The American workforce no longer fits between 9 and 5.
According to the most recent Bureau of Labor Statistics figures, about 14% of full-time wage and salary workers — roughly 15 million people — usually work a non-daytime schedule. Across all wage and salary workers, full- and part-time, the figure rises to 16%. Add the 36% of workers McKinsey now classifies as participating in independent or supplemental gig work, the millions of working parents managing daycare-bound schedules, and the professionals whose calendars are full of back-to-back meetings from 9 until the dinner hour, and the population of people who can't reliably take a phone call between 10am and 4pm Tuesday through Thursday represents at least a third of the workforce — depending on overlap and definition.
Recruiting hasn't moved.
The default workflow is still: post the job, collect applications, have a recruiter call candidates between 10am and 4pm Tuesday through Thursday, leave voicemails for the no-answers, hope for callbacks during the same window. Anyone whose life doesn't fit those hours gets ghosted. Anyone the recruiter doesn't reach within 48 hours self-removes from the funnel. The result is a hiring system that screens out the candidates most likely to actually show up to work — because the act of being available for a 2pm phone call is itself a screening signal, and it's screening for the wrong thing.
There's a name for what a different system can do: Inclusion Arbitrage.
The arbitrage is this. The candidates who can't take a 2pm phone call aren't worse candidates. In many cases they're better — they're already employed, already busy, already managing complex lives. They're not unreachable. They're just unreachable during the hours your recruiter happens to work. Any company that can engage them on their schedule captures a talent pool the rest of the market is structurally blind to. The labor isn't hidden. The willingness to work isn't hidden. What's hidden is just the time slot.
This isn't theory. We watched it happen in a recent pilot.
What 53 Inbound Calls Revealed
A medtech company opened a Catheter Assembler role in March 2026. The ad ran on Indeed and pulled 123 applications in three days. We ran AI voice pre-screens against all 123. The platform called outbound. For anyone who didn't answer, it left a voicemail with a callback number and a code so they could complete the screen on their own time.
What happened next is the part worth paying attention to.
Of the 123 candidates, 70 received an outbound call from the platform. 26 of them completed the screen during that initial call. A 37% completion rate on outbound — roughly what every recruiter operating today already lives with. The rest didn't pick up, said they couldn't talk, or got partway through and bailed.
Then the inbound calls started. 53 candidates called back on their own schedule. 52 of them completed the full screen. A 98% completion rate.
Same job. Same questions. Same AI on the line. The only variable that changed was who chose the moment.
The 2.6× lift between outbound and inbound completion is the operational signature of a structural pattern: when you remove the scheduling tax, the candidates who self-select in are dramatically more committed to finishing what they started. They already decided to give you the next twelve minutes of their life. They're not negotiating with their kids in the next room or looking at the clock wondering if they'll be late picking up from school. They picked the moment because the moment worked.
Eight of those 78 completed pre-screens landed after 6pm Toronto. One wrapped just past midnight.
The Candidates Who Don't Answer at 2pm
Who calls back at 11:47pm to talk about a manufacturing job?
Not the unmotivated. The unmotivated already self-selected out at the application step. The 11pm callback is a candidate who saw the voicemail at 5:30pm during a shift change, decided they wanted the job enough to follow up, made dinner, put the kids to bed, sat down on the couch, and pulled the slip of paper with the callback code out of their pocket.
That candidate is already telling you everything you need to know about their work ethic. The traditional workflow never hears from them.
Who calls back at 6:45pm on a Thursday? The candidate who just finished an 11-hour day at a job they're underpaid at and is actively looking to move. Who calls back at 8pm? The single parent who just got the toddler down. Who calls back during a 12:15pm break? The hourly worker whose shift starts at 6am and who is one of the most reliable candidates you could possibly hire.
These are not edge cases. They represent a structural share of the workforce — 14% on non-daytime full-time schedules per BLS, 27% of part-time workers, the 36% participating in independent or gig work per McKinsey, the working parents and shift-rotators and professionals whose calendars don't bend to a recruiter's 2pm slot. Across overlapping definitions, the population easily exceeds a third of everyone you'd want to hire.
