AWS logo in LAX airport in advance of its conference in 2024
News Analysis

AWS Wants to Sell You AI Hiring. Remember 2018?

6 minute read
Lance Haun avatar
By
SAVED
Amazon's new AI hiring tool, Connect Talent, could be technically flawless and still fail for reasons unrelated to the product itself.

The history of technology companies trying to fix recruiting is long. The part where they succeed is much shorter.

Last week, Amazon Web Services (AWS) announced Connect Talent, an AI-powered agentic hiring platform that conducts voice interviews, scores candidates on competency frameworks and hands recruiters a dashboard with transcripts and rankings. AWS states it's built on decades of the company's own hiring science, and that it's designed for employers who need to hire a lot of people fast.

Those claims hit a bit differently when you know the backstory.

In 2014, Amazon's engineers started building an internal AI recruiting tool. By 2018, Reuters reported the company had quietly scrapped it after discovering it had taught itself to penalize resumes containing the word "women's," downgraded graduates of certain women's colleges and favored verbs statistically more common on male engineers' resumes. The model had learned from Amazon's own historical hiring data and faithfully reproduced what it found there.

Amazon didn't say much about it at the time. Now the company is trying to sell an AI hiring tool to other employers.

Table of Contents

Tech Giants Fail at HR Tech in a Very Specific Way

Amazon isn't the first large technology company to decide that HR tech is a natural adjacency.

Google launched Hire by Google in 2017, built on the Bebop acquisition and positioned as the obvious ATS for smaller companies already living in G Suite. People said it had a real shot, including me. It was shuttered by September 2020, after Google decided it wasn't strategic enough to keep funding.

Facebook has tried its hand at job postings multiple times. Its first effort was back in 2012, in a partnership that never really panned out. Five years later, it launched Jobs, using the argument that the social graph was a natural fit for connecting employers and workers. It scaled up, then ran into civil-rights lawsuits over ad targeting tools that allowed employers to exclude candidates by age, gender and religion. Facebook shut it down globally in February 2022. Tried again. Shut it down again in 2023. Quietly relaunched a more modest version in Marketplace in late 2025, this time scoped to local entry-level work, which is a very different pitch than the original one.

The causes for failure were different each time. Google's problem was commitment. Facebook's problem was a broken product architecture before launch. Amazon's 2018 problem was that the training data carried the bias of years of actual hiring decisions.

What they share is a structural reality that keeps reasserting itself: HR tech is not a natural adjacency for companies whose core business is search, cloud infrastructure or social media.

Selling into the HR market at the enterprise level requires CHRO relationships built over years, compliance expertise that doesn't transfer from other verticals and training data that's relevant to hiring.

The companies that have actually built durable positions here — LinkedIn with its professional identity graph, Indeed with its two-sided job-search marketplace, Paradox with its 189 million candidate conversations before Workday acquired it last year — got there by owning something specific and defensible, not by deciding that hiring was close enough to what they already did.

AWS is a cloud platform. The channel advantages are real, and they matter. The product heritage is not.

Amazon's Version of the Problem Comes With Extra Baggage

Most tech giants walk into HR tech with a disadvantage. Amazon walks in with a credibility gap and a documented case study.

The 2018 bias incident isn't ancient history. It's the founding example of the regulatory environment Connect Talent is launching into. New York City Local Law 144, enforced since July 2023, requires bias audits, candidate notice 10 business days in advance of using an automated employment decision tool and public disclosure of results. Illinois HB 3773 took effect Jan. 1, 2026. The EU AI Act's employment provisions are enforceable August 2026. When researchers and regulators were writing those frameworks, they were largely thinking about cases like Amazon's.

Connect Talent enters a compliance environment that was, in part, built in response to what Amazon did before. But even if you believe they’ve solved that problem, there’s a more difficult issue. Do companies really want to copy Amazon’s homework?

