Fair hiring, from sourcing to offer
Prisma monitors every stage of hiring, from sourcing and CV screening to interviews and offers, against your anti-discrimination policy and regulatory requirements, whether the work is done by people or by AI. It applies consequences and feedback that improve behavior, and keeps the evidence.
Discover the Prisma platformAt a glance
- Department
- Compliance and human resources
- Regulatory context
- Employment anti-discrimination laws (such as Title VII, the ADEA and the ADA in the United States), rules on automated employment decision tools (such as New York City Local Law 144), and the EU AI Act, which classes AI used in recruitment and in decisions about workers as high-risk
- What is monitored
- Hiring decisions at every stage, including sourcing, CV screening, assessments, interviews, shortlisting and offers, whether made by people or by AI
- What Prisma delivers
- Full-population monitoring against your policy, intervention with consequences and feedback before an outcome stands, remediation that improves and stabilizes behavior, and an audit trail for every decision
Fairness is decided at every stage of hiring
Hiring is a chain of small decisions. Where a job is advertised and how it is worded. Which applications are screened in or out. How a CV is ranked. Which questions are asked in an interview and how the answers are scored. Who is shortlisted, and what is offered. Each of these is made by a person, by an AI tool or by both, and each can treat similar candidates differently for reasons that have nothing to do with the job.
Anti-discrimination monitoring is the practice of checking those decisions, at every stage, against the organization's anti-discrimination policy and the rules that apply. It asks whether a decision rested on job-related criteria, whether a protected characteristic or something that stands in for one influenced it, and whether similar candidates were treated in the same way.
For compliance and human resources, this is where fairness is either shown or only claimed. The same question follows people beyond hiring, into promotion, pay and performance decisions.
Where discrimination hides in a hiring process
Discrimination is rarely written down. It tends to work through proxies: a postcode, a school, a graduation year or a gap in employment that stands in for race, age or sex. It sits in job advertisements worded in a way that discourages some groups, in interview questions that drift toward family, health or origin, and in scoring that changes with who is being scored.
AI adds its own risks. A tool trained on past hiring decisions can reproduce the patterns in them, and a ranking can look neutral while consistently favoring one group. When people and AI work in the same process, a problem at one stage can be carried through every stage after it.
What the rules ask of hiring
In the United States, laws such as Title VII, the Age Discrimination in Employment Act and the Americans with Disabilities Act prohibit discrimination in hiring on grounds such as race, color, religion, sex, national origin, age and disability. They reach both intentional unequal treatment and neutral practices that disproportionately exclude a protected group, which is known as adverse impact.
Rules are also being written for the tools. New York City requires a bias audit and notice to candidates for automated employment decision tools, and the EU AI Act classes AI used to place job advertisements, filter applications or evaluate candidates as high-risk, as it does AI used for decisions on promotion, termination or performance. High-risk AI is expected to come with human oversight, monitoring while it operates, and records.
Reading every hiring decision against your policy
Prisma evaluates each hiring decision, at every stage, against the anti-discrimination policy and regulatory expectations your organization has set, and returns a result of PASS, FAIL or AMBIGUOUS, together with the reasoning behind it. A PASS means the decision meets policy. A FAIL records why it does not. An AMBIGUOUS result means the situation does not allow a clear answer, so it is escalated for a human decision instead of forcing certainty.
The evaluation looks at whether the decision rests on job-related criteria, whether a protected characteristic or a likely proxy appears to have influenced it, whether similar candidates were treated consistently, and whether the reasoning is recorded.
Stopping an unfair outcome before it stands, and fixing what caused it
When a decision falls outside policy, Prisma intervenes before the outcome stands. The outcome is restricted or escalated to a human, and consequences and feedback are applied, to the person or the AI behind it, so that the decision can be put right while it still matters.
Remediation turns that into better behavior. Consequences and feedback improve how sourcing, screening, interviewing and selection are carried out, for people and for AI alike, so the same problem does not repeat and performance improves and stabilizes over time. Patterns by stage become visible, and the evidence of how each decision was judged is kept as an audit trail.
One standard for people and for AI
Sourcing tools suggest candidates, models rank CVs, assistants schedule and summarize interviews, and software scores recorded answers. People do the same work, and often the two are mixed in one process. Prisma applies a single governance standard to hiring decisions made by people and by AI, so an AI ranking is read against the same policy and receives the same PASS, FAIL or AMBIGUOUS result as a decision made by hand. Because it sits across the workflow and not inside a single tool, the same oversight holds at every stage.
Don’t just take our word for it
Frequently asked questions
Anti-discrimination monitoring is the review of hiring decisions, at every stage, against your anti-discrimination policy and the rules that apply. It checks that each decision rests on job-related criteria, that protected characteristics and their proxies did not influence it, and that similar candidates were treated consistently.
Any stage that has a decision and a policy to follow: sourcing and job advertising, CV screening and ranking, assessments, interviews and their scoring, shortlisting and offers. The same approach applies to the HR decisions that follow hiring, such as promotion, pay and performance evaluation. Prisma monitors decisions made by people and by AI.
Disparate treatment is intentionally treating people differently because of a protected characteristic. Adverse impact is when a neutral practice, such as a screening rule or a test, disproportionately excludes people in a protected group. Monitoring has to look for both.
The four-fifths rule is a guideline from the Uniform Guidelines on Employee Selection Procedures. If the selection rate for one group is less than four-fifths, or 80 percent, of the rate for the group selected most often, enforcement agencies generally regard it as evidence of adverse impact. It is a screening guideline and not a legal definition, and your policy decides how it is applied.
Local Law 144 applies to automated employment decision tools used in hiring and promotion decisions in New York City. Organizations that use them must have an independent bias audit within the past year, publish a summary of the results, and give candidates notice that the tool is being used.
The EU AI Act classes AI used for recruitment and selection, including placing targeted job advertisements, filtering applications and evaluating candidates, as high-risk. It does the same for AI used in decisions about promotion, termination, task allocation and performance. High-risk AI is expected to come with human oversight, monitoring while it operates, and record keeping.
Yes. A neutral-looking factor such as a postcode, a school, a graduation year or a gap in employment can stand in for a protected characteristic and produce the same result. Prisma looks at whether a protected characteristic or a likely proxy appears to have influenced a decision, not only whether one was used directly.
PASS means the decision meets your policy. FAIL means it does not, and the reason is recorded. AMBIGUOUS means the situation does not allow a clear result, so it is escalated for a human decision instead of forcing certainty.
Prisma intervenes before the outcome stands. The outcome is restricted or escalated to a human, and consequences and feedback are applied, to the person or the AI behind it. Remediation then improves behavior: how screening, interviewing and selection are carried out gets better, and performance improves and stabilizes over time. The evidence is kept as an audit trail.
No. Your applicant tracking, sourcing and screening tools keep doing their work. Prisma monitors the decisions that come out of them and out of the people using them, intervenes when a decision falls outside policy, and keeps the evidence. Your organization keeps the hiring decision.
Prisma is delivered as an SDK that runs inside your own environment. It can also be delivered through Palqee's managed cloud and interface if you need it.
