03The full analysis
A data-forward outlook on how staffing firms and recruiting agencies are absorbing AI in 2026, why the gap between AI adopters and laggards is now the clearest divide in the industry, and where contingent demand is actually shifting.
The global staffing market sits near $650 billion in 2025, roughly flat after a soft 2024, but the aggregate number hides the only statistic that matters right now: top-performing staffing firms are four times more likely to use AI than their peers - Bullhorn. A market that has not grown in two years is producing winners and losers at a rate the industry has not seen in a generation, and the line separating them is not size, geography, or vertical. It is whether a firm has put AI into the work of sourcing, screening, and matching, or whether it is still selling the same recruiter hours at the same margins it sold in 2019.
This is the central tension of the 2026 staffing market, and it is why a flat top-line number is the least interesting fact about it. A staffing firm is, at its economic core, an arbitrage on human attention: it finds, vets, and places people faster and more reliably than the hiring company could do alone, and it captures a margin for that speed and reliability. When the cost of the underlying intelligence (the sourcing, the screening, the matching) collapses, the arbitrage changes shape. The firms that re-tool their operations around cheap intelligence widen their margins and take share. The firms that treat AI as a marketing checkbox keep the same cost structure into a market that no longer rewards it. The result is a bifurcation that the headline market size cannot show you.
This outlook is written for the people who have to act on that bifurcation: agency owners deciding where to invest a constrained technology budget, enterprise buyers of contingent labor deciding which suppliers will still be competitive in three years, and investors trying to separate durable repricing from momentum. We size the market honestly, segment by segment. We trace the Bullhorn consolidation play (Textkernel and TargetRecruit folded into an agentic stack) because it is the clearest single signal of where the category's center of gravity is moving. We map the talent marketplaces and contractor platforms (Upwork, Fiverr, Toptal, Deel, Mercor, Scale AI, and the rest) that are absorbing the demand traditional staffing once owned. And we close with a point of view a buyer or investor can actually use. The companion to this piece is our State of AI in Recruiting: 2026, which examines the recruiting workflow itself; here we examine the firms and platforms that sell recruiting as a service.
Contents
- How to read the 2026 staffing market
- Market size: a flat top line over a bifurcating base
- The segment view: where contingent demand is actually shifting
- The Bullhorn thesis: rolling up the agentic staffing stack
- AI adoption as the dividing line between growth and flat
- The contractor platforms: Upwork, Fiverr, and the marketplace squeeze
- Global employment infrastructure: Deel, Remote, and the EOR layer
- The vetted-talent networks: Toptal, Andela, Turing, A.Team, Braintrust
- The AI-native frontier: Mercor, Scale AI, and the data-labor economy
- On-demand and hourly: Instawork, Wonolo, and the gig-staffing layer
- The economics: what AI does to the staffing margin
- A buyer and investor playbook for 2026
1. How to read the 2026 staffing market
A staffing market outlook is only useful if it separates the cyclical from the structural, because in 2026 the two are easy to confuse and the confusion is expensive. The cyclical story is straightforward and largely behind us: hiring boomed through 2021, contracted hard through 2023 and 2024 as employers digested over-hiring and rate uncertainty, and is now stabilizing into a flat-to-modest-growth posture. The structural story is the one that determines who wins: AI is changing the unit economics of the staffing business itself, independent of the cycle. A firm that reads 2026 as purely cyclical will wait for the market to recover and discover that the market recovered for someone else.
The single most important framing for the year is that staffing is a margin business under input-cost compression. The inputs that a staffing firm buys (recruiter time spent sourcing, screening, scheduling, and coordinating) are exactly the inputs AI makes dramatically cheaper. This is not a peripheral efficiency story. It strikes the cost structure that defines the industry. When a recruiter who could carry 4 to 6 open requisitions can now carry 12 to 15 with AI doing the first-pass sourcing and screening, the firm that deploys that recruiter either takes more revenue per head or undercuts competitors on price while holding margin. Both outcomes take share from firms that have not made the change. The whole 2026 story flows from this one mechanism.
The second framing is that demand is not disappearing, it is relocating. The work that traditional staffing firms have historically placed (the contract software engineer, the temporary warehouse worker, the interim finance professional) is increasingly being intermediated by platforms that look nothing like a staffing agency: AI-native talent marketplaces, employer-of-record infrastructure, and on-demand labor apps. This does not mean staffing volume is shrinking in aggregate. It means the channel through which contingent demand is satisfied is fragmenting, and a staffing firm that defines its addressable market as "the way we have always placed people" is watching its real addressable market migrate to channels it does not operate in. Reading the market correctly means reading where the demand is going, not just where it has been.
The third and final framing is that the platform layer is consolidating while the labor supply is fragmenting, and these two trends are not contradictory. On the supply side, the universe of people available for contingent work has never been larger or more globally distributed, spread across freelance marketplaces, vetted networks, on-demand apps, and AI-labor platforms. On the software and infrastructure side, the tools that staffing firms use to operate (the CRM, the ATS, the sourcing and matching engine) are consolidating fast, with Bullhorn the clearest example of a platform absorbing point solutions to assemble an end-to-end agentic stack. A firm navigating 2026 must therefore make two distinct decisions: which fragmented labor pools to access, and which consolidating platform to build its operations on. Conflating these two decisions is one of the most common strategic errors we see.
