The State of AI in Recruiting: 2026

A research-grade map of where artificial intelligence actually sits inside the hiring funnel in 2026, what it has changed, and what it has not.

AI adoption in human resources nearly doubled in a single year, climbing from 26% of organizations in 2024 to 43% in 2025 - SHRM. That is one of the fastest technology adoption curves any operational function has ever recorded, faster than the spread of applicant tracking systems in the 2000s and faster than the cloud migration of HR systems in the 2010s. Recruiting sits at the leading edge of it. Among HR professionals specifically, 69% now use AI to support recruiting, up from 51% the year before, making talent acquisition the single most AI-saturated activity in the entire people function.

But here is the uncomfortable part of the same data set: average cost-per-hire and time-to-hire have both increased over the past three years, a period that correlates almost exactly with the arrival of generative AI in the recruiter's toolkit - SHRM Benchmarking. Adoption is racing ahead of measurable outcomes. The tools are everywhere. The results are uneven. And the gap between the two is the most important story in recruiting technology this year.

This report does not treat AI as a magic input or an existential threat. It treats recruiting as a system that converts a defined set of inputs (open roles, candidate supply, recruiter time, hiring-manager attention) into a defined output (a hired person who stays and performs). When one of those inputs, intelligence, becomes radically cheaper, the system reorganizes around the new cost structure. The interesting questions are structural: which steps in the funnel get absorbed by cheap intelligence, which steps resist it, where the value migrates, and what talent leaders should build versus buy versus ignore.

We will move from the macro labor backdrop, through adoption data by workflow stage, into what is genuinely working versus what is marketing, then to the vendor landscape shifts that AI has triggered, the candidate-side AI arms race, and finally a decision framework for talent leaders. The companion to this piece is our Talent Acquisition Technology Market Map: 2026, which maps the full vendor ecosystem category by category. Read them together.

Contents

1. The labor market backdrop: why AI arrived now
2. The adoption picture: how fast and how deep
3. AI across the recruiting workflow, stage by stage
4. What is actually working versus what is hype
5. The new economics: when intelligence becomes a commodity input
6. Vendor landscape shifts: agents, consolidation, and the rebundling of recruiting
7. The candidate-side arms race: AI versus AI
8. Measurement: why time-to-hire went up while everyone bought AI
9. Risk, regulation, and the governance gap
10. A decision framework for talent leaders in 2026
11. The 2027 outlook

1. The labor market backdrop: why AI arrived now

Technology adoption is never purely about the technology. It is about the conditions that make a buyer willing to change behavior. To understand why recruiting embraced AI so quickly between 2024 and 2026, you have to start with the labor market itself, because the macro environment created exactly the kind of pressure that makes operational leaders try new tools.

The headline condition is a market that has gone quiet without going soft. As of the March 2026 reading, US job openings held at 6.9 million while the hires rate ticked up to 3.5% - BLS JOLTS. The quits rate sat at 2.0%, meaning fewer people are voluntarily leaving, and layoffs remained low at a 1.2% rate. Economists describe this as a low-hire, low-fire equilibrium: employers are neither expanding aggressively nor cutting deeply. For recruiters, that produces a peculiar pressure. Requisition volume is lower than the 2021 to 2022 frenzy, but each open role carries more scrutiny, more internal competition for budget, and a higher bar for the quality of the eventual hire.

When volume falls but the cost of a bad hire rises, the rational response is to invest in efficiency and precision rather than raw throughput. That is precisely the environment where AI tooling becomes attractive. A recruiting organization that cannot justify headcount expansion can still justify software that makes its existing recruiters more productive. This is why budget intent diverged so sharply from hiring intent: two-thirds of talent acquisition leaders plan to increase technology spending in 2026, with more than half specifically allocating budget toward new recruiting platforms - HR Executive. The tools are getting funded even as the req load stays flat.

The second condition is the maturation of the underlying models. The generative AI that recruiters reach for in 2026 is not the same technology that existed when ChatGPT first went viral. The frontier has moved through several generations. OpenAI's current flagship is GPT-5.5, released in the spring of 2026 with a one-million-token context window and materially fewer hallucinations than its predecessors - OpenAI. Anthropic's strongest model is Claude Opus 4.8, which the company positions around reliability and is the strongest computer-use and browser-agent model it has tested - Fortune. The practical consequence is that the same recruiting task that produced embarrassing output in early 2024 produces usable output in 2026. Capability crossed the threshold of usefulness for real work, and adoption followed capability.

The third condition is competitive fear. 84% of talent leaders worldwide say they will use AI in 2026, and 93% of recruiters plan to increase their AI use - LinkedIn via HR Dive. Numbers that high are not driven by careful return-on-investment analysis. They are driven by the perception that everyone else is doing it and that falling behind is the larger risk. Roughly a third of HR professionals believe their own organization trails its competitors on AI - SHRM. That perception, accurate or not, is itself a powerful adoption engine. When the labor market goes quiet, the models get good, and the fear of being left behind spreads, you get the curve we are now living through.

{
  "title": "US Labor Market Snapshot, March 2026",
  "subtitle": "JOLTS rates showing the low-hire, low-fire equilibrium recruiters operate in",
  "type": "bar",
  "xKey": "metric",
  "yKeys": [
    {
      "key": "rate",
      "label": "Rate (%)"
    }
  ],
  "data": [
    {
      "metric": "Hires rate",
      "rate": 3.5
    },
    {
      "metric": "Quits rate",
      "rate": 2
    },
    {
      "metric": "Layoffs rate",
      "rate": 1.2
    },
    {
      "metric": "Openings rate",
      "rate": 4.4
    }
  ],
  "yAxisFormatter": "percent",
  "source": "US Bureau of Labor Statistics, JOLTS March 2026",
  "sourceUrl": "https://www.bls.gov/news.release/jolts.nr0.htm"
}

2. The adoption picture: how fast and how deep

It is tempting to read a single adoption number and declare the transformation complete. The data resists that reading. AI in recruiting is simultaneously the fastest-spreading operational technology in recent memory and one of the shallowest in actual depth of use. Understanding both halves of that statement is the key to interpreting everything else in this report.

Start with the speed. The jump from 26% organizational adoption in 2024 to 43% in 2025 represents a near-doubling in twelve months - SHRM. When you widen the lens to any use of AI anywhere in the recruiting process, the figure climbs higher still: 69% of companies report using AI in some capacity for hiring - SelectSoftware Reviews. Those are extraordinary penetration rates for a function that, only a few years ago, ran largely on manual resume review and phone screens. The direction of travel is unambiguous, and the slope is steep.

