Fair, transparent, and compliant AI
Building reliability into AI
You define what matters
You articulate exactly what you need: specific years of experience, team management skills, legal requirements. Each criterion should be clear, measurable, and defensible.
Then assign a priority to each criterion. Must-haves automatically flag candidates for accelerated human review when missing. Strong signals fast-track qualified candidates. Nice-to-haves factor into final scoring. Final rejection always requires a human decision.
Evidence, not opinions
The AI searches for concrete evidence in applications. Specific examples, quantified results, skills mentioned explicitly. It flags exactly what it found.
The same criteria apply to all candidates identically, reducing the unconscious bias that natural human screening introduces.
Full transparency
Every score shows which criteria were met, what evidence was found, and exactly why it scored that way. Complete transparency. Never a black box.
Override any decision instantly with one click. Disagree with the score? Decide the outcome yourself. The override is recorded on that candidate's record. Adjust the underlying criteria and the system re-assesses every candidate against the new definition. You stay in control.
Refine with every hire
As you hire, learn what actually predicts success. Adjust priorities, add criteria, remove what didn't work. Changes apply instantly—even retroactively.
Your criteria evolve with experience. Build a repeatable process that improves with each hire instead of the same generic description.
No training on your data
Customer and applicant data is never used to train, fine-tune, or otherwise improve any AI model — neither ours nor any third-party provider's. AI is inference-only against the criteria you define.
No AI-generated rankings
AI evaluates each criterion independently. The weighted score that ranks candidates is built using procedural code — never AI.
No AI processing of special-category data
Special-category data (health, ethnicity, religion) is not collected by default. When customers opt in, it is stored separately and never routed through AI.
Continuous monitoring
Weekly automated checks for bias, hallucinations, and regressions in performance ensure our AI is always accurate and fair.
Improving candidate experience and hiring outcomes
Fairness
Systematic criteria reduce unconscious bias and ensure consistent evaluation for all candidates.
Transparency
Every AI decision is explainable with clear reasoning and evidence from candidate materials.
Human control
AI recommends, you decide. Hiring teams retain full authority over all hiring decisions.
Frequently Asked Questions
What is First's approach to AI ethics?▼
We treat AI as decision-support, never as the decision-maker — humans stay in the decision loop for every consequential outcome. We do not train any AI model on your data, do not generate AI rankings (the weighted score that ranks candidates is built with procedural code, not AI), and do not route special-category data through AI by default. Weekly automated checks for bias, hallucinations, and regressions in performance ensure our AI stays accurate and fair. Our full commitments are summarised in the section above.
What does First's AI actually do?▼
It screens candidates based on criteria you define—skills, experience, requirements like right to work, behaviors. The AI finds evidence in CVs and applications, scores each candidate against your criteria, and ranks them so your team can review the strongest matches first.
How do I set up screening criteria?▼
You define exactly what matters for the role. For each criterion, you decide if it auto-rejects (must-have), fast-tracks (strong signal), or just scores (nice-to-have). Built-in guidance helps you create fair, job-relevant criteria that are legally defensible.
What if I don't have formal job requirements?▼
That's the point. First forces you to articulate what actually matters instead of vague job descriptions. You'll discover which criteria actually predict success. As you hire and learn, you adjust criteria—your system gets smarter with every hire.
Can the AI auto-reject candidates?▼
No. The AI flags candidates who don't meet must-have criteria for accelerated human review, but final rejection always requires a human decision. Your team can override any AI assessment at any time. Fully automated rejection is not available.
How does First reduce bias?▼
By replacing subjective judgments with systematic criteria. You define what matters; the AI looks for evidence (not opinions) in CVs and applications. The same criteria apply identically to every candidate, reducing unconscious bias that natural human screening introduces.
How is this different from other AI hiring tools?▼
Most AI tools compare CVs to job descriptions—which introduces bias and lacks transparency. First systematizes your hiring logic. You control what the AI looks for. It finds evidence, not opinions. Every decision is explainable and can be overridden.
Can I see the evidence behind each score?▼
Yes. Every score includes which criteria were met, exact quotes from the CV or application that triggered each score, and plain-language reasoning. You see exactly what the AI found and why. It's not a black box.
What if I disagree with the AI's score?▼
Override it instantly with one click. Your judgment always wins. The override is recorded on that candidate's record. When you adjust the underlying criteria, the system re-assesses every candidate against the new definition. You're the expert; the AI applies what you define.
Where is candidate data stored?▼
In London, United Kingdom (eu-west-2) and Dublin, Ireland (eu-west-1) on secure cloud infrastructure (Amazon and Vercel). All data is encrypted at rest and in transit. We use enterprise-grade security with role-based access controls and regular audits.
How long do you keep candidate data?▼
Applicant data is retained for 6 months by default; employers may set a different retention period, communicated via their own privacy policy, and can ask the candidate's permission to keep data for longer. Candidates can delete their data at any time via links in our emails or by emailing privacy@firststage.co. Customer data is retained for 12 months after contract termination, then permanently deleted.
Is First GDPR compliant?▼
Yes. We have a Data Processing Agreement (DPA) available for review. All subprocessors comply with GDPR. Data is hosted in London, United Kingdom (eu-west-2) and Dublin, Ireland (eu-west-1). International transfers use Data Privacy Framework and Standard Contractual Clauses.
Can I audit my hiring decisions?▼
Yes. Complete audit trails show every decision with timestamps: AI recommendations, criteria met/unmet, evidence cited, and which team member made the final call. Perfect for compliance reviews or defending hiring decisions.
Can candidates access their data?▼
Yes. Under GDPR, candidates can request all data we hold on them. We respond within 30 days with their application data, AI scores, reasoning, and evidence cited. They can also request deletion anytime.