2. Market size: a flat top line over a bifurcating base
The honest headline for 2026 is that the global staffing market is large, mature, and not growing, and that this is the least important thing about it. Staffing Industry Analysts, the authoritative source for market sizing, valued the global staffing market at roughly $620 billion in 2024 and expects it to remain essentially flat in 2025, held back by trade and tariff uncertainty and broader macroeconomic and geopolitical drag - Staffing Industry Analysts. Consensus industry estimates put the 2025 figure near $650 billion, a number that captures both the scale of the market and its stagnation. For context, the market peaked around $648 billion in 2022 after the post-pandemic hiring surge, then contracted through 2023 and 2024 before stabilizing - Staffing Industry Analysts.
The phrase "soft 2024" understates the depth of the contraction in some segments, which matters because the recovery is uneven. The decline was not a gentle plateau. It was a multi-year drawdown concentrated in the most cyclical segments, with double-digit revenue declines in 2023 and 2024 in parts of the market before the rate of decline narrowed to low single digits in 2025 - Staffing Industry Analysts. The firms that lived through this experienced it not as a soft patch but as the worst stretch since the financial crisis, and the experience hardened a generation of agency owners against the assumption that volume always returns. That hardening is part of why AI adoption is accelerating: a firm that cannot grow its way out of a downturn has to operate its way out, and operating its way out means automating the cost structure.
The US market, the world's largest single staffing market, illustrates the stabilization-without-recovery pattern precisely. The US staffing industry generated roughly $178.7 billion in 2025 and is forecast to grow only about 1 to 2% annually through 2027, a recovery so modest that it barely registers against inflation - Staffing Industry Analysts. SIA expects most major staffing markets to return to moderate growth in 2026 after a difficult 2025, but "moderate" is the operative word. No one is forecasting a return to the growth rates of 2021. The base case for the next several years is a large market growing slower than the broader economy, which makes share-taking the only meaningful path to growth for an individual firm, and AI the primary mechanism for taking it.
The strategic reading of these numbers is that the aggregate has become a distraction, and the variance beneath it is the real story. A flat market is, almost by definition, a market where one firm's gain is another firm's loss, because there is no rising tide lifting all boats. This is what makes the AI dividing line so consequential in 2026 specifically: in a growing market, a laggard can still grow on the tide and never notice it is losing relative position. In a flat market, the laggard's lost share shows up immediately as declining revenue, and the adopter's gained share shows up as growth that looks almost impossible against the flat backdrop. The market size tells you the pie is not growing. Everything interesting is happening in how the slices are being recut.
3. The segment view: where contingent demand is actually shifting
Aggregate staffing numbers conceal a set of divergent segment stories, and a buyer or investor who reasons from the aggregate will misjudge every one of them. The staffing market is not one market. It is a collection of segments (IT, healthcare, industrial, professional, clerical) with their own demand drivers, their own AI exposure, and their own trajectories. Understanding where contingent demand is shifting requires disaggregating the flat top line into its moving parts, because the parts are moving in opposite directions and the net result only looks flat because the moves partially cancel.
Healthcare staffing is the segment that best illustrates the normalization story, because it ran the hottest during the pandemic and therefore had the furthest to fall. The segment generated roughly $39.4 billion in 2025, a 6% decline from 2024, as the extraordinary travel-nurse demand of the pandemic years finished unwinding, and SIA projects only modest growth of around 2% for 2026 - Staffing Industry Analysts. Within healthcare, the demand is relocating toward the more durable subsegments: locum tenens revenue is set to grow around 5% in both 2026 and 2027, while travel nursing, the pandemic-era boom segment, is forecast to stay roughly flat. The lesson for a healthcare staffing firm is that the segment is not recovering uniformly. It is recomposing toward physician and specialized clinical placement and away from the high-volume nursing surge that defined 2021 and 2022.
IT staffing tells a starkly different story, and it is the segment where the AI question is most acute because IT staffing sells exactly the skills AI most directly affects. The segment has been weak, with revenue essentially flat and forecast to grow only about 1% in 2026 and 2027, and the list of largest IT staffing firms has actually shrunk as the segment consolidated - Staffing Industry Analysts. The structural pressure on IT staffing is twofold: enterprise IT hiring has been cautious, and the contract software roles that IT staffing firms place are precisely the roles that AI-native talent marketplaces and vetted developer networks are now competing for most aggressively. An IT staffing firm in 2026 is squeezed between weak end demand and intensifying competition from platforms that vet and match developers faster and at lower cost than a traditional agency can.
The cross-segment trend that matters most for everyone is the structural shift toward contingent labor as a permanent strategy rather than a stopgap. Roughly 76% of companies now use freelancers or contractors to build more adaptable teams, treating the contingent workforce not as overflow capacity but as a deliberate way to scale without the fixed cost of full-time hires during an uncertain period - StaffingHub. This is genuinely good news for the staffing industry in the abstract: the demand for flexible labor is structurally rising. The complication is that the channel through which that demand is satisfied is fragmenting away from traditional agencies and toward platforms. The contingent pie is growing even as the traditional-staffing slice of it stays flat, which is the clearest possible signal that demand is relocating rather than disappearing.
For a staffing firm, the practical implication of the segment view is that the safest strategic posture is to follow the demand into its new channels rather than defend the old ones. A healthcare firm should lean into locums and specialized clinical placement where demand is durable. An IT firm should confront the reality that it competes with AI-native developer platforms and either match their vetting speed with AI of its own or move upmarket into the relationship-heavy, hard-to-automate placements where a human firm still wins. And every firm should recognize that "contingent demand is rising" and "traditional staffing is flat" are both true at once, which means the growth is available but only to firms willing to operate in the channels where the demand has moved.