Now the depth, which tells a different story. Despite the high penetration, only 18% of companies say they are using AI broadly across hiring processes, and only 37% of talent acquisition professionals actively integrate generative AI into their daily workflows - SelectSoftware Reviews. In SHRM's deeper cut, recruiting is the highest-adoption HR practice area at 27% of organizations, yet that still leaves nearly three-quarters of organizations not running AI across their core recruiting motion - SHRM. The picture that emerges is broad but thin: a great many teams have touched AI for a task or two, far fewer have rebuilt their process around it.

The depth gap widens further when you slice by company size. Extra-large organizations with 5,000 or more employees adopt AI in HR at 60%, while small organizations sit at 33% - SHRM. This is the opposite of what early commentators predicted. The conventional wisdom held that nimble small companies would out-adopt lumbering enterprises. In practice, enterprises have the requisition volume, the budget, and the dedicated talent-operations teams to justify and implement AI tooling, while small teams often lack the volume to make the tooling pay off. AI in recruiting is, for now, an enterprise-led phenomenon.

The most forward-looking signal in the adoption data is the appetite for autonomy. 52% of talent leaders plan to add autonomous AI agents to their teams in 2026 - Korn Ferry. This is a qualitatively different ambition than using AI to draft a job description. An agent does not assist a recruiter with a task; it owns the task end to end, takes actions, and reports back. The fact that a majority of talent leaders intend to deploy agents within the year tells you the center of gravity is shifting from AI-as-feature to AI-as-coworker. Whether the agents deliver is a separate question, addressed later in this report, but the intent is now mainstream.

{
  "title": "AI Adoption in HR: 2024 vs 2025",
  "subtitle": "Organizational adoption nearly doubled year over year",
  "type": "grouped-bar",
  "xKey": "year",
  "yKeys": [
    {
      "key": "anyHR",
      "label": "Any AI in HR"
    },
    {
      "key": "recruitingUse",
      "label": "Use AI for recruiting"
    }
  ],
  "data": [
    {
      "year": "2024",
      "anyHR": 26,
      "recruitingUse": 51
    },
    {
      "year": "2025",
      "anyHR": 43,
      "recruitingUse": 69
    }
  ],
  "yAxisFormatter": "percent",
  "source": "SHRM 2025 Talent Trends",
  "sourceUrl": "https://www.shrm.org/topics-tools/research/2025-talent-trends/ai-in-hr"
}

The interpretation that matters for a talent leader reading this is that you are neither early nor late. If your team has adopted AI for a handful of tasks, you are squarely in the mainstream, not ahead of it. The competitive frontier has moved from whether you use AI to how deeply and how well you have integrated it into the actual decision points of your hiring process. Surface adoption is now table stakes. Depth is where the advantage lives, and depth is exactly what most organizations have not yet achieved.

3. AI across the recruiting workflow, stage by stage

To understand AI in recruiting with any precision, you have to abandon the idea of AI as a single thing applied to a single process. Recruiting is a sequence of distinct stages, each with its own data, its own decisions, and its own tolerance for automation. AI has penetrated each stage to a very different degree, and the pattern of penetration is not random. It tracks a clear logic: the more repetitive and rules-based the stage, the deeper the automation; the more it depends on judgment, relationships, and accountability, the more the human remains in control.

The cleanest way to see this is to walk the funnel from top to bottom. At the top sits sourcing, the act of finding candidates who might fit. In the middle sit screening and assessment, the act of filtering the many down to the few. Then comes interviewing and the scheduling logistics around it. At the bottom sit selection, offer, and the handoff to onboarding. AI behaves differently at each layer, and mapping that behavior is the foundation for every buying and process decision a talent leader makes.

graph TD
    A["Sourcing & Outreach<br/><i>high automation, agent-ready</i>"] --> B["Screening & Matching<br/><i>high automation, accuracy-gated</i>"]
    B --> C["Assessment & Skills Testing<br/><i>moderate automation, validity-gated</i>"]
    C --> D["Interview Scheduling<br/><i>near-full automation, logistics</i>"]
    D --> E["Interviewing & Intelligence<br/><i>assist mode, notetaking & analysis</i>"]
    E --> F["Selection & Offer<br/><i>human-led, AI advises</i>"]
    F --> G["Onboarding Handoff<br/><i>workflow automation</i>"]

3.1 Sourcing and outreach

Sourcing is where AI has gone deepest, because sourcing is fundamentally a search-and-pattern-matching problem, and search-and-pattern-matching is exactly what modern models excel at. The task is to scan an enormous population of profiles, identify the small subset that matches a role, and reach out to them at scale with relevant, personalized messages. None of that requires the kind of contextual judgment that resists automation. It requires breadth, speed, and the ability to draft passable copy, all of which are now cheap.

The numbers confirm the depth of penetration. 59% of recruiters are currently using AI to surface hidden-gem candidates, those who would not surface through a conventional keyword search but match on deeper signal - LinkedIn via Pin. The platforms enabling this operate at a scale that would have been unimaginable a few years ago. Sourcing engines now index between 800 million and over a billion public profiles aggregated from dozens of sources, then apply matching models on top. AI-native sourcing engines aggregate upward of a billion profiles and rank them against a role in seconds, a category we profile in depth in our market map.

The most consequential development in this layer is the move from suggestion to action. LinkedIn's own Hiring Assistant, its first AI agent, reached general availability in late 2025, and the early results from charter customers are striking: 62% fewer profiles reviewed, more than four hours saved per role, and 69% higher InMail acceptance rates - LinkedIn via Pin. Read those numbers carefully. Fewer profiles reviewed is not a sign of less work done; it is a sign of better filtering upstream, so that recruiter attention lands only on candidates worth the attention. The higher acceptance rate suggests the outreach is genuinely more relevant, not merely more voluminous. This is AI improving the quality of a stage, not just its speed.

The practical takeaway for a talent leader is that sourcing automation has crossed from experimental to expected. If your sourcers are still building boolean strings by hand and copy-pasting outreach templates, you are operating below the current efficiency frontier. The risk in this layer is not that automation fails to work; it is that automation works so well it floods candidates with near-identical AI-drafted messages, eroding the very response rates it was meant to improve. The teams winning here are the ones using AI to find better-fit candidates and reserving genuine personalization for the messages that matter, not the ones blasting the largest possible volume.