Your hiring funnel either captures these people or it doesn't. The default funnel doesn't.
Why This Is Arbitrage and Not Just Access
It's tempting to read this as "AI lets us screen more candidates, which is nice." That misses the actual claim.
Arbitrage is a price difference between two markets for the same asset. The asset here is a hireable human being. The two markets are the daytime workforce (where every recruiter is fishing in the same pond, paying agency fees, getting ghosted, losing offers to counter-offers) and the everyone-else workforce (where almost no one is fishing because reaching them costs more than the average recruiter desk can afford).
The price difference is enormous. In the daytime market, a Catheter Assembler in Kitchener with three years of experience commands a premium because every other manufacturer is trying to hire the same person. In the everyone-else market, that same Catheter Assembler — perhaps already employed across town and willing to switch for the right shift — is functionally invisible because no one is reaching them at 7:30pm on a Tuesday when they could actually take the call.
The arbitrage closes the moment the rest of the market figures out how to fish in the second pond. It hasn't yet. There's a window — possibly several years long — in which any company willing to deploy structured AI pre-screening with 24/7 availability captures candidates the rest of the market literally cannot see.
This is why we're writing about it instead of holding it close. The arbitrage isn't going to last forever. The companies that move first don't just get better access. They build a hiring funnel that's operationally larger than their competitors' funnel, while their competitors are still posting jobs at 11am Wednesday and wondering why nobody's calling back.
A note on the quality question: the operational claim — that async, structured pre-screening produces dramatically higher completion rates and reaches candidate populations the synchronous workday misses — is supported by industry benchmarks and by the data from our own pilots. The stronger claim — that this produces demonstrably better hires — is the hypothesis we're testing in the field, not yet a settled result in the peer-reviewed literature. Anyone making the strong-form claim today is selling more than they can prove. We're making the operational claim with confidence and the quality claim with appropriate humility.
What "Doing It Right" Actually Requires
The naive version of this argument — "use AI to screen candidates 24/7" — is missing most of what makes Inclusion Arbitrage work. There are four pieces, and you can't skip any of them.
One: structured questions, not chatbot vibes. Every candidate gets the same script, scored against the same framework. The current best meta-analytic estimate (Sackett, Zhang, Berry & Lievens, 2022) places structured interviews at r = 0.42 against actual job performance — the highest validity of any single selection method when methodological corrections are properly applied. Unstructured interviews land at r = 0.19. The gap is the structure. (The older Schmidt & Hunter 1998 figures of 0.51 and 0.38 are still widely quoted, but were revised downward when the original range-restriction corrections were shown to be systematically overstated.)
What that 0.19 figure means in practice: the unstructured phone screen is the lowest-validity selection method in widespread use. Better than nothing, but the weakest signal in the recruiter's toolkit. Replacing it with a structured equivalent is one of the highest-leverage moves available to any hiring team — and it's one of the few moves the I/O psychology literature has been agreeing on for forty years.
Two: voice, not text. This is non-obvious and worth dwelling on. The candidates Inclusion Arbitrage captures often can't easily complete a text-based screening flow — they're driving home, they're in a noisy break room, they're putting kids to bed with one hand. Voice meets them where they are. Text loses them. Most "AI screening" platforms are text-first because text is easier to build. They're solving a different problem.
Three: callback codes, not pure outbound. If the only mode is "the platform calls you," you're back to solving the recruiter scheduling problem with an AI instead of a human. You haven't moved the ball. The unlock is the inbound channel: a number the candidate can call, a code they can enter, a screen they can complete on their schedule. That is what flips the 37% completion rate to 98%.
Four: structured reports, not transcripts. A 12-minute audio file isn't a screening output. The output is a structured report with the candidate's responses to each question, their stated qualifications, their availability, the recommendation tier the AI placed them in based on the rubric, and whatever flags the system surfaced for human review. The recruiter still owns the decision. The AI just ensures the recruiter has the same information about every candidate, generated in seconds, available to compare side by side.