4. The Bullhorn thesis: rolling up the agentic staffing stack
If you want to understand where the staffing technology market is going, watch what Bullhorn is buying, because Bullhorn is the dominant operating system for the staffing industry and its acquisition strategy is the clearest single statement of the category's direction. Bullhorn's CRM and ATS sit at the center of how tens of thousands of staffing firms run their daily operations, which gives the company a unique vantage point and a unique incentive: it can see exactly which capabilities its customers most need to compete, and it has the distribution to deliver them at scale. Over the past two years, Bullhorn has used that position to assemble an agentic AI stack through acquisition, and the shape of that stack tells you what the firms running on Bullhorn will be able to do that their competitors cannot.
The first major piece was Textkernel, the Amsterdam-based leader in sourcing AI, which Bullhorn acquired to deepen its parsing and matching capabilities. Textkernel is not a minor tuck-in: it powers more than 2,000 customers globally, including eight of the top ten staffing agencies in the world, and its parsing-and-matching technology is among the most widely deployed in the industry - Staffing Industry Analysts. The strategic logic is precise: sourcing and matching are the most computationally tractable parts of the staffing workflow, the parts where AI delivers the clearest leverage, and owning the best-in-class engine for them lets Bullhorn put that leverage directly into its core platform rather than relying on third-party integrations. Bullhorn explicitly framed the deal as a move to accelerate its AI strategy - Bullhorn.
The second piece was TargetRecruit, a Houston-based provider of front- and middle-office staffing solutions built natively on Salesforce, acquired in 2025. The deal grew Bullhorn's Salesforce user base to nearly 150,000 users and meaningfully strengthened its position in healthcare staffing, particularly the locums and per diem segments that, as the segment analysis above shows, are exactly where durable healthcare demand is concentrating - Bullhorn. The TargetRecruit acquisition reveals a second layer of the thesis beyond AI: Bullhorn is not only buying intelligence capabilities, it is buying into the Salesforce ecosystem and into the specific high-growth verticals where staffing demand remains strong. A roll-up that combines a sourcing-AI engine with a Salesforce-native middle-office platform in healthcare is a roll-up designed to own the full operational stack for the segments that will still be growing.
The capstone is Bullhorn Amplify, launched at the company's Engage Boston conference on May 28, 2025, which is where the acquired pieces become an agentic product rather than a collection of features. Amplify is built to let firms scale impact and revenue without adding headcount, deploying a digital workforce across the core recruiting workflow with skills the company labels Source, Match, Screen, and Submit - Bullhorn. Powered by Salesforce Agentforce and Textkernel's search and match engine, it can surface and contact top candidates automatically based on AI-driven relevancy scores, conduct and score initial screening interviews around the clock, and assemble candidate submission packets including resume redesigns. The agentic candidate-matching agent Bullhorn and Textkernel ship leverages more than 90 actions to drive the recruiting workflow - Salesforce AppExchange. This is the point of the roll-up: the acquired parsing engine, the Salesforce-native operations layer, and the agentic orchestration combine into a system that does the staffing workflow rather than merely tracking it.
The validation that this thesis is correct comes from the customers adopting it at the high end of the market. The Adecco Group, one of the largest staffing companies in the world, expanded its Bullhorn partnership specifically to accelerate its digital transformation with agentic AI - Bullhorn. When the largest players in the industry commit to an agentic platform strategy, it tells you that the agentic stack is not a small-firm experiment or a vendor talking point. It is becoming the operating assumption of the industry's leaders. For a staffing firm choosing its technology foundation in 2026, the Bullhorn roll-up represents a clear bet that the future of the business is agentic, and the firms that build on that foundation will have a structural cost-and-speed advantage over firms still stitching together point solutions or running on legacy systems with AI bolted on as an afterthought.
5. AI adoption as the dividing line between growth and flat
The clearest and most actionable finding in the entire 2026 staffing landscape is that AI adoption now predicts firm-level performance better than any other variable, and the data behind this claim is unusually direct. Bullhorn's 2026 GRID Industry Trends Report, based on a survey of nearly 2,300 recruitment professionals conducted in late 2025, found that top-performing staffing firms are four times more likely to leverage AI - Bullhorn. This is not a soft correlation buried in a footnote. It is the headline of the most widely cited annual benchmark in the staffing industry, and it reframes AI from an efficiency nice-to-have into the single strongest observable marker separating the firms that are growing from the firms that are stuck flat.
The supporting figures from the same report make the mechanism concrete, because they show AI affecting the specific operational metrics that drive a staffing firm's economics. 78% of firms reporting more than 25% revenue growth use AI tools in their ATS, and 56% of the highest-growth firms report average placement times under 10 days - Bullhorn. On the workflow level, 55% of firms say AI screening improved their KPIs by more than 25%, and 46% say AI cut screening time in half or better. These are not abstract productivity claims. Placement speed and screening throughput are the operational levers that determine how many fills a recruiter produces per quarter, and how many fills per recruiter is the proximate driver of both revenue and margin. AI is improving exactly the metrics that translate most directly into financial performance, which is why the adoption-performance link is so tight.
The independent productivity data corroborates the Bullhorn findings from an entirely different methodology, which is what gives the dividing-line thesis its weight. The American Staffing Association's 2026 Staffing Productivity Report, built from roughly 1.6 million monthly data points, found that recruiter call time reached 286 minutes per week in the first quarter of 2026, the highest on record and double the level of early 2024, while recruiter interactions with candidates and clients jumped 60% year over year - American Staffing Association. Recruiters now use an average of 1.36 AI tools, up from a single tool in early 2024. The picture these numbers paint is of recruiters doing dramatically more relationship and conversation work, precisely because AI has taken over the sourcing and screening busywork that previously consumed their hours. The productivity is real, it is measured by an independent body, and it is concentrated among the firms that have adopted AI tools.