3.2 Screening and matching

Screening is the stage where the promise of AI is largest and the peril is sharpest. The task is to reduce a large applicant pool to a short list, which means making consequential decisions about real people based on incomplete information. Done well, AI screening removes drudgery and surfaces candidates that tired human reviewers would miss. Done badly, it encodes bias at scale and rejects qualified people for reasons no one can explain. The stakes are why this layer attracts the most regulatory attention and the most internal anxiety.

The efficiency case is real and well documented. Talent acquisition professionals using generative AI report a 20% reduction in overall workload, the equivalent of reclaiming one full workday per week - SHRM. A large share of that reclaimed time comes from automating first-pass resume review, the most tedious and least differentiated task in the recruiter's week. When 89% of HR professionals using AI in recruiting report that it saves time or increases efficiency, screening is where much of that time saving originates - SHRM.

The quality case is more contested and more interesting. The strongest evidence does not support replacing human screening with AI screening; it supports combining them. Organizations that combine AI screening with human-led final interviews cut time-to-hire by 40% while improving first-year retention by 25% - SelectSoftware Reviews. The retention improvement is the number to dwell on. Faster hiring that produces worse hires is not a win; it is a way to fail more efficiently. The fact that the combined model improves retention suggests AI screening, used as a filter rather than a decider, genuinely raises the quality of who reaches the human stages.

That distinction, filter versus decider, is the entire ballgame in this layer. The organizations getting value treat AI screening as a way to rank and surface, with a human making the actual cut. The evidence that this is the prevailing approach is encouraging: recruiter judgment overrides AI recommendations in 58% of organizations - SelectSoftware Reviews. That is not a sign of distrust in the tools; it is a sign of healthy process design, where AI informs and humans decide. The danger lies with the organizations in the minority who let the model auto-reject, because that is where unexplainable, unaccountable, and potentially unlawful decisions get made at scale.

There is a deeper reason the filter-versus-decider line matters, and it is worth making explicit because it drives both the legal and the quality dimensions of screening. A model that ranks candidates is making a recommendation that a human can inspect, question, and overrule. A model that rejects candidates is making a decision, and decisions in hiring are subject to a body of employment law that predates AI by half a century and does not care whether the discriminating agent is a person or an algorithm. When a screening model auto-rejects, the organization has effectively delegated a legally consequential decision to a system that, in most deployments, cannot produce a defensible, individualized rationale for why a specific person was rejected. That is not merely a compliance risk; it is a quality risk, because the same opacity that makes a rejection hard to defend also makes it hard to learn from. A filter that surfaces the wrong candidates gets corrected by the human who reviews them. A decider that rejects the wrong candidates produces no feedback signal at all, because no one ever sees the people it wrongly screened out.

The matching technology underneath screening has also matured in a way that changes what is possible and what is dangerous. Early resume-screening tools matched on keywords, which produced the infamous failure mode of rejecting strong candidates who happened to describe their experience in different words than the job description used. Modern systems match on semantic meaning and on inferred skills rather than literal terms, which genuinely reduces that particular failure and is the engine behind the hidden-gem candidate surfacing that 59% of recruiters now rely on. But semantic matching introduces a subtler risk: because the model is inferring fit from patterns in historical data, it can reproduce whatever biases were latent in who got hired before. The keyword era rejected people for using the wrong words. The semantic era can reject people for not resembling the historical winners. Neither failure is acceptable, and the second is harder to detect precisely because it feels more sophisticated. This is why adverse-impact testing, examined in the governance section, is not optional for any organization running modern screening models.

3.3 Interviewing, scheduling, and interview intelligence

The interview stage splits cleanly into two very different problems, and AI has treated them differently. The first problem is logistics: coordinating calendars, sending reminders, rescheduling, and confirming. The second problem is the interview itself: the conversation, the evaluation, and the record of what was said. Logistics has been almost fully automated. The conversation has been augmented, not replaced, and the augmentation is reshaping how interviews get evaluated.

On the logistics side, conversational AI has effectively solved interview scheduling for high-volume hiring. Paradox, the conversational recruiting platform whose assistant is named Olivia, works for more than 500 companies and reports cutting time-to-hire in half for high-volume employers, with customer results including Chipotle hiring 75% faster and one large retailer saving tens of thousands of hours weekly - Paradox. The significance of Workday acquiring Paradox for approximately one billion dollars in 2025 is that the largest enterprise HR vendor decided conversational scheduling and screening was core enough to own outright - Global Legal Chronicle. When a function gets acquired into the system of record, it has stopped being a novelty.

The conversation side is where the more subtle transformation is happening, through a category called interview intelligence. These tools join interviews, transcribe them, structure the content, and surface insights. The recruiter is still asking the questions and forming the judgment, but the record is now complete, searchable, and analyzable. 66% of recruiters intend to increase their use of AI for pre-screening interviews in 2026 - LinkedIn via HR Dive, and the interview-intelligence vendors have grown into serious businesses on the back of that demand. Metaview raised a 30.1 million euro Series B led by Google Ventures and counts roughly 2,500 customers including Sony, Brex, and Deel - EU-Startups.

The strongest validation of the interview-intelligence category came when Zoom acquired BrightHire in November 2025 - CB Insights. Zoom owns the video infrastructure where interviews increasingly happen; acquiring an interview-intelligence layer means the conversation platform itself now wants to capture and analyze the hiring conversation. The deeper implication is that the interview is becoming a structured data asset rather than an ephemeral conversation. Once every interview is transcribed and analyzed, organizations can finally audit interviewer quality, detect inconsistency, and measure which questions actually predict success. That is a profound shift in a stage that, for decades, produced almost no usable data at all.

3.4 The data layer: skills, matching, and talent intelligence

Underneath every stage of the funnel sits a layer that AI has quietly made into the most strategically important part of the modern recruiting stack: the data layer. Sourcing, screening, and matching all depend on a representation of who candidates are and what they can do, and AI has transformed that representation from a pile of unstructured resumes into a structured graph of skills, experiences, and trajectories. This is the layer that talent-intelligence platforms occupy, and it is where the largest proprietary data assets in recruiting now live. The scale is genuinely hard to fathom: the leading platforms maintain graphs spanning more than a billion candidate profiles, hundreds of millions of job records, and skills ontologies running into the tens of millions of distinct skills - Knowlee.