The company has spent years in active conflict with workers over conditions, surveillance-driven performance quotas and unionization efforts. The Teamsters organized roughly 25 units with 10,000 workers by the end of 2024, and a pre-Christmas strike that year was the largest against the company in U.S. history. NLRB proceedings found CEO Andy Jassy had made anti-union statements that violated federal labor law. Fortune reported in July 2025 on a performance review system that employees described as deliberately opaque and designed to move people out.

Not only that, Amazon's warehouse turnover runs around 150% annually. It may have figured out how to hire employees in a high-volume environment but I’m not sure anyone is impressed that the process manages to identify people who will stay.

When AWS states Connect Talent is informed by its hiring science, buyers should think about what that science has produced internally. That's not a rhetorical point. It's a due diligence question.

The Candidate Side Has Its Own Problem

Colleen Aubrey, AWS SVP of Applied AI Solutions, acknowledged on a press briefing call that the voice agent is still being refined to sound convincingly human. That's a notable admission for a product currently being shown to enterprise buyers.

It matters because the candidate market has already formed opinions about AI interviewers. Fortune documented widespread hostility to them in August 2025, with job seekers saying they'd rather risk staying unemployed than go through another AI screening.

Connect Talent is launching into a candidate headwind with a voice agent its own team acknowledges needs more work.

Learning Opportunities

AWS's marketing frames this as "humorphism," a philosophy of AI that adapts to how humans work rather than the reverse. Whether that's a real design principle or a brand word for a product that isn't ready yet is something only production deployments will answer.

How Connect Talent Might Gain Traction

None of this means Connect Talent goes nowhere.

AWS's procurement channel and IT relationships are a genuine advantage that Google Hire and Facebook Jobs never had. Large enterprise buyers carry significant committed AWS spend, and adding Connect Talent can be a purchasing convenience rather than a deliberate platform decision.

That's not nothing.

For a buyer who needs to hire 50,000 seasonal workers in eight weeks, already lives inside AWS and doesn't want to run a separate vendor evaluation, the friction of trying Connect Talent is low enough that some will try it.

AWS is explicit that the target is high-volume hiring: seasonal retail, logistics, healthcare staffing, hospitality. Amazon hired roughly 250,000 seasonal workers in 2025. The product is, in some sense, Amazon packaging its own peak-season playbook. The compliance instrumentation it ships with, audit logs, name redaction in recruiter dashboards, recruiter override authority on every AI decision, is at least built to what the regulatory floor requires in 2026. That's more than Amazon's 2018 tool had, and roughly on par with what other HR technology solutions have, at least on the surface.

If Connect Talent succeeds, it won't be because it out-competed LinkedIn's data moat or Workday's CHRO relationships. It'll be because a subset of high-volume buyers treated it as an AWS infrastructure decision rather than an HR technology decision.

How narrow is that lane? At the enterprise level, I have to think it’s pretty narrow.

The Trust Question Doesn't Have a Workaround

Google Hire failed because Google didn't care enough. Facebook Jobs failed because the product architecture was broken from the start. Amazon's 2018 tool failed because it learned from the wrong data and nobody caught it until years later.

Connect Talent could avoid all three failure modes and still run into a problem that has nothing to do with its product. HR technology requires a level of trust from buyers, from candidates and increasingly from regulators, that isn't easy to establish. LinkedIn earned it by being where professional identity lives. Workday earned it by being where HR data lives. Paradox earned it by processing hundreds of millions of frontline conversations before anyone was paying close attention.

Amazon is asking HR leaders to trust that this time is different. That the company that built a biased AI recruiting tool in 2014, that runs a warehouse operation with 150% annual turnover, has now built something worth putting between your employer brand and your candidates.

That's not an impossible case to make. But Amazon hasn't made it yet — and it’s hard to see how it could build a formidable competitor without stronger enterprise tie-ins beyond AWS.

Editor's Note: What else is happening in the hiring technology space?

About the Author
Lance Haun

Lance Haun is a leadership and technology columnist for Reworked. He has spent nearly 20 years researching and writing about HR, work and technology. Connect with Lance Haun:

Main image: adobe stock