The crucial nuance, and the reason the dividing line is widening rather than holding steady, is that agentic AI penetration is still early, which means the gap is poised to grow before it closes. The Bullhorn report found that only about 10% of agencies have implemented agentic AI across their full workflow, with 45% citing data concerns as the biggest obstacle to adoption - HR Tech Feed. This is the most important strategic fact in the report. The firms that are four times more likely to use AI are mostly using it in narrow, assistive ways, not yet in full agentic workflows. The frontier (the full agentic stack that Bullhorn Amplify and similar platforms enable) is still largely unexploited. The leaders' current advantage comes from basic AI adoption. The next, larger advantage will come from agentic deployment, and only one firm in ten has reached it. The dividing line between growing and flat firms is going to widen substantially over the next two years as the leaders move from assistive AI to agentic AI while the laggards are still worrying about data readiness.
The reason data concerns dominate the list of obstacles deserves explicit treatment, because it points to where the real work of adoption lies. Agentic AI is only as good as the data it acts on, and a staffing firm's competitive asset is its candidate and client database. A firm with years of messy, duplicated, poorly structured records cannot simply switch on an agentic workflow and expect it to source and match effectively, because the agent will act on bad data and produce bad results. This is why data readiness is the binding constraint on agentic adoption rather than the technology itself, which is largely available off the shelf now. The firms that will win the next phase are the ones investing now in cleaning, structuring, and enriching their data so that when they deploy agentic workflows, the agents have a high-quality foundation to act on. The strategic implication is that the work of preparing for agentic AI starts well before the AI is deployed, and the firms that have not started that work are further behind than the headline adoption statistics suggest.
For a staffing firm, the message could not be clearer or more urgent: AI adoption is no longer optional, it is the difference between growth and stagnation in a flat market. The firms hesitating because of data concerns are making a defensible short-term decision (bad data does produce bad AI results) that compounds into an indefensible long-term position (competitors are pulling away while they wait). The correct response is not to wait for the market to recover or for the technology to mature, both of which are arguments for inaction. The correct response is to start the data work now, deploy assistive AI immediately where the data supports it, and build toward agentic workflows on a cleaned data foundation, because the firms that do this are the ones that will be four times more likely, then ten times more likely, to be the top performers as the gap widens.
6. The contractor platforms: Upwork, Fiverr, and the marketplace squeeze
The public freelance marketplaces are the most visible front in the relocation of contingent demand, and their 2026 financials reveal a category being reshaped by AI from two directions at once. Upwork and Fiverr are the two scaled public players, and both are simultaneously benefiting from AI as a new demand category and being squeezed by AI as a substitute for the lowest-value work on their platforms. Reading their results correctly means holding both effects in mind at once, because the net direction of the category depends on which effect dominates over time.
Upwork reported full-year 2025 gross services volume above $4.0 billion and revenue of $788 million, up 2.4% year over year, with a record adjusted EBITDA of $226 million at a 29% margin - The Motley Fool. The story inside those numbers is the bifurcation that defines the whole category: AI-related work is one of the company's top growth drivers and average GSV per active client rose to a record above $5,100, while categories like writing and translation face headwinds from AI displacement and the smallest transactional contracts (below $300) are eroding. Upwork is, in effect, watching AI hollow out the bottom of its marketplace (the commodity tasks a model can now do directly) while AI work and higher-value engagements grow at the top. The platform is getting more valuable per client even as the floor of low-value gigs gives way.
Fiverr shows the same pattern with a sharper pivot toward higher value. The company grew revenue 10% in 2025, accelerating from 8% in 2024, with services revenue (its higher-margin offering) growing 18% year over year in Q4 to represent a third of total revenue - SEC. The most telling metric is that high-value projects over $1,000 grew 23% and now represent roughly 15% of marketplace GMV, even as the total count of active buyers declined. Fiverr is deliberately trading the low-end, high-churn buyer for the high-value, AI-augmented project, and its Fiverr Go platform, which lets talent create and price AI versions of their services, is the clearest expression of a marketplace trying to capture rather than be displaced by AI. The strategic bet is that the platform can sell AI-augmented human work at higher prices than it ever sold pure human gig work, and the early services-revenue growth suggests the bet is paying off.
The first-principles read on the contractor marketplaces is that AI is a great filter that pushes value up the stack. The commodity end of freelance work (basic copywriting, simple translation, formulaic design) is exactly what AI does well, so that work is being absorbed by models directly rather than routed to a human freelancer through a marketplace. This destroys the low-value gig that platforms like Fiverr were built on. But the same dynamic increases the value of the work that genuinely requires human judgment, taste, domain expertise, and accountability, and it creates an entirely new category of demand (people who can build, prompt, fine-tune, and deploy AI) that the platforms are racing to capture. The marketplaces that win will be the ones that successfully move their supply and demand up the value stack faster than AI erodes the bottom. Upwork's record per-client spend and Fiverr's surging high-value project share suggest both are managing this transition, though neither is immune to the squeeze.
For a staffing firm or contingent-labor buyer, the contractor marketplaces are both a competitor and a signal. They are a competitor because they intermediate exactly the project-based contingent work a staffing agency might otherwise place, and they do it with lower overhead and global reach. They are a signal because their financials are a real-time readout on which kinds of contingent work AI is absorbing (the commodity bottom) and which it is making more valuable (the expert top). A buyer reading these results should infer that the contingent work worth routing through a high-touch staffing relationship is the high-judgment, high-accountability work, while the commodity work is migrating to self-serve marketplaces or being automated outright. The marketplaces are showing the whole industry where the human premium will and will not survive.