The reason this layer matters more than any single feature above it is that the data is the moat. A screening model is only as good as its representation of the candidate, and a matching engine is only as good as its understanding of how skills relate to one another and to outcomes. The platforms that have spent years assembling these graphs, and the models that have learned the relationships within them, possess an asset that a new entrant cannot easily replicate, because the asset is accumulated rather than coded. This is why the talent-intelligence category commands the attention it does in enterprise buying, and why the suite vendors have been so acquisitive in it. Owning the data layer means owning the foundation that every AI capability above it stands on.

The most consequential application of the data layer is the shift toward skills-based hiring, which AI has moved from aspiration to operational reality. For decades, organizations screened on proxies, degrees, prior titles, and pedigree, because evaluating actual skill was too expensive to do at scale. When intelligence becomes cheap, evaluating skill at scale becomes feasible, and the proxies lose their justification. 75% of recruiters now say skills-based hiring will be their top priority, and companies with the most skills-based searches are 12% more likely to make a quality hire - Recruit AI Suite. The practical effect is that the data layer is not just making existing recruiting faster; it is enabling a different basis for hiring decisions altogether, one grounded in demonstrated capability rather than credential. The talent-intelligence platforms that own this layer are profiled category by category in our market map.

4. What is actually working versus what is hype

After two years of breathless claims, the recruiting industry now has enough operating history to separate the genuine value from the marketing. The honest answer is that some applications of AI in recruiting are delivering measurable, repeatable results, while others remain aspirational, oversold, or actively counterproductive. A research-grade assessment requires naming both, because the cost of believing the hype is wasted budget and damaged candidate experience, while the cost of dismissing the real gains is falling behind competitors who captured them.

Begin with what is working, because the evidence here is solid. The clearest win is recruiter productivity on repetitive tasks. The one-workday-per-week reclaimed by generative AI users is not a vendor claim; it shows up consistently across independent surveys - SHRM. The second clear win is candidate experience in high-volume hiring, where conversational AI answers questions instantly, schedules without friction, and keeps applicants warm. The third, and most strategically important, is the combined screening model that improves both speed and retention when AI filters and humans decide. These three are real, and a talent leader who has not captured them is leaving value on the table.

{
  "title": "Measured Outcomes from AI in Recruiting",
  "subtitle": "Documented results from surveys and vendor case studies, 2025-2026",
  "type": "bar",
  "xKey": "outcome",
  "yKeys": [
    {
      "key": "value",
      "label": "Improvement (%)"
    }
  ],
  "data": [
    {
      "outcome": "Workload reduction (GenAI users)",
      "value": 20
    },
    {
      "outcome": "Time-to-hire cut (AI + human screen)",
      "value": 40
    },
    {
      "outcome": "First-year retention lift (combined)",
      "value": 25
    },
    {
      "outcome": "Fewer profiles reviewed (LinkedIn HA)",
      "value": 62
    },
    {
      "outcome": "Higher InMail acceptance (LinkedIn HA)",
      "value": 69
    }
  ],
  "yAxisFormatter": "percent",
  "source": "SHRM, SelectSoftware Reviews, LinkedIn",
  "sourceUrl": "https://www.selectsoftwarereviews.com/blog/ai-recruiting-statistics"
}

Now the hype, which is equally important to name. The most overstated claim is fully autonomous hiring, the notion that an agent can own a requisition end to end and produce a hire without meaningful human involvement. The aspiration is real, the appetite is real (52% of talent leaders want agents), but the delivered reality in 2026 is augmentation, not autonomy. The same research that shows enthusiasm for agents also shows that only 22% of respondents believe their leaders can effectively manage mixed human-AI teams - Korn Ferry. Organizations want autonomous agents but admit they are not ready to manage them. That gap between desire and readiness is where most agent deployments will underdeliver in the near term.

The second overstated claim is AI-driven candidate assessment that predicts job performance from video, voice, or game-based signals. The category has real vendors and real customers, but the predictive validity claims frequently outrun the peer-reviewed evidence, and the regulatory environment is tightening around exactly these tools. A talent leader should treat assessment claims with the same skepticism they would apply to any vendor selling a proprietary predictive model: ask for the validation study, ask about adverse-impact testing, and ask what happens when a rejected candidate demands an explanation.

The third area to approach carefully is the gap between perception and measurement. 98% of hiring managers say AI has improved the hiring process, yet the same period saw cost-per-hire and time-to-hire rise - SelectSoftware Reviews. When perception and hard metrics diverge that sharply, the responsible interpretation is that the perceived improvement is partly real (tasks feel easier) and partly a halo effect (new tools feel like progress). The discipline of measuring actual funnel outcomes, not satisfaction surveys, is what separates the teams genuinely benefiting from the teams who merely feel modern.

A useful way to sort the working from the hyped is to ask where the value claim is bounded versus where it is open-ended. The bounded claims tend to hold. Conversational AI that schedules interviews in high-volume hiring makes a narrow, verifiable promise: it coordinates calendars without human involvement, and it either does or does not. It does, which is why the category was worth a billion dollars to Workday. Interview intelligence makes a similarly bounded promise: it captures and structures the interview record, and it demonstrably does. The open-ended claims are the ones to scrutinize, because they promise outcomes that depend on factors far outside the tool. A vendor claiming its assessment predicts job performance is making an open-ended claim that depends on the validity of its model, the representativeness of its training data, and the stability of the relationship between whatever it measures and actual performance, none of which the buyer can easily verify. The reliable heuristic is that bounded efficiency claims about specific tasks tend to be true, while open-ended outcome claims about prediction and quality tend to be oversold.

The high-volume hiring case deserves particular emphasis because it is the clearest example of AI delivering genuine, compounding value rather than marginal convenience. In frontline and hourly hiring, the constraints are speed and candidate drop-off: the best applicants take other jobs while a slow process deliberates, and every hour of friction loses candidates. Conversational AI attacks exactly this constraint, engaging applicants instantly, screening them through natural conversation, and scheduling them before they disengage. Paradox reports cutting time-to-hire in half for high-volume employers, with results such as a major fast-casual chain hiring 75% faster - Paradox. This is not a halo effect; it is a structural fit between what the tool does and what the hiring context needs. The lesson generalizes: AI delivers the most reliable value where its capability maps cleanly onto a specific, measurable constraint, and the most dubious value where it promises to improve something diffuse and hard to measure.

