Trust in AI Employees Starts with Hiring. Ema Recruiter is now Warden Assured
Ema’s AI Employees run work across HR, Finance, and IT on one governed platform. Ema Recruiter, which opens the hire-to-retire lifecycle in Ema’s HR Hub, is now Warden Assured: Warden AI tests it for bias every month.
Enterprises now give AI Employees work that used to wait in HR, Finance, and IT queues: answering pay and leave questions, processing invoices, resetting access, and screening applicants. Each of those tasks touches a person, a payment, or a system. Buyers want evidence that the AI does the work the way the vendor says it does. In hiring, they want that evidence from someone other than the vendor.
Recruiter is where the employee lifecycle starts in Ema’s HR Hub. The candidate it screens today is the employee Ema onboards next month and supports for years after. Fairness in the first decision sets the standard for every decision that follows.
Earlier this month, we published our own audit of how Ema Recruiter scores candidates. It showed our method and our numbers. This post covers the next step. A third party now runs the tests, on its own data, on a fixed schedule.
Warden tests Ema Recruiter the way your recruiters use it
Warden AI audits AI systems used in hiring. It tests each system from the outside. Warden sends inputs, records the outputs, and compares results across demographic groups.
For Ema Recruiter, the inputs are a job description and a candidate profile. The output is the match score. Warden built 49,140 test samples and each set of profiles shares the same skills, experience, and qualifications and differs only in demographic signals.
If the score moves when only the demographic signal changes, the test catches it.
Warden runs two tests:
- Disparate impact analysis (group bias). Warden counts the share of each group that scores above the median and compares it to the top scoring group. An impact ratio of 0.80 or higher passes. This matches the four-fifths rule, a guideline established by the U.S. Equal Employment Opportunity Commission (EEOC).
- Counterfactual analysis (individual bias). Warden swaps one demographic signal on the same profile and checks whether the outcome holds. A consistency score of 95% or higher passes.
Why do AI recruiters need to be tested?
Independent testing gives buyers proof they can check. Recruiting teams ask us one question before they let AI touch a shortlist.
Does it score candidates differently because of who they are? Our own audit answered that with data. Buyers, legal teams, and candidates also want a third party to check the work.
Regulators ask for the same. New York City Local Law 144 requires an independent bias audit for automated employment decision tools. In January 2026, and October 2025, California and Illinois rules on AI in employment took effect.
The Colorado AI Act takes effect January 1, 2027. The EU AI Act classifies recruitment AI as high risk, with those obligations starting December 2, 2027.
AI systems also change - models evolve, prompts change, and new features ship. A one-time audit describes one version of a product. Warden audits Ema Recruiter monthly, so the results follow the product as it changes.
Ema Recruiter cleared all seven protected classes
Warden tested sex, race and ethnicity, sex and race combined, age, disability, religion, and sexual orientation. Every class came back clear on both tests. The table shows the lowest result in each class.
Ema Recruiter scores candidates on the job, by design
Ema Recruiter compares a candidate profile to a job description and returns a match score. We build and test it around these guardrails:
- Score against the role: The match score reflects the skills, experience, and qualifications the job needs.
- Test the proxies: Bias often enters through signals that stand in for a protected class, such as a pronoun, a graduation year, a university, or a society membership. Our internal audit and the Warden audit both change these signals and check the score.
- Prove the model reads quality: In our internal audit, deliberately weakened profiles scored 1.8 to 2.6 points lower. The model responds to qualifications.
- Separate bias from noise: We measured how often the model’s own variation changes a shortlist, 4.4% in our audit. A bias finding has to clear that baseline.
- Audit every month: Warden tests Ema Recruiter on a monthly cadence and can run an audit on demand after a product update.
- Keep recruiters in the decision: Ema Recruiter ranks candidates so recruiters spend time on the applicants who fit.
Fair hiring is where hire-to-retire starts
Most AI recruiting tools stop at the offer letter. Ema Recruiter sources candidates, runs outreach across channels, and handles automated screening, and the hire it helps make keeps working with Ema long after.
When a candidate accepts, the HR Hub’s onboarding AI Employees set up payroll, identity, and devices, and build a 30-day plan. Employee support AI Employees answer pay, leave, and benefits questions in seconds. On the last day, offboarding AI Employees close out access and paperwork. All of them share one context graph, so what Ema learned at hiring carries forward. Nobody has to re-enter it in five systems.
Hiring is the most regulated AI decision in HR, which makes it the right place to prove the platform. The controls behind this result run under every AI Employee in the Hub: immutable audit trails, human approval where it matters, automatic PII redaction, and ISO 42001 certification. At Wipro, that platform serves more than 250,000 employees and cut HR response times from five days to seconds.
Customers get audit evidence they can share
- Evidence for vendor review: Your legal, compliance, and procurement teams can review the Warden report during evaluation.
- Support for your compliance work: Independent audit results support the documentation many AI hiring rules ask employers to keep.
- Results for the version you run: Monthly audits cover Ema Recruiter as it ships today.
- Answers for candidates: When a candidate asks how AI scores them, your team can point to third-party results.
- Less work for your team: You spend less time building your own bias tests for the tool.
Hire with evidence you can check every month
Hiring teams want AI that speeds up screening and gives every candidate the same shot. Ema Recruiter now has independent monthly evidence on both counts. Because it runs on the same platform as the rest of Ema’s HR Hub, that standard follows every hire from the first screen to the last day.
Book a demo of Ema Recruiter and the HR Hub, or see Ema’s listing in the Warden Assured directory.