7. Global employment infrastructure: Deel, Remote, and the EOR layer
A category that did not exist at scale a decade ago has become one of the most consequential layers in the contingent-work economy: the employer-of-record (EOR) and global payroll infrastructure that lets companies hire workers anywhere without establishing a local entity. This layer matters to the staffing outlook because it is the plumbing that makes the global, distributed, contingent workforce operationally possible, and the company that owns the plumbing captures a strategic position adjacent to (and increasingly overlapping with) traditional staffing. The growth and valuations in this layer are among the strongest in the entire talent-tech market, which is itself a signal of where capital believes the future of work is heading.
Deel is the runaway leader and one of the most valuable private companies in the category. In October 2025, Deel closed a $300 million Series E at a $17.3 billion valuation, led by Ribbit Capital with Coatue and Andreessen Horowitz participating - Calcalist. The financial profile behind the valuation is exceptional for the category: Deel surpassed $1 billion in annual recurring revenue in 2025, recorded its first $100 million revenue month, and reached its third straight year of profitability, while serving more than 35,000 businesses and 1.5 million workers across 150+ countries and processing $22 billion in payroll annually - Deel. Deel's unified system handles the full lifecycle (hiring, onboarding, payroll, compliance, benefits, and offboarding) across full-time employees, contractors, and EOR arrangements, which positions it not as a staffing firm but as the infrastructure on top of which distributed hiring happens.
Remote is the most direct competitor in the EOR layer, valued around $3 billion in its most recent funding, with revenue that reached roughly $600 million in 2023 and a workforce that has scaled substantially since - CB Insights. Oyster HR reached a $1 billion valuation as a global employment platform serving 180+ countries, and Velocity Global rounds out the set of scaled EOR providers focused on contingent and distributed workforce management - Contrary Research. The category is consolidating around a handful of well-capitalized players, with Deel pulling decisively ahead on revenue and valuation, which suggests the EOR layer is following the same winner-take-most dynamic that characterizes most infrastructure markets: the platform with the broadest country coverage, the deepest compliance capability, and the most workers under management gets harder to displace with every additional worker it onboards.
The strategic significance of the EOR layer for the staffing outlook is that it competes with traditional staffing for the same fundamental job while operating on a completely different model. A company that needs to engage a contractor in another country can either go through a staffing agency that handles the relationship, or it can use an EOR platform to employ the worker directly on a compliant basis. The EOR model captures the compliance and payroll value that staffing firms historically bundled into their markup, and it does so at software margins rather than services margins. As distributed and contingent work becomes the default rather than the exception, the EOR layer captures an increasing share of the value that used to flow to staffing intermediaries, particularly for the longer-term, direct-relationship engagements where the hiring company wants to manage the worker itself and just needs the legal and payroll infrastructure.
For a staffing firm, the EOR layer is best understood as both a threat and a potential partner, and the right posture depends on the firm's model. For firms whose value was primarily compliance and payroll handling of distributed contractors, the EOR platforms are a direct disintermediation threat, because they deliver that value at lower cost. For firms whose value is genuine talent discovery, vetting, and relationship management, the EOR layer can be a partner that handles the infrastructure while the firm focuses on the parts that are hard to automate. The investor read is cleaner: the EOR layer is one of the highest-quality growth stories in talent tech, with Deel's profitable billion-dollar ARR standing in sharp contrast to the flat traditional staffing market, and the valuation gap between the two reflects a market that believes infrastructure for distributed work will keep capturing value as the world of work keeps distributing.
8. The vetted-talent networks: Toptal, Andela, Turing, A.Team, Braintrust
Between the open marketplaces and the AI-native platforms sits a category built on a different premise: that the value is not in volume or in raw matching but in curation. The vetted-talent networks pre-screen their supply heavily, admitting only a small fraction of applicants, and sell access to that curated pool at a premium. This model directly competes with high-end staffing for professional and technical placements, and its 2026 trajectory shows a category being reshaped by AI both as a tool the networks use and as a force redefining what "vetting" means.
Toptal is the established leader in premium freelance talent, with more than $200 million in annual revenue and over 40% year-over-year growth, positioning itself as one of the largest fully distributed workforces in the world across developers, designers, finance experts, and product managers - CompWorth. Toptal's model is the purest expression of the curation premise: it famously admits a tiny percentage of applicants and charges accordingly, betting that buyers will pay more for a network where vetting is the product. The durability of Toptal's growth in a flat market is evidence that the curation premium is real, but it also makes Toptal a prime target for AI-native challengers who argue they can vet more rigorously and more cheaply with AI than Toptal can with its human-led screening.
Turing, Andela, A.Team, and Braintrust each occupy distinct positions in the vetted-network category, and their funding tells a story of a category still attracting capital despite the flat broader market. Turing raised a $111 million Series E in March 2025 at a $2.2 billion post-money valuation, repositioning from remote-developer matching toward AI specialists and AI training work - Sacra. Andela has raised $378 million total and uses AI to match companies with software engineers from emerging markets, expanding its capabilities through its acquisition of technical-assessment company Woven in January 2026 - PitchBook. Braintrust, the blockchain-based network where talent retains its earnings and the platform charges no intermediary markup to the freelancer, raised an $80 million Series B in February 2026, bringing total funding to $242.5 million across 7 rounds, with over 600,000 community members and 1,000+ companies - Crunchbase. A.Team rounds out the set as a network for assembling cross-functional product teams of senior freelancers, targeting the higher-complexity end where a single contractor is insufficient and a coordinated team is the unit of work.