5. The new economics: when intelligence becomes a commodity input

To reason clearly about where AI takes recruiting next, it helps to strip the question down to its economic core. Recruiting is, at bottom, the conversion of inputs into a hire. The inputs are candidate supply, recruiter labor, hiring-manager time, and information about who is good. For the entire history of the profession, the binding constraint was information: finding candidates was hard, evaluating them was harder, and the recruiter's value came from doing that scarce, expensive cognitive work. AI changes the binding constraint by making the cognitive work cheap. When a previously scarce input becomes a commodity, the economics of the whole system reorganize.

The first-order effect is the one everyone notices: tasks that consisted purely of cheap-now cognition get automated, and the people who did only those tasks lose their economic moat. Building boolean searches, screening resumes against a checklist, and scheduling interviews were all valuable precisely because they were tedious and time-consuming for humans. The moment a model does them in seconds, that value collapses. This is why the productivity gains concentrate in exactly these activities, and why the recruiters most exposed are those whose role was primarily mechanical.

But the first-order effect is not the interesting one. The interesting question is where the value goes when it leaves the commoditized tasks, and the answer is that it migrates to the parts of recruiting that cheap intelligence cannot supply. Three things resist commoditization. The first is accountability: someone has to own the decision, defend it to a hiring manager, and stand behind it when it goes wrong, and a model cannot bear that responsibility. The second is relationship: a hesitant senior candidate is closed by a human who understands their unspoken concerns, not by an optimized email sequence. The third is judgment under ambiguity: deciding whether an unconventional background is a risk or an asset is exactly the kind of contextual call that resists pattern-matching. Value flows toward these three, away from the mechanical middle.

This is the same structural pattern that played out in every prior technology shift, and it is worth testing the analogy rather than asserting it. When spreadsheets made calculation cheap, the accounting profession did not disappear; it moved up the value chain from computation to advice and judgment. When design software made layout cheap, designers moved from production to strategy and taste. The recurring dynamic is that automating the mechanical layer of a knowledge profession does not eliminate the profession; it raises the floor of what a practitioner must contribute to be worth employing. The recruiter who was valuable for doing the mechanical work must become valuable for the judgment, relationships, and accountability that sit above it.

The data is consistent with this reframing rather than with mass displacement. SHRM finds organizations are 5.7 times more likely to shift job responsibilities than to displace jobs as a result of AI, with only 7% reporting any job displacement while 39% report shifts in responsibilities - SHRM. The recruiting role is being rewritten, not erased. This also explains the most counterintuitive finding in the entire talent-trends literature: that the top recruiting priority for 2026 is not AI skills at all. 73% of talent acquisition leaders rank critical thinking and problem-solving as their top priority, while AI skills rank only fifth - Korn Ferry. Leaders intuitively understand that when the machine handles the mechanical layer, the human contribution that remains is precisely judgment. That intuition is economically correct.

There is a second economic mechanism worth naming, because it explains why the volume of hiring activity can rise even as the cost per activity falls. When the marginal cost of a task collapses, the rational quantity of that task increases. If sourcing a qualified candidate cost an hour of recruiter time, an organization rationed sourcing carefully and pursued only the most promising leads. If sourcing a qualified candidate costs pennies of model inference, the organization can afford to pursue far more leads, engage passive candidates it would previously have ignored, and maintain relationships with talent pools it could never have justified before. This is the same dynamic that played out when cloud computing made compute cheap: usage did not fall, it exploded, because cheap compute made previously uneconomical workloads economical. Cheap intelligence will similarly expand the surface area of recruiting activity, which is part of why the function is not contracting even as individual tasks automate.

A rigorous analysis has to pressure-test its own conclusion, so consider the strongest counter-argument: that this time really is different, and that AI will not merely automate the mechanical layer but will eventually absorb the judgment layer too, leaving little for the human recruiter to do. The Mercor and Micro1 cases, examined later, are the leading evidence for this view, because they demonstrate AI systems making genuine assessment and matching decisions, not merely assisting them. The counter-argument deserves to be taken seriously rather than dismissed. But the evidence in 2026 does not support the strong version of it for general recruiting. The contexts where AI-native models are making real hiring decisions are precisely the contexts where the decision is well-bounded and the volume is enormous, such as vetting thousands of specialized contractors against a clear specification. The contexts that resist automation are those where the decision is ambiguous, the relationship is load-bearing, and accountability cannot be delegated. The frontier will keep moving, but the structural reason judgment resists commoditization, that someone must own and defend the decision, does not dissolve simply because the models improve.

The practical implication for how a talent leader should think about their own team is significant. The temptation is to ask which recruiters to cut now that AI does part of their job. The first-principles answer is that the question is backward. The right question is which recruiters can be elevated, by handing the mechanical work to AI, into the higher-value work of judgment, relationship, and accountability that the business will pay more for, not less. The organizations that frame AI as a way to do the same work with fewer people will capture the smaller, first-order prize. The organizations that frame it as a way to redeploy human attention toward what only humans can do will capture the larger, second-order one.

6. Vendor landscape shifts: agents, consolidation, and the rebundling of recruiting

AI has not only changed how recruiters work; it has restructured the industry that sells them tools. The defining dynamic of 2025 and 2026 is consolidation, and the defining cause of that consolidation is AI. When intelligence becomes the differentiating layer, the large platforms that own the system of record have a powerful incentive to absorb the best AI capabilities rather than let nimble startups build a wedge into their accounts. The result is a wave of acquisitions that is rebundling functions the prior decade had unbundled. We map this in full in our Talent Acquisition Technology Market Map: 2026, but the AI-specific logic deserves attention here.

The headline deals all share a structure: a suite vendor buys an AI-native point solution to fold its intelligence into the platform. Workday acquired Paradox for approximately one billion dollars to own conversational candidate experience - Global Legal Chronicle. SAP acquired SmartRecruiters to fold AI-powered talent acquisition into SuccessFactors - SAP. Zoom acquired BrightHire to bring interview intelligence into the video layer - CB Insights. Even among the standalone players, the same logic operates at smaller scale: Ashby acquired Talent Llama to add AI interviewing, and Phenom acquired Included AI for agentic people analytics and Be Applied for skills-first assessment - Knowlee. The pattern is unmistakable. AI capability is being pulled into platforms.

