For Beth Wolfe, Senior Director of Recruiting at Daxko, applicant fraud wasn't a theoretical risk. It was a daily frustration — and an increasingly expensive one.

Over her decade-plus at the company, Beth had watched fraudulent applicants grow from an occasional anomaly into a pattern she couldn't ignore. Fake candidates were showing up in her queue on nearly every remote tech role, looking great on paper, scoring well on every dimension she cared about. And by the time her team figured out they weren't real, they'd already spent significant time — and money — finding out.

The cost of manual fraud detection and early point solutions

Before Gem's Fraud Detection Agent, Daxko was using a standalone fraud detection tool that required manually copying and pasting each candidate's name, email, and phone number — one at a time — to run a check. At $1.80 per check, the cost added up fast. On one particularly difficult AI engineer role, a bulk upload to screen the full applicant pool ran nearly $800 for a single job.

But the dollar cost wasn't even the biggest problem. The workflow itself created a compounding inefficiency that's easy to underestimate: because the check happened outside the ATS, recruiters were already deep into reviewing a candidate — reading the resume, checking LinkedIn, building a mental picture of who this person was — before finding out they were fake.

The other limitation of the standalone tool: no IP or location data. Name, email, and phone signals only. Which meant a meaningful category of fraud — candidates applying from overseas under a fabricated US-based identity — was invisible.

Gem helped surface 30% of suspicious applicants already in the pipeline

When Daxko joined Gem's Early Access program for Fraud Detection Agent, the data that surfaced confirmed what Beth already suspected. Across active remote tech roles, fraud rates were significant:

• Senior Product Manager: 27.9% flagged

• SRE Manager: 27.4%

• UX Designer: 21%

• Product Owner: 13%

Assuming Daxko had 1,000 applicants across their roles and a roughly 30% fraud rate, that's approximately 300 applications that would otherwise require manual review. And these aren't the quick dismissals — a clearly unqualified candidate might take 10 to 30 seconds to pass over. A fraudulent candidate who looks qualified is a different story entirely.

At five minutes per fraudulent application, 300 flagged candidates represents approximately 25 hours of review time — saved. That figure doesn't include screening calls avoided, interview time protected, or the downstream cost of a fraudulent candidate making it further into the process.

The IP location signal proved to be a decisive differentiator. A candidate with an established US work history applying from Pakistan. An email address flagged against an FBI list with ties to state-sponsored fraud. These aren't signals a name-email-phone check can surface — and for Beth, they're not abstract risks either. A sister company under the same private equity firm had previously experienced a security incident after inadvertently hiring a fraudulent candidate who gained access to internal systems.

How Daxko uses Gem to catch fraud before they fall in love with the profile

Beth's current workflow in Gem inverts the old process entirely. She opens a role, sorts by AI match score, and checks the fraud risk level — before she reviews a single qualification. The result: she never gets emotionally invested in a candidate she's about to reject.

High-risk flags get a quick scan and a bulk dismissal. Medium-risk flags get examined in context — occasionally there's a legitimate explanation, like an internal candidate whose email was recently migrated after an acquisition. Low-risk, high-match candidates get her full attention.

The counterintuitive insight Gem's Fraud Detection Agent surfaced: fake candidates score disproportionately high on AI matching. Bots are engineered to mirror job descriptions precisely — which means the candidates who look most promising are often the most worth scrutinizing.

The two features working in tandem accelerated results beyond what either could do alone. Beth recently filled a Senior Product Owner role on Daxko's payments team — historically one of the more difficult positions to close given the niche skill set required — faster than usual. Fraud Detection Agent cleared the fraudulent applicants. AI matching surfaced the qualified ones. The shortlist came together quickly.

89% lower cost per check. 25 hours saved per role. One contract.

The financial case is straightforward. At $0.20 per check, Gem's Fraud Detection Agent represents an 89% reduction in cost per check compared to Daxko's previous standalone solution — with more signal, broader coverage, and zero manual workflow overhead.

For recruiting leaders who already use Gem ATS, the consolidation value compounds that further. No separate vendor relationship. No standalone contract. No additional approval process at renewal. Fraud detection baked into the platform means one line item, one conversation, and one less tool to manage.

"Having it baked into where you're already working — to me, that alone is a decision maker," Beth said. "You're saving a click, a copy-paste, a screen switch. You're just in one place, going next, next, next. And then having the IP address data on top of that — knowing where people are actually applying from — that's something you just can't get from a standalone tool that only has email and phone."