I spent most of last week in Las Vegas for the HR Tech Conference, the largest trade show focusing on—you guessed it— human resources technology.
On the way in, my Lyft driver told me about his former life as the owner of a small creative agency in Dubai. The agency produced commercials, and he charged in the low six figures for each spot. One day a longtime customer called to tell him that they were moving to another agency that would produce similar spots using AI for only $5,000 apiece. This was back when AI was still generating people with six fingers, and he warned them that the quality of the work would be terrible. The customer’s answer was that at that price, they could redo the work twenty times over and still come out ahead.
The conversation set the tone for a week where the inevitability of AI was on full display. Last year everyone was talking about agents. This year they are still talking about them, but the stories have evolved to become more unique to each company, and the products felt more differentiated.
Here are the trends I saw.
A generational line is being drawn between AI-native products and everything else. One founder told me his twelve-month-old product beats a three-year-old rival for no reason other than being newer. I take claims like that with a huge grain of salt (show me what your product can do that theirs can’t), but when things are moving this rapidly, three years already feels like a legacy product.
Investment in recruiting technology is down.George LaRocque, who has tracked HR technology investment for more than nine years, says overall HR tech funding is holding steady, landing between $4.8 and $6.2 billion every year since 2023. Recruiting technology is the segment that is lagging. “Roughly $590 million so far, against more than $3 billion a year at the 2021 and 2022 peak,” he told me. “But across nine-plus years of data, that looks like a correction from an outlier, and it’s already above 2024’s full-year low. Capital flowing into recruiting tech is funding new approaches to the work, applying AI and agentic models to rethink all or part of the hiring process or deliver capabilities that were previously aspirational. Capital is selective, not absent.”
Someseea shakeout coming. Fama CEO Ben Mones told me that he had never seen so many bankers and investors in the hall. His explanation is that there are more assets on the market than anyone knows what to do with. Companies that raised money between 2021 and 2024 now have backers underwater because they have no generative AI story, and they are trying to figure out what their peers think those assets are worth.
Data is the differentiator. A year ago an agent was a product. This year it is a feature any competent team can build in a week, and the competition is over data. Findem’s Hariharan Kolam said it most directly: “An agent is not innovation. What is a real innovation here is three things: context, which is the data around which decisions will be made; a methodology, which is how the work happens; and evidence and verification.” Every vendor has its own vocabulary for the same claim. iCIMS calls it an intelligence layer. Darwinbox calls it a context graph. SAP talks about a foundational data layer. For buyers, this means that demos are not telling you the most important thing: what does the product know about you?
Under strain for years, the application funnel has finally broken. Applying now costs nothing, so a candidate, or a bot acting for one, can produce tailored resumes and submit them to hundreds of jobs in an afternoon. The application used to be a signal of candidate interest in a role. Now it does not even mean the candidate knows the role exists. Terry Terhark of NXTThing RPO had just come from the Indeed conference, where flooded recruiter inboxes were the main topic. “It’s so easy for a candidate to go out and say, tell me the hundred jobs that I matched to, and create a custom resume for each one and apply,” he said.
Fraud is a huge and growing problem. When generating a plausible application costs nothing, there is no longer any friction between a fake candidate and a real one. Yardstik CEO Andrew Johnson sees fraud rates of 10 to 15 percent in gig and remote roles. There was enough going on in screening that I am going to cover it in a separate article.
Pricing is moving off seats and tokens. Buyers are balking at token-based bills because they cannot forecast them. Darwinbox, an India-based HCM vendor that is expanding into the US market, will announce fixed per-company AI pricing within weeks, scaled to employee count. Ema, which sells what it calls AI employees and has raised $140 million from Accel, ADP, and Workday, charges per block of employee requests. Findem’s new ReadySlate product charges per candidate unlocked. The unit of pricing is becoming the outcome, not the compute.
Headless is becoming table stakes. Many of the companies at HR Tech touted their MCPs or that their services could be used easily outside of their platform. iCIMS put its new hiring agent inside Microsoft Teams because that is where hiring managers are active, not in the ATS. Findem exposed its entire platform through MCP, and Kolam says every customer has some kind of builder project going on, from recruiters vibe-coding away the most annoying part of their job to a few trying to build their own HR system. Mones described a future with no dashboard at all, where a customer sends a document describing what it wants and the vendor does the rest.
Transparency has become a feature. Few vendors mentioned the Workday and Eightfold lawsuits, but the cases are forcing recruiting products to show at least some of the work that their black boxes are doing. Fama’s agents trace every decision back to a specific line in the customer’s own policy, and the audit trail is part of the product. Marovi, which offers an AI-enabled version of executive search that compresses the first weeks of the process into a scored shortlist, opened the same evaluation engine to candidates so that an executive can enter a role and see how they would be rated and why.
Size matters, and talent acquisition vendors are quietly trying to becomeworkforce vendors. If the big differentiator is moving from product to structured data about people and what they can do, then the most valuable people to understand are the ones a company already employs. This has been the pitch of the big HCM vendors, and the recruiting point solutions are scrambling to adapt. CodeSignal, a hiring-assessment company for most of its life, is moving into skills intelligence for the existing workforce and the training to close the gaps it finds. Findem rebranded from talent intelligence to people intelligence.
My biggest takeaway from HR Tech this year is not any one product. It is the speed. Pricing models, products, and fundamental assumptions about candidates have all changed in a very short time, and the pace of change is only accelerating.