{
  "title": "Major Talent-Tech Acquisitions, 2025",
  "subtitle": "Suite vendors absorbing AI-native point solutions (disclosed and estimated values)",
  "type": "bar",
  "xKey": "deal",
  "yKeys": [
    {
      "key": "value",
      "label": "Deal Value ($M, disclosed/est.)"
    }
  ],
  "data": [
    {
      "deal": "Workday / Paradox",
      "value": 1000
    },
    {
      "deal": "Zoom / BrightHire",
      "value": 36
    },
    {
      "deal": "Metaview Series B (raise)",
      "value": 33
    }
  ],
  "yAxisLabel": "USD millions",
  "source": "Global Legal Chronicle, CB Insights, EU-Startups",
  "sourceUrl": "https://globallegalchronicle.com/workday-acquires-paradox-in-1-billion-deal/"
}

The consolidation is not happening in a vacuum; it is occurring against a funding backdrop that is simultaneously cooler and more concentrated. Total HR technology investment was roughly 3.7 billion dollars in 2025, a 40% decrease year over year by one measure - Crunchbase News. Yet within that cooler total, talent acquisition led all categories with about one billion dollars raised across 197 deals, and the M&A engine ran hot, with roughly 90 to 100 companies acquired every quarter and 110 acquisitions in the third quarter of 2025 alone - Venero Capital. The capital that is flowing is flowing toward AI-differentiated recruiting, and the weaker players are being acquired or absorbed rather than funded independently.

The most economically interesting development is not the consolidation of incumbents but the emergence of AI-native businesses that did not exist before the technology made them possible. The clearest example is Mercor, founded by Thiel Fellows, which built an AI interviewer that assesses candidates and matches them to roles. Mercor raised a 350 million dollar Series C in October 2025 at a 10 billion dollar valuation, a fivefold jump in eight months, on the back of more than 450 million dollars in annual recurring revenue - TechCrunch. Micro1 built a similar AI recruiter named Zara and raised at a 500 million dollar valuation - TechCrunch. These are not recruiting tools sold to recruiters; they are AI-native recruiting businesses that compete with the recruiting function itself.

The first-principles reading of these new entrants is the most important strategic point in this section. When intelligence becomes cheap, the threat to incumbent recruiting tools is not only that competitors add AI faster. It is that entirely new business models become viable, models where the AI does not assist a recruiter but replaces the need for that part of the recruiter's job at the company that would have hired them. Mercor's customers are largely AI labs that need to vet enormous numbers of specialized human contributors; the company found a hiring problem that was too large and too specialized for traditional recruiting and solved it with an AI-native approach. The lesson for buyers is to watch not only how incumbents add AI, but where AI-native challengers find hiring problems that the old model served poorly. A new lineage of autonomous sourcing agents, which work large candidate indexes and conduct outreach with little ongoing human direction, sits in exactly this AI-native category, alongside the agent layers the incumbents are now bolting on.

7. The candidate-side arms race: AI versus AI

Every analysis of AI in recruiting that looks only at what employers are doing is missing half the system. Candidates have access to the same technology, and they are using it aggressively. The result is an arms race in which AI on the employer side increasingly confronts AI on the candidate side, and the most important emerging risk in recruiting is not that the tools fail to work but that they can be gamed, faked, and overwhelmed by adversaries using the same tools.

The benign version of candidate-side AI is the use of generative tools to write resumes, tailor applications, and prepare for interviews. This is now ubiquitous and largely legitimate; a candidate using AI to articulate genuine experience more clearly is doing nothing different in kind from using a spelling checker. The volume effect, however, is real and corrosive. When every applicant can generate a polished, keyword-optimized application in seconds, application volume explodes and the signal-to-noise ratio of any individual application collapses. This is part of why AI screening became necessary on the employer side: it is partly a defense against AI-inflated application volume on the candidate side. The two halves of the arms race feed each other.

The malign version is far more serious, and the data on it is alarming. Hiring fraud has industrialized. Gartner projects that by 2028, one in four candidate profiles worldwide will be fake - The Hire Hub. In a 2026 survey, 59% of hiring managers reported suspecting candidates of using AI to misrepresent themselves, and 91% of US hiring managers have encountered or suspected AI-generated interview answers during online meetings - The Hire Hub. The barrier to entry has collapsed: someone with no image-manipulation experience can reportedly build a fake candidate capable of passing a video interview in roughly seventy minutes.

{
  "title": "The Hiring Fraud Problem in 2026",
  "subtitle": "Survey and projection data on AI-enabled candidate fraud",
  "type": "bar",
  "xKey": "metric",
  "yKeys": [
    {
      "key": "value",
      "label": "Share (%)"
    }
  ],
  "data": [
    {
      "metric": "Mgrs suspecting AI misrepresentation",
      "value": 59
    },
    {
      "metric": "Mgrs encountering AI interview answers",
      "value": 91
    },
    {
      "metric": "Firms that hired a fraudulent candidate",
      "value": 41
    },
    {
      "metric": "Fake candidate profiles by 2028 (proj.)",
      "value": 25
    }
  ],
  "yAxisFormatter": "percent",
  "source": "Gartner and 2026 hiring-manager surveys via The Hire Hub",
  "sourceUrl": "https://www.thehirehub.ai/blog/deepfake-candidates-how-to-protect-your-hiring-process"
}

The structural consequence of industrialized fraud is that identity verification and trust become first-class problems in recruiting, not afterthoughts. 41% of companies have already hired a fraudulent candidate without knowing it - The Hire Hub, and the threat has attracted organized criminal infrastructure, including synthetic identity kits sold on dark-web markets and state-sponsored actors using fake hires to infiltrate companies. This elevates the background-verification category from a compliance checkbox to a strategic line of defense, a shift we examine in the market map. It also means the interview-intelligence tools that record and analyze interviews acquire a second purpose: they become part of the evidentiary record that a candidate is who they claim to be.

It is worth understanding why remote hiring specifically amplified this risk, because the answer points to where the vulnerability lives and how to close it. In an in-person process, identity verification happened implicitly: a candidate who showed up to an office was, at minimum, a real person whose appearance matched their later self. Remote hiring removed that implicit verification without replacing it with anything, creating a gap that AI-enabled fraud rushed to fill. The video interview, which felt like a verification step because you could see a face, turned out to be the weakest link, because a face on a screen can now be synthesized in real time. The organizations most exposed are those that hire fully remotely for roles with access to sensitive systems or data, which is precisely the profile that state-sponsored infiltration targets. The closing move is not to abandon remote hiring but to insert deliberate verification where the implicit version used to be: liveness checks, document verification, and, for sensitive roles, a verified in-person or escalated identity step before access is granted.