The first-principles tension running through this entire category is that AI is simultaneously the networks' best tool and their biggest threat to differentiation. Vetting is, at its core, an assessment problem, and AI is rapidly getting better at assessment: it can evaluate code, conduct technical interviews, score work samples, and rank candidates at a scale and cost no human-led vetting operation can match. This is why Turing pivoted toward AI work and why Andela bought a technical-assessment company. The networks that built their premium on human curation now face challengers (covered in the next section) who argue that AI vetting is not just cheaper but better, because it is more consistent, more scalable, and free of the human bottleneck that limits how many candidates a network can rigorously assess. The vetted networks' response has been to adopt AI vetting themselves while leaning on the relationship, community, and team-assembly value that AI does not directly replicate.
For a buyer of high-end contingent talent, the vetted networks remain a strong option precisely because curation is genuinely valuable for senior and specialized roles where a bad hire is expensive and the buyer lacks the internal capacity to vet rigorously. The right network depends on the job: Toptal and A.Team for premium individual experts and assembled teams, Turing and Andela for global engineering and AI talent at scale, Braintrust for buyers who want to avoid the intermediary markup. For an investor, the category is interesting but contested: the curation premium is real and the networks are growing in a flat market, but the AI-native challengers are attacking the core differentiator (vetting) with a fundamentally cheaper and more scalable approach, which puts long-term pricing power in question. The networks that survive will be the ones that move their value beyond vetting (into community, team assembly, and ongoing relationship management) before AI commoditizes the screening that built them.
9. The AI-native frontier: Mercor, Scale AI, and the data-labor economy
The most consequential development in the entire contingent-work landscape is the emergence of platforms that do not merely use AI to staff humans but exist to staff humans for AI, and in doing so have built the fastest-growing talent businesses the market has ever seen. This is the AI-native frontier, and it is where the deepest disruption lives, because these platforms found a hiring problem the traditional staffing stack served poorly (vetting and deploying enormous numbers of specialized human contributors to train and evaluate AI models) and solved it with an AI-native approach that traditional firms cannot easily replicate.
Mercor is the defining company, and its trajectory is without precedent in talent tech. The company built an AI system that interviews candidates through a 20-minute domain-specific AI video interview, assesses their skills, and matches them to work, and on the strength of that model it raised a $350 million Series C in October 2025 at a $10 billion valuation, a fivefold increase in eight months, on the back of revenue that reached roughly $760 million annualized at the end of 2025 - TechCrunch. That figure then accelerated further, crossing $1 billion run rate by early 2026 and reaching about $2 billion in gross annualized revenue by June 2026 (materially lower net of contractor payouts), as Mercor entered talks to double its valuation to $20 billion - TechCrunch. Mercor operates a network of more than 30,000 experts (scientists, doctors, lawyers, and software engineers) who earn over $85 per hour on average, with the company paying out more than $1.5 million per day to contractors - CNBC. These are not the numbers of a recruiting tool. They are the numbers of a company that found a category of hiring the old model could not serve and built a new model to serve it.
Scale AI demonstrates the same data-labor economy at even larger scale, and its 2025 events reveal how strategically valuable the human-labor layer underneath AI has become. In June 2025, Meta agreed to take a 49% non-voting stake in Scale AI for $14.3 billion, valuing the company at $29 billion, with Scale's founder taking a senior role inside Meta as part of the deal - Wikipedia. Scale's business runs on a vast contingent workforce: it coordinates over 100,000 contractors performing data labeling and annotation through platforms including Outlier, which focuses on the LLM data work that requires contributors with advanced degrees and specialized expertise - Sacra. Scale reached roughly $870 million in revenue in 2024 with a run rate heading well past that into 2025. The Meta deal is the clearest possible signal that the human-labor layer feeding AI development is considered strategically essential, worth billions to lock up.
The first-principles insight that makes this category the most important signal in the outlook is that when intelligence becomes cheap, the binding constraint shifts to the specialized human judgment that trains and evaluates the intelligence. The AI labs do not need help writing code or generating text; the models do that. What they need, at enormous volume, is specialized humans (the doctors, lawyers, scientists, and expert engineers) who can produce the high-quality training data and the rigorous evaluations that make the models better. This is a hiring problem characterized by massive scale, deep specialization, and a need for fast, rigorous assessment, which is precisely the profile where an AI-native vetting-and-matching platform outperforms a traditional staffing firm by orders of magnitude. Mercor and Scale did not build better staffing agencies. They built a new kind of talent platform for a category of demand that did not meaningfully exist five years ago and is now one of the fastest-growing in the economy.
For the staffing industry, the AI-native frontier is less an immediate competitive threat to most firms' core business than a strategic preview of where the entire model is heading. Most staffing firms will not be vetting AI trainers in 2026. But the pattern (finding a hiring problem the traditional stack serves poorly and solving it with AI-native assessment and matching) is exactly the pattern that will spread from AI-data labor into adjacent specialized-contingent categories. The firms that study Mercor's AI-interview-and-match model are studying the future of their own vetting function. For an investor, the valuation gap is the loudest statement in the market: Mercor at $10 billion and Scale at $29 billion dwarf the valuations of mature traditional staffing companies with far larger revenue, because the market is pricing AI-native talent platforms as categorically more valuable than the firms that merely help recruiters do their jobs. Whether that gap is justified or inflated is the open question we examine in our Talent Acquisition Tech Market Map: 2026, but the capital has already cast its vote, and it voted for the AI-native model.
