Screening Contract Engineers in 2026: A Hiring Manager’s Playbook

The single most expensive mistake in contract engineering hiring is letting someone start work before you know they can do it. The next most expensive is letting your fear of that first mistake push you into a screening pipeline so cumbersome that the best contractors decline to engage. The point of a well-designed contractor-screening process is to find the narrow band where you have enough signal to commit without forcing senior practitioners through hoops that signal you don’t value their time.

We have been living inside this problem for the past nine months — partly because Basanite uses contract engineers ourselves, partly because contractor hiring is one of the cleanest commercial use cases for our platform, and partly because the AI-cheating wave has hit this category harder than any other. Below is what we have learned about screening contractors well in the current environment.

The cost equation, properly stated

Most companies dramatically underestimate what a bad contractor hire actually costs. The naive calculation is the day rate times the days they were on the team before you ended the engagement. The real cost includes:

• Onboarding overhead. A senior engineer at your company spends 4–8 hours per week for the first 2–3 weeks getting a contractor productive. Assume £80/hour fully loaded; that is £640–1,920 of internal time you do not recover.
• Code-quality remediation. Work shipped by a misaligned contractor in the first two weeks rarely gets reverted; it gets refactored. The refactor work is invisible because it shows up as routine engineering time, but it is real and costs another 10–30 internal engineer-hours.
• Trust-rebuild on adjacent decisions. Engineers on the team start spending more time reviewing the contractor’s work, asking more questions in pull requests, and second-guessing architectural choices the contractor proposed. The drag persists for a week or two after the contractor leaves.
• The direct day-rate cost. Senior UK contractors are £600–900/day; senior US contractors are $800–1,500/day. A two-week paid trial that ends in a no is £6,000–18,000 of day rate alone.

Aggregated, a single failed senior-contractor trial week typically costs £10,000–25,000 once you account for everything. Companies running 20 trials a year to fill 10 contractor seats are spending £100,000–250,000 a year on failed trials. The cost is real even when it doesn’t show up as a line item.

Why the traditional triage steps have broken

The standard pre-trial triage funnel has been: portfolio review, take-home test, technical phone screen with a senior engineer. Each step has degraded.

Portfolio review is now nearly useless as a stand-alone signal. GitHub profiles are inflated with AI-generated side projects. Personal websites are themselves AI-generated. Case studies on a contractor’s site can be entirely fictional with no detectable provenance. We are not saying these are common — most contractors are honest — but we are saying that you cannot use portfolio quality as an independent signal because the variance is now dominated by how much AI assistance went into the portfolio rather than how skilled the contractor is.

Take-home tests are the cleanest example of the cheating wave. We covered the underlying tooling in detail in our piece on AI cheating in interviews. A take-home calibrated for four hours of solo human effort can be completed by a modern agent in 15–45 minutes. Reviewer judgment has not been recalibrated for this. Reviewers are still rating AI-generated take-homes highly because the code looks good — clean, tested, documented. The grading signal has become almost orthogonal to the underlying capability.

Technical phone screens with a senior engineer still work, but they are expensive. A 45-minute technical screen consumes 60 minutes of senior engineer time (15 minutes of context-switch and write-up overhead). At £80/hour fully loaded, that’s £80 per screened candidate. Companies running 30 screens to fill a single contractor seat are spending £2,400 in screening cost per hire, and that is before any trial-week cost.

The Toptal alternative

The dominant outsourced solution to this problem is Toptal — or, more accurately, Toptal and its category competitors (Andela, Turing, Arc, the niche specialist marketplaces). The deal is roughly: they screen, you pay 30% over what the contractor would charge directly, and the marketplace absorbs the screening cost.

The Toptal model works for some companies. The trade-offs to understand before signing:

• You pay 30% forever, not just for the screening. The marketplace takes its cut for the entire duration of the engagement. A 12-month £900/day contract through Toptal costs you £900 × 250 days × 1.30 = £292,500 instead of £225,000 direct.
• The screening is generic. Toptal screens for general engineering competence, not for whether the contractor fits the specific role at your company. The signal floor is decent but the signal-to-noise ratio for any individual hire is not high.
• You don’t own the relationship. A contractor you sourced direct can be hired full-time later for the cost of a finder’s fee. A Toptal contractor cannot be poached without paying Toptal a buyout, which is often a six-figure sum for senior roles.
• The pool is curated, not exhaustive. The best 5% of contractors mostly don’t need marketplaces, and the worst 50% can’t pass the marketplace screen. You’re fishing in a narrow band.

For small companies hiring one contractor a year, paying the 30% is often the right call — the cost of building an internal screening pipeline isn’t justified. For companies running ongoing contractor engagement, the 30% compounds fast enough that doing it internally pays back inside the first 4–6 contractor hires.

The third path: rigorous direct screening

The case for doing it yourself rests on whether you can build a screening process that hits three properties at once.