The economics of the arms race also explain why it will not resolve into a stable equilibrium any time soon. Both offense and defense draw on the same falling cost of intelligence, so every improvement in detection is met by an improvement in evasion, and vice versa. This is the familiar dynamic of every security domain, from email spam to payment fraud, and recruiting has now joined that category whether it wanted to or not. The practical consequence for a talent leader is a mindset shift: hiring is no longer only a matching problem, it is partly an adversarial-trust problem, and adversarial-trust problems are never solved, only managed. Budgeting for verification, building it into the process by design rather than bolting it on after an incident, and accepting that some friction is the price of trust are the marks of an organization that understands the new reality rather than one still operating as if the candidate on the other end is automatically who they say they are.

The first-principles lesson here is that any time a screening system becomes valuable, adversaries are incentivized to defeat it, and cheap intelligence lowers the cost of defeating it just as it lowers the cost of building it. A talent leader cannot treat AI adoption as a one-sided efficiency play. Every automated stage is also an attack surface. The teams that will navigate this well are the ones that pair automation with verification, that keep enough human touchpoints in the process to catch what the machines miss, and that treat the candidate-side arms race as a permanent condition rather than a temporary nuisance. The naive read of AI in recruiting is that it makes hiring easy. The sophisticated read is that it makes hiring a contest between increasingly capable systems on both sides of the table.

8. Measurement: why time-to-hire went up while everyone bought AI

The single most instructive data point in this entire report is the one that looks like a contradiction. Adoption of AI in recruiting exploded, satisfaction with AI is nearly universal, and yet average cost-per-hire and time-to-hire both increased over the three-year period that AI adoption accelerated - SHRM Benchmarking. If the tools work, why did the headline metrics move the wrong way? Resolving this apparent paradox is essential, because the wrong conclusion (the tools do not work) is as dangerous as the naive one (the tools are magic).

The resolution begins with recognizing that time-to-hire and cost-per-hire are system-level outcomes influenced by far more than tool efficiency. The same low-hire, low-fire labor market that drove AI adoption also lengthened hiring cycles for a reason that has nothing to do with software: when requisition volume is scarce and the cost of a bad hire is high, organizations deliberately slow down. They add interview rounds, require more sign-offs, and hold out for a better candidate. The increase in time-to-hire is partly a feature of risk-averse hiring behavior in a cautious market, not a failure of the AI that was deployed inside that slower process. The tool got faster while the surrounding process got more deliberate, and the process effect dominated the metric.

The second part of the resolution is the explosion in application volume driven by candidate-side AI, discussed above. When AI lets every candidate apply to far more roles with far less effort, the number of applications per requisition surges. Even a recruiting team using AI to screen faster can find its total cycle time lengthening, because the pool it must process has grown faster than its processing speed improved. The per-application efficiency went up; the per-requisition volume went up faster. This is a textbook case of a local improvement being swamped by a system-level change, and it is exactly why measuring tool-level efficiency (time saved per task) tells you almost nothing about funnel-level outcomes (time-to-hire).

The third and most actionable part of the resolution is that most organizations are not measuring the right things, or measuring at all. 56% of organizations do not formally measure the success of their AI investments, and only 16% use any custom return-on-investment metric - SHRM. When a majority of buyers cannot say whether their AI spend is paying off, the high satisfaction numbers reveal themselves for what they partly are: a feeling of progress untethered from evidence. The teams that will actually capture value from AI are the minority who instrument their funnel, isolate the effect of the tooling from the effect of the market, and track outcomes that matter, especially quality of hire and retention, rather than vanity metrics about tasks automated.

This points directly to the metric that talent leaders should obsess over in 2026: quality of hire. It is the outcome that the combined AI-plus-human model demonstrably improves (the 25% retention lift), it is the metric that resists gaming by application volume, and it is the one a majority of practitioners now believe AI can help measure. 61% of talent acquisition professionals believe AI will improve how they measure quality of hire - Recruit AI Suite, and LinkedIn's framework of weighting demand, retention, and mobility offers a credible starting structure. A leader who shifts the scorecard from speed and cost toward quality of hire will both measure AI's real contribution more accurately and steer their team toward the outcome that actually matters to the business.

9. Risk, regulation, and the governance gap

The fastest way to turn an AI advantage into an AI liability is to deploy the tools faster than the governance to control them. The recruiting function is uniquely exposed here, because hiring decisions are legally consequential, demographically sensitive, and increasingly regulated. The data reveals a governance gap that is wide, persistent, and, for many organizations, genuinely dangerous. Closing it is not a compliance nicety; it is a precondition for capturing AI's value without incurring legal and reputational damage that dwarfs the efficiency gains.

The first dimension of the gap is policy maturity. Roughly half of organizations, 49%, have an AI use policy in place, but only 25% describe their policies as clear and future-proof, while 54% report their policies are too restrictive or too tool-specific to be useful - SHRM. A policy that is either absent or badly designed offers little protection. The pattern suggests most organizations wrote policy reactively, in response to a specific tool, rather than building a durable framework for evaluating any AI used in hiring decisions. That reactive posture leaves them exposed every time a new tool enters the stack.

The second and more alarming dimension is regulatory awareness. Among HR professionals working in jurisdictions that already regulate AI in hiring, 57% are unaware that such regulations exist, and among those who are aware, only 12% have implemented compliant policies - SHRM. Read those two numbers together and the implication is stark: the overwhelming majority of organizations operating under AI hiring regulations are not compliant, and most do not even know they are subject to the rules. As enforcement matures, this is a liability waiting to crystallize, particularly around automated screening and assessment tools that make or heavily influence rejection decisions.

The third dimension is organizational ownership, and it contains a structural problem. HR rarely leads AI rollouts; IT, legal, and compliance typically drive them, and 52% of organizations do not involve HR directly in AI strategy - SHRM. This is a recipe for tools that are technically deployed but operationally and ethically ungoverned, because the function with the deepest understanding of hiring fairness and candidate experience is excluded from the decisions about which AI to deploy and how. The governance gap is, at root, an ownership gap.

