incorrect details, so a person confirms the details before they are locked in.
Speed applies equally to correct and incorrect details, so a person confirms the details before they are locked in.
Code Clan has come across candidates who appeared to be impersonating someone else or misstating their location, and cases like these are detected and filtered at early stages.
Code Clan manages payroll, contracts, compliance and tax for engineers hired across international markets. Its international team expansion service suits companies that want global talent without becoming experts in each jurisdiction's employment rules.
Stage 4 · Onboarding
Stage 4: How Do You Confirm the Right Person Is Doing the Work After Day One?
The person who interviewed has to be the person who does the work, so verification doesn't end at the offer.
Code Clan's onboarding includes general training and field-specific training, covering:
- privacy
- customer data
- work ethics
- the code of conduct
Staged access, early check-ins, and a short live work task are normally handled before engagement starts.
Some mismatches only show once work begins. A hire can match the role well and do very well in vetting, yet have limited capacity or commitment, for example because of another job. That is hard to see in advance, because they look like exactly the right fit.
Many engineers work more than one role, and that is not a problem in itself. What matters is whether the client's role gets the contribution it was hired for, and Code Clan's systems flag low contribution in the first month.
AI speeds up onboarding with documentation, codebase summaries, and routine answers. An alert can show that contribution is low, and a person judges commitment and decides what to do about it.
For businesses extending an engineering team across borders, Code Clan's scaling teams service covers ongoing management after the hire.
At a Glance
What Should You Check at Each Stage?
The table below pulls the four stages together.
Looking Ahead
What Could Change Next as AI Models Improve?
Each new generation of AI models tends to handle more of the work that candidates are asked to do. Three developments from credible sources follow.
Fixed-length exercises are likely to get easier for AI. METR reports that the length of software task leading models complete with 50% success, measured by how long a skilled person takes, has roughly doubled every seven months since 2019.
The authors discuss limits beyond the self-contained software tasks they tested (METR, July 2026). A fixed take-home test is likely to be solved by AI tools more often over time.
Employers are also likely to look for AI skills directly. In IBM's survey of 1,500 chief human resources officers and equivalent executives, 71% named the ability to supervise, validate, and override AI outputs as the workforce's most essential skill (IBM, September 2026).
Globally, postings requiring at least one AI-related skill rose by about 69% between 2024 and 2025 (PwC, June 2026). If that continues, candidates may be asked to show how they work with AI, a different question from whether they used it.
AI-led interviews are also being studied. In a field experiment with about 70,000 applicants, those randomly assigned to an AI voice interviewer were 12% more likely to receive a job offer than those interviewed by a human recruiter.
The gains carried through to job starts and retention, and human recruiters still evaluated the interviews and made the hiring decisions (Jabarian and Henkel, September 2026). It is a single study in one hiring setting, so not a general rule.
Candidates are likely to want advance notice when an AI system is part of an interview, though that is our expectation and not a measured result.
Some of today's signals will weaken as tools improve, and some human checks will matter more. Long pauses and a second screen are visible now and may be harder to see later.
Checks that rest on a person's own history, such as references, a conversation about their current team, and cross-checkable records, don't depend on spotting a tool. This is our expectation and not a measured result.
If you're weighing a remote or offshore engineering hire and want to talk it through, Code Clan offers a discovery call to see whether it's a good fit.
FAQ
Frequently Asked Questions
Can AI Interview Tools Be Trusted to Vet Engineers?
They help with consistency and record-keeping, such as transcripts a second reviewer can check. They can't judge depth of understanding or confirm identity, so a person needs to review the interview.
Is a Take-home Test Still Worth Setting?
It can be, with a live follow-up where the candidate explains their choices. Anthropic redesigned its test and didn't ban AI tools, which shows why an unsupervised exercise alone is a weak signal.
How Can a Company Hire Offshore Engineers Safely?
Check at each of the four stages, not just one. Cross-check work history after matching, test for depth in interviews, have a person confirm contract and payroll details before they are set, and watch contribution in the first weeks.
How Does Code Clan Vet Engineers?
Code Clan matches candidates in detail against the job description, runs hard and soft skill assessments, and carries out reference and ID verification. Its vetting assesses technical capability, communication, reliability and real-world execution.
