EPAM a leading digital and AI transformation company, partnering with global enterprise organisations to ambitious start ups. We employ over 62,000 employees operating in more than 55 countries across six continents. The talent function at EPAM operates at significant global scale, supporting remote, hybrid, and on-site hiring models across multiple regions, with a mature talent attraction capability closely integrated with the company’s operations, security, legal, and compliance teams.
Problem Statement:
As remote hiring expanded globally, the Company identified candidate fraud as a potential area of concern. This included cases of skills exaggeration, impersonation, organized fraud, and activity involving sanctioned regimes. These issues may present risks such as financial loss, intellectual property exposure, regulatory challenges, and reputational impact.
Strategic Alignment:
Addressing candidate fraud supports EPAM’s broader business priorities, including risk management, client trust, ethical delivery, and regulatory compliance. Enhancing hiring integrity also helps protect delivery quality, safeguard clients, and maintain workforce credibility, especially in sensitive environments.
Goals & Success Metrics
The primary goal was to move candidate fraud detection from informal intuition to a repeatable, defensible, and evidence-based framework that could operate at scale.
Quantitative goals included:
- Reduction in confirmed fraudulent hires.
- Improved early detection rates across the interview lifecycle.
- Reduced downstream incidents linked to employee fraud or misrepresentation.
Qualitative goals included:
- Increased confidence among recruiters and hiring managers in identifying and escalating concerns.
- Clearer decision-making standards that reduced ambiguity and bias.
- Preservation of a fair and respectful candidate experience.
Secondary goals included:
- Improving cross-functional collaboration between Talent Operations , security, legal, and compliance.
- Establishing governance models that could evolve with emerging fraud tactics.
Stakeholder Engagement
This work required deep collaboration across multiple functions.
Internal stakeholders included Talent Operations, Recruitment, , HR leadership, hiring managers, corporate security, legal, compliance, and ethics teams. TA played a frontline detection role, while final determinations always sat with cross-functional human reviewers.
External stakeholders included candidates, interview platform providers, identity verification and anti-cheating technology vendors, and regulators in certain jurisdictions.
Clear role delineation was established:
- TA and hiring managers observed and documented indicators.
- Security and fraud specialists assessed patterns and evidence.
- Legal and compliance advised on sanctions, regulatory exposure, and appropriate responses.
- No automated system was permitted to make a final fraud determination.
Approach / Actions Taken
Design & Planning
The Company adopted a structured fraud framework that defined:
- What can be falsified (identity, face, voice, persona, qualifications).
- Observable indicators for each fraud type.
- Immediate mitigation actions during interviews.
- Evidence collection standards.
Processes & Methods
Fraud patterns were grouped into two core categories: human imposters and AI-mediated fraud. Six common fraud scenarios were documented, each with clear indicators and response guidance.
A repeatable micro-interview flow was introduced, combining rapport-building with light identity verification prompts to confirm continuity between the candidate, their story, and their environment.
Technology / Tools
Anti-cheating and identity tools were assessed for:
- Video and audio integrity analysis.
- ID and face matching (where legally permitted).
- Device, screen, and behavioural monitoring.
- Post-interview reporting and evidence quality.
Tools informed decisions but never replaced human judgement.
Ethics & Governance
Strong guardrails were applied around privacy, consent, disability inclusion, and bias. Interviewers were trained to distinguish between genuine fraud signals and behaviours linked to nerves, neurodiversity, or disability.
Communication & Enablement
Recruiters and hiring managers were trained on recognising indicators, collecting evidence, and escalating concerns appropriately, without accusing candidates in real time.
Outcomes & Results
Quantitative Results
- Earlier identification of fraudulent activity in the hiring funnel.
- Reduced incidence of post-hire fraud discovery.
- Improved consistency in escalation and review processes.
Qualitative Results
- Increased confidence across TA and hiring managers.
- Stronger collaboration between TA, security, and legal teams.
- Greater organisational clarity around what constitutes fraud versus acceptable candidate behaviour.
Unexpected Benefits
- Broader awareness of digital identity risk beyond hiring.
- Improved interview quality through greater emphasis on practical problem-solving.
- Stronger governance maturity across remote hiring practices.
Challenges & Do-Overs
Unexpected Roadblocks
- Rapid evolution of AI-enabled fraud tactics.
- Balancing fraud detection with candidate experience and fairness.
- Avoiding over-reliance on tools or false positives.
Key Learnings
- Evidence matters more than suspicion.
- Practical tasks outperform polished CVs in risk detection.
- Automated tools must support, not replace, human decision-making.
Advice for Peers
- Assume fraud exists in remote hiring environments.
- Build cross-functional ownership early.
- Train interviewers to observe, document, and escalate calmly and consistently.
Next Steps / Ongoing Work
EPAM continues to refine its framework as fraud techniques evolve, strengthening training, updating detection patterns, and scaling governance across regions.
Ongoing work includes deeper integration between TA systems and fraud oversight functions, and continued evaluation of emerging tools through an ethics-first lens.
Additional Resources
Olga presented this Case Study on stage at the Australasian Talent Conference (ATC2025) with detailed insights around each kind of AI fraud they experienced. Here is a link to the detailed Slide Deck.