The practical response is not to slow adoption to a crawl; it is to build governance that moves at the speed of adoption. That means three things in prose rather than as a checklist: first, a durable evaluation framework that any new hiring AI must pass, covering validity, adverse-impact testing, explainability, and a documented human-decision point; second, genuine HR ownership of that framework, with talent leaders at the table when tools are selected rather than informed after the fact; and third, a candidate-facing commitment to transparency about where and how AI is used in the process. Organizations that treat governance as an enabler of confident adoption, rather than as a brake on it, will deploy more AI, not less, because they will be able to defend what they deploy. The ones that treat governance as an afterthought will eventually deploy less, after a regulatory or reputational event forces a retreat.

10. A decision framework for talent leaders in 2026

Everything in this report points toward a small set of decisions that a talent leader actually controls. The macro market, the model capabilities, the vendor consolidation, and the fraud arms race are conditions to navigate, not levers to pull. The levers are these: which stages of your funnel to automate and how deeply, how to redeploy the human attention that automation frees, what to measure, and how to govern the whole thing. This section turns the analysis into a framework for pulling those levers well.

The first principle is to automate by stage, not by slogan. The funnel analysis in section three is the map: automate sourcing and scheduling aggressively because the value of human involvement in those stages is low and the efficiency gains are high; automate screening as a filter that ranks and surfaces while keeping a human as the decider; and keep selection, offer, and the closing of senior candidates firmly human, with AI in an advisory role. The mistake to avoid is uniform automation, applying the same level of AI to every stage regardless of how much judgment each stage requires. The teams that win match the depth of automation to the judgment-intensity of the stage.

The second principle is to redeploy rather than reduce. The economic analysis in section five showed that AI's value migrates toward judgment, relationships, and accountability. The leaders who frame AI as a headcount-reduction tool capture the small prize and demoralize the team that remains. The leaders who frame it as a way to lift recruiters out of mechanical work and into the high-value work the business pays more for will capture the larger prize and build a more capable, more retained team. This is also why the top-priority skill for 2026 is critical thinking, not AI proficiency: the skill that becomes scarce and valuable is exactly the one the machine cannot supply. Hire and develop for judgment, not for prompt-writing.

The third principle is to measure quality of hire above all else. Section eight established why: it is the outcome AI demonstrably improves in the combined model, it resists the volume gaming that distorts time-to-hire, and a majority of practitioners now believe AI can help measure it. A leader who reorients the scorecard around quality of hire, using a structure like demand, retention, and mobility, will both evaluate their AI investments honestly and steer the organization toward the outcome that matters. The corollary is to stop celebrating tasks automated and time saved as if they were outcomes; they are inputs, and inputs that improve while outcomes stagnate are a warning, not a victory.

The fourth principle is to govern at the speed of adoption, with HR owning the framework. Section nine showed the governance gap is wide and the ownership gap is its root cause. The leader's job is to insist that talent acquisition sits at the table when hiring AI is selected, to maintain a durable evaluation standard every tool must pass, and to be transparent with candidates about AI's role. Done well, governance is not a brake; it is what lets you adopt confidently and defend what you adopt. The companion to all four principles is to keep a clear-eyed view of the vendor landscape, which our market map provides, so that buying decisions are made against an accurate picture of who does what and who owns whom.

The builders furthest along in autonomous sourcing tend to converge on the same argument, and it maps onto this framework: the point of recruiting automation is not to remove the recruiter but to remove the parts of the job that never needed a human, so the human can spend their time where judgment actually changes outcomes. That is the disposition this report recommends: aggressive automation of the mechanical, disciplined protection of the human, and relentless measurement of the only thing that ultimately matters, the quality of who you hire and whether they stay.

11. The 2027 outlook

Forecasting in a field moving this fast is hazardous, but the structural forces identified throughout this report point in consistent directions, and a research house owes its readers a view. The honest forecast is not that AI will suddenly do everything, nor that the current limitations are permanent. It is that the trajectory already visible in the 2026 data will extend, with the center of gravity shifting from AI-as-feature to AI-as-agent, and with the human role narrowing toward the irreducible core of judgment and trust.

The clearest near-term trend is that the agent ambition becomes agent reality in narrow domains first. The 52% of talent leaders who plan to deploy autonomous agents will mostly deploy them in high-volume, well-bounded contexts, frontline and hourly hiring, where the decisions are repetitive and the cost of error per decision is low. Autonomous agents will struggle longer in complex, senior, and specialized hiring, where context and relationship dominate. Expect the headlines to outrun the reality, and expect the genuine progress to be concentrated where the problem is structurally suited to autonomy. The models themselves will keep improving on the cadence the labs have set, but model capability has not been the binding constraint for some time; integration, governance, and trust have been, and those move at organizational speed, not model speed.

The second trend is that consolidation continues and the rebundling completes. With 90 to 100 companies acquired every quarter and the suite vendors actively absorbing AI capabilities, the standalone point-solution era of recruiting tech is closing. By 2027 expect the major HCM and recruiting suites to offer native AI across sourcing, screening, scheduling, and interview intelligence, with the surviving independents differentiated by either depth in a specific vertical or by an AI-native model that the suites cannot easily replicate. Buyers will increasingly choose between an integrated suite that does everything adequately and a best-of-breed layer that does one thing exceptionally, the same build-versus-buy tension that has structured enterprise software for decades, now playing out in recruiting.

The third trend is that trust and verification become a defining battleground. As fraudulent profiles head toward the projected one-in-four by 2028, the recruiting stack will increasingly be evaluated not only on how efficiently it finds and screens candidates but on how reliably it verifies they are real. Identity, provenance, and the evidentiary record of the hiring process will move from the periphery to the core. The arms race between employer-side and candidate-side AI is not a passing concern; it is a permanent structural feature of a world where the same cheap intelligence is available to both sides of the hiring table.

The throughline of all three trends is the conclusion this report has built toward from its first principle. When intelligence becomes a commodity input, the mechanical layer of recruiting gets absorbed, the value migrates to judgment and trust, and the winning organizations are the ones that automate aggressively where automation belongs while protecting and elevating the human contribution where it is irreplaceable. The teams that internalize that structure will spend the next year capturing real advantage. The teams that chase the autonomous-hiring headline without doing the structural work will spend it generating impressive-sounding metrics that do not show up in the quality of who they hire. The data already tells you which path leads where.

This report reflects the state of AI in recruiting as of June 2026. The labor market, model capabilities, vendor ownership, and regulatory landscape change rapidly. Verify current figures and tool capabilities before making procurement or process decisions. For the full vendor ecosystem, read our companion Talent Acquisition Technology Market Map: 2026.