10. On-demand and hourly: Instawork, Wonolo, and the gig-staffing layer
At the opposite end of the skill spectrum from the AI-data-labor frontier sits the on-demand hourly staffing layer, and it deserves attention because it represents the technological transformation of the highest-volume, most operationally intensive segment of traditional staffing: filling shifts. The fundamental job here is to connect businesses with hourly workers (in hospitality, warehousing, logistics, manufacturing, and retail) and fill open shifts within hours rather than days. This is the segment where traditional light-industrial and hospitality staffing agencies have operated for decades, and it is being reshaped by platforms that turn shift-filling into a marketplace transaction rather than a phone-and-spreadsheet operation.
Instawork is the scaled leader in the category, operating a network of over 9 million background-checked and skills-verified local workers who fill shifts, temp-to-hire, and full-time roles for businesses across hospitality, warehouse and logistics, manufacturing, and retail - Contrary Research. The platform's value proposition is speed and reliability at scale: a business can post a shift and have a vetted worker confirmed within hours, with the platform handling the matching, the background checks, and the payment flow that a traditional agency would manage manually. Wonolo occupies an adjacent position, an on-demand staffing marketplace founded in 2014 that focuses on blue-collar sectors like warehousing, event staffing, and delivery, with total funding of roughly $190.9 million including a $140 million Series D - CB Insights. Both platforms are doing to light-industrial and hospitality staffing what Upwork did to freelance project work: turning a relationship-and-phone-based agency function into a self-serve, technology-mediated marketplace.
The strategic dynamic in the gig-staffing layer is that the value of the agency in high-volume hourly staffing was always operational efficiency, and operational efficiency is exactly what software does better than people. A traditional light-industrial staffing branch employs recruiters who maintain relationships with a pool of hourly workers, field shift requests from clients, and manually match workers to shifts by phone and text. This is labor-intensive coordination work with thin margins, and it is precisely the kind of coordination a marketplace platform can automate. Instawork and Wonolo capture the matching, vetting, and payment functions in software, deliver faster fills at lower overhead, and scale across markets without opening physical branches. The on-demand platforms are not adding AI as a feature to a staffing agency; they are replacing the staffing agency's operational model with a marketplace, and AI now sharpens the matching and demand-forecasting that make the marketplace efficient.
For a traditional hourly staffing firm, the on-demand layer is the most direct disintermediation threat in the entire outlook, because it attacks the high-volume, low-margin core of the light-industrial and hospitality staffing business with a structurally lower-cost model. The defensible response is to compete on the things a marketplace handles poorly: complex compliance, dedicated on-site management for large accounts, and the relationship depth that enterprise clients with specialized requirements still value. For a buyer of hourly labor, the on-demand platforms offer compelling speed and cost for standard shift-filling, while a traditional agency may still win for accounts that need hands-on management, specialized worker pools, or compliance handling that exceeds what a self-serve platform provides. The gig-staffing layer is the clearest example in the market of contingent demand relocating from agencies to platforms, and it is happening fastest in the highest-volume segment because that is where the operational efficiency of software delivers the largest advantage over human coordination.
11. The economics: what AI does to the staffing margin
Everything in this outlook ultimately comes back to a single economic question that determines who wins: what does cheap intelligence do to the margin structure of a staffing business? Answering it from first principles is more useful than any individual data point, because the margin mechanics explain why AI adoption predicts performance, why the valuation gaps look the way they do, and why demand is relocating. A staffing firm earns its margin by performing a sequence of functions (sourcing, screening, matching, coordinating, and managing the relationship) more efficiently than the hiring company could, and capturing a markup for that efficiency. AI changes the cost of each function differently, and understanding the differential is the key to the whole picture.
The functions AI compresses most are sourcing and screening, which is why those are exactly where the documented productivity gains concentrate. Sourcing is a search-and-rank problem over large candidate populations, and AI does search-and-rank better and faster than any human. Screening is a first-pass assessment problem, and AI does first-pass assessment at a scale no recruiter can match. When these two functions, which historically consumed the majority of a recruiter's hours, collapse in cost, the firm faces a strategic choice about what to do with the freed capacity. It can keep its headcount and dramatically increase output per recruiter (taking more revenue per head), or it can hold output and reduce headcount (cutting cost), or it can pass some of the savings to clients as lower prices (taking share). The Bullhorn and ASA data show the leading firms choosing the first option: recruiters carrying more requisitions, conducting more interactions, and producing faster placements, which expands revenue without proportionally expanding cost. That is margin expansion, and it is available only to the firms that have made the AI investment.
The functions AI compresses least are relationship management and complex judgment, which is where the durable human value concentrates and where the staffing margin of the future will be earned. Persuading a passive candidate to consider a move, navigating a delicate counter-offer, understanding the unstated political dynamics of a hiring manager's team, and standing behind a placement with accountability are functions that require trust, context, and human judgment that AI does not replicate. This is why the productivity data shows recruiter interaction time rising even as AI takes over the busywork: the human is being redeployed from the functions AI does well to the functions AI does poorly. The firms that understand this redeploy their recruiters toward relationship and judgment work and let AI handle the rest. The firms that do not understand it either resist AI (and keep their recruiters trapped in the busywork AI should be doing) or naively try to automate the relationship work (and lose the trust that justified their margin in the first place).
The structural conclusion is that AI bifurcates the staffing market into a high-volume, low-touch tier that platforms increasingly own, and a high-judgment, high-touch tier where human firms retain durable margin, and the danger is being caught in the middle. The high-volume tier (commodity freelance gigs, standard hourly shifts, basic contract roles) is migrating to marketplaces and AI-native platforms because that is where software's efficiency advantage is decisive and the human premium is smallest. The high-judgment tier (executive search, specialized technical placement, complex team assembly, sensitive contract negotiation) is where human firms armed with AI tooling will keep winning, because the human value is real and AI amplifies rather than replaces it. The firms in genuine trouble are the ones stuck in the middle: too commoditized to defend a human premium, too slow to adopt AI to compete on the platform model's efficiency. The strategic imperative is to pick a side of the bifurcation deliberately and commit to it, rather than being slowly squeezed out of the middle by platforms below and AI-augmented specialists above.