1. Robust to current AI cheating tooling. The screening signal needs to survive a candidate using Cluely or Interview Coder. We covered the specific tooling in our piece on interview-coder detection.
2. Cheap enough per candidate to be a real triage filter. If each candidate costs £200 to screen, you can afford to screen 30 candidates for £6,000 to find one good hire. If each candidate costs £80 in senior engineer time, you can only afford to screen 8.
3. Tolerable enough that strong candidates engage. Senior contractors decline interviews with hoops that signal you don’t respect their time. Anything that takes more than 60–90 minutes of their unpaid time before you commit money is a hard sell to the top of the market.

The design that survives all three constraints, in our experience, is a two-round structure roughly mirroring what we use at Basanite:

Round one: 20–30 minute structured voice conversation. Adaptive questions grounded in the contractor’s actual CV. Probes specific past projects. Cheap to deliver if you can automate it; £5–15 per candidate at scale. Surfaces the cognitive, judgmental, and tacit-knowledge dimensions that distinguish senior contractors from confident generalists. We discussed why this design is robust to AI assistance in our piece on CV-grounded interviews.

Round two: 60–90 minute instrumented coding task. A realistic ticket against a representative codebase. The contractor uses their own AI tooling — Cursor, Claude Code, Copilot, whatever they actually work with. Instrumentation captures their prompts, their verification behaviour, and their override judgment. A short reflection conversation at the end cross-references what they did against what they understand themselves to have done. This is the design pattern we call the AI Collaboration Workbench, and it is what gives you a defensible read on whether a contractor can actually ship work in your codebase rather than just having clean GitHub commit messages.

Together, the two rounds take around 90 minutes of the contractor’s time and 5 minutes of yours — the report you receive is the basis of a 30-minute final human conversation where you sanity-check fit and discuss the engagement. The total fully-loaded cost is around £40 per candidate. At 30 candidates per hire, that is £1,200 of screening to avoid £10,000–25,000 of failed trial-week risk. That ratio works.

What to actually evaluate

The dimensions that matter for a senior contractor are not the dimensions that matter for a permanent hire. A contractor needs to be productive in week one. The signals that matter:

• Calibration. Can the contractor estimate task complexity accurately, scope appropriately, and admit uncertainty when they hit it? Bad contractors over-promise and produce shocked “it’s harder than I thought” messages in week three.
• Code-base archaeology. Can they read existing code well enough to make changes that fit the codebase’s prevailing style? Contractors who refactor your codebase to match their preferred style on day three are a known failure mode.
• Communication cadence. Do they surface blockers early, ask the right questions, and write pull-request descriptions that don’t require a meeting to understand?
• AI orchestration. Can they direct an AI agent productively without either over-trusting it (shipping hallucinated solutions) or under-using it (re-doing work the agent could have done in 20% of the time)? This is now table stakes for any senior contractor in 2026. We dug into the broader pattern in hiring engineers who can use AI.
• Verification rigor. Do they test their own work before declaring it done? Contractors who ship to production and then say “I couldn’t reproduce” when the bug appears are an expensive failure mode.

Notice what is not on this list: leetcode-style algorithmic puzzle solving, knowledge of obscure language trivia, system-design from cold start. Those things matter for permanent senior hires being considered for staff and principal roles. They don’t matter for whether a contractor can pick up your repo on Monday and ship a ticket by Friday.

The 90-minute screening sketch

For a hiring manager building this internally, the minimum viable version of a robust contractor screening process looks like:

First, run a 20-minute structured voice conversation. Either deliver it yourself (one senior engineer’s morning, costs £300/day in their time) or use an AI interviewer. The questions probe the contractor’s most recent two projects, the trade-offs they faced, and a specific decision they would do differently in hindsight.

Then, give the contractor a 60-minute coding task against a small representative repo. Be explicit that AI use is required, not optional. Instrument what you can — at minimum, get a screen recording with audio narration where the contractor talks through their reasoning as they work.

End with a 10-minute reflection call. Pick two decisions visible in the recording and ask the contractor to explain why they made them. Genuine contractors explain fluently from working memory. Substitutes — yes, we have caught two — pause, hedge, and produce answers that don’t match the recording.

Total contractor time: 90 minutes. Total your time: 40 minutes spread across the conversation, the recording review, and the reflection call. Cost: £50 in your time, £15 in tooling. Output: a defensible go/no-go decision before you commit a single day of trial-week budget.

Where this leaves you

The companies that get contractor hiring right in 2026 are the ones that have stopped trying to make 2023’s screening stack work and have built an explicit two-round structure tuned to the new environment. The choice between building it yourself and outsourcing to a marketplace is mostly a function of volume — at fewer than 4–5 contractors a year, pay the marketplace; at more than that, the math favours direct screening.

If you want to read about the underlying design principles in more depth, our piece on AI-resistant interview design is the technical companion to this one. If you want to see what the Basanite version of this looks like, the FAQ covers the mechanics, and the homepage covers the pitch.