If you're missing any one of these four, you're not doing Inclusion Arbitrage. You're doing some adjacent thing that probably has its own merits but won't generate the 2.6× lift, won't capture the 11pm callback, won't surface the candidate the rest of the market can't see.
What to Watch For When You Try It
A few patterns we've seen across the pilots we've run. None of these are published industry benchmarks — they're our observations, offered as provisional reference points for anyone running their own first pilot.
Inbound completion will dwarf outbound completion. The lift won't be exactly 2.6× — that was one pilot — but the direction is structural. If you don't see at least a 1.5× lift in your own data, something's wrong with your callback flow. Probably the callback code is too long, or the IVR is asking the candidate to verify their identity before connecting them to the screen, or the prompt isn't clear that they can hang up and try again later.
The after-hours bucket will be smaller than you expect, and more concentrated than you expect. Across a single mid-volume role, you might see 8–12 after-hours completions out of 70–80. That sounds modest. The pattern matters more than the count. Almost all of those after-hours completions will cluster between 6pm and 10pm on weekdays, with a long tail extending past midnight and into weekends. The candidates in that long tail are disproportionately the ones whose lives are most structured around someone else's schedule — which often correlates with the candidates most worth meeting.
Your "Hold" tier will be larger than recruiters predict. When recruiters do unstructured phone screens, they advance roughly 65–75% of candidates to a second conversation, because unstructured screens give candidates a thousand subtle ways to talk their way through. Structured AI pre-screens with role-specific technical questions are unforgiving. If a candidate doesn't have direct experience with the manufacturing environment the role requires, they can't bluff their way past a question about cleanroom protocols. Across our pilots, we've seen Hold tiers cluster in the 35–45% range. This isn't the AI being mean. It's the AI being honest about a fact recruiters had been politely ignoring.
Your hire will probably come from the Review tier, not the Proceed tier. This surprises people. The Proceed tier is small (5–10% in most pilots), and Proceed candidates often have other options. The Review tier is where founder judgment adds the most value — it's the queue of candidates with mixed signals, where a 15-minute conversation can resolve the ambiguity in either direction. In the Vena pilot, the eventual hire was a candidate who landed in the Review tier on the AI's evaluation. The AI didn't make the call. It just made the founders' shortlist visible.
The Larger Argument
If you take only one thing from this piece: the 9-to-5 phone screen is no longer a neutral filtering mechanism. It's an active selection bias against the modern workforce.
This isn't primarily a moral argument, though there is a moral argument to be made. It's an economic one. The hiring funnel that excludes parents, shift workers, immigrants holding two jobs, and professionals whose calendars are full from 9 to 6 is a hiring funnel that's leaving money on the table. The candidates the funnel excludes aren't worse. They're often demonstrably different — already employed, already managing complex schedules, already proving they can be reliable in environments more demanding than yours.
The arbitrage is real. It's measurable. It has a 2.6× completion-rate signature in pilots we've run, and we expect to see similar signatures across other roles, other industries, other geographies as more companies start running structured pre-screening with proper inbound channels. Whether the resulting hires are demonstrably better is the question the next two years of running this at scale will answer.
The companies that move first capture a talent pool the rest of the market can't see. The companies that wait will eventually have to do the same thing, after their competitors have already eaten their lunch.
The hours recruiters don't work are the hours your best candidates are free.
That's the arbitrage. That's the edge. That's what we built Virvell to industrialize.
The Vena Medical pilot — 123 applicants, 78 completed pre-screens, 1 hire, 21 days from job posting to background-check clearance — is documented as a full case study at virvell.ai/case-studies/vena-medical.
See how Inclusion Arbitrage works in practice
If you want to talk about whether the structural pattern in this piece applies to a role you're hiring for right now, grab 15 minutes.
Book 15 Minutes