The investor translation of these mechanics explains the valuation gaps that recur throughout this outlook. The market assigns software multiples to the platforms (EOR infrastructure, AI-native talent platforms, on-demand marketplaces) because they capture staffing value at software margins and scale without proportional headcount. It assigns services multiples to traditional staffing firms because their model still ties revenue to human capacity, even AI-augmented human capacity. The most valuable companies in the contingent-work economy (Deel at $17.3 billion, Mercor at $10 billion, Scale at $29 billion) are the ones that turned a staffing function into software-margin infrastructure. The traditional firms that will close the valuation gap are the ones that use AI to push their own economics toward software-like operating leverage (more output per head, faster placement, expanding margin) rather than remaining linear services businesses. The firms that do not will keep trading at services multiples in a flat market, which is to say they will not be where the value accrues.
12. A buyer and investor playbook for 2026
Everything in this outlook converges on decisions that buyers of contingent labor, owners of staffing firms, and investors actually control, and the purpose of a research-grade outlook is to make those decisions sharper. The macro environment (flat market, uneven segment recovery, accelerating AI) is not controllable. What is controllable is where to source contingent talent, which technology foundation to build a firm on, how to position against the bifurcation, and where to deploy capital. This closing playbook turns the analysis into action.
For the buyer of contingent labor, the first principle is to route the work to the channel that matches its value tier, because the cost and quality differences across channels are now large. High-volume, standardized work (hourly shifts, commodity tasks, basic contract roles) belongs on the platforms (Instawork and Wonolo for hourly, Upwork and Fiverr for project work, EOR providers like Deel for compliant direct engagement) where the software model delivers speed and cost that no agency can match. High-judgment, high-stakes work (specialized engineering, senior professionals, complex teams) belongs with the vetted networks (Toptal, A.Team, Turing, Andela) or with AI-augmented staffing firms that bring genuine relationship and vetting depth. The error to avoid is paying agency markups for commodity work that a platform handles better, or routing high-stakes placements through a self-serve marketplace that cannot provide the judgment and accountability the work requires.
For the owner of a staffing firm, the playbook is the most urgent, because the AI dividing line is widening and the window to get on the right side of it is closing. The first move is to start the data-readiness work immediately, since data concerns are the binding constraint on agentic adoption and the work takes time. The second is to deploy assistive AI now wherever the data supports it (sourcing and screening, where the documented productivity gains concentrate) and build toward agentic workflows on a platform designed for them, of which the Bullhorn stack is the clearest example given its roll-up of Textkernel, TargetRecruit, and Amplify into an agentic system. The third is to pick a side of the bifurcation deliberately: either compete on platform-like efficiency in high-volume staffing, or move decisively into the high-judgment tier where human value is durable, but do not get caught in the squeezed middle. A firm that does these three things turns a flat market into a share-taking opportunity. A firm that waits becomes the share that the adopters take.
For the investor, the playbook is to distinguish durable repricing from momentum, and the bifurcation framework is the tool for doing it. The software-margin platforms (EOR infrastructure, AI-native talent platforms, on-demand marketplaces) are repricing staffing value into software economics, and the strongest of them (Deel's profitable billion-dollar ARR being the cleanest example) represent durable value creation, not just momentum. The AI-native frontier (Mercor, Scale) is genuinely differentiated where it has traction (AI-data labor) but its valuations price in a generalization into broader hiring that has not yet been proven, so the disciplined position is to watch which way the evidence breaks before assuming the model generalizes. The traditional staffing firms are mostly not the place value accrues unless they demonstrably push their economics toward software-like operating leverage through AI, which most have not yet done. The capital is flowing to the platforms and the AI-native models for a reason rooted in margin mechanics, and the investor who understands those mechanics can tell the difference between a platform that has structurally repriced a staffing function and a company that has simply attached AI to its marketing.
The single point of view that ties this entire outlook together is that 2026 is the year the staffing industry's flat top line stops being a description of the industry and starts being a smokescreen over a violent recomposition beneath it. The market is not growing, but underneath that stillness, AI is recutting the margins, the platforms are absorbing the high-volume demand, the AI-native models are creating entirely new demand, and the gap between the firms that adopted AI and the firms that did not is widening into the defining divide of the category. A buyer, owner, or investor who reads the flat number and concludes the industry is sleepy will be blindsided. The one who reads beneath it (who follows the demand into its new channels, builds on the consolidating agentic platform, picks a side of the bifurcation, and deploys capital where the margin mechanics point) will be positioned exactly where the value is moving. The companion analyses to this outlook, our State of AI in Recruiting: 2026 and our Talent Acquisition Tech Market Map: 2026, examine the workflow and vendor layers of the same transformation; here the message is simply that the staffing business itself is being repriced, and the repricing rewards the deliberate and punishes the passive.
This outlook reflects the staffing and agency technology landscape as of May 2026. Market sizes, valuations, funding rounds, and adoption figures change rapidly in a fast-moving market; verify current figures before any procurement or investment decision. For the workflow-level view of how AI is changing recruiting, read our State of AI in Recruiting: 2026, and for the vendor-by-vendor map of the recruiting technology stack, read our Talent Acquisition Tech Market Map: 2026.