Technology
Navigating EV Manufacturing Quality with Karthick Ravichandran
Karthick Ravichandran’s core argument is one the automotive industry is learning under pressure: quality planning ...
Adithya Sekhar Gummadi argues that as AI becomes embedded in HR decision-making, trust depends on traceability, fairness, and human accountability—not speed alone. Drawing on his consulting work and doctoral research, he introduces an ethical blueprint centered on lineage-awareness, bias audits, and human-in-the-loop governance to turn AI from a black box into a defensible process. His framework shows that ethical rigor, when designed into automation from the start, strengthens both operational efficiency and organizational trust.
The integration of artificial intelligence into human resources is rapidly moving from a theoretical advantage to an operational reality. As organizations deploy AI-driven tools to score candidates and recommend compensation, they simultaneously confront significant risks regarding bias, transparency, and accountability. While research shows a high median ROI for AI in HR, addressing the inherent complexities of adoption requires a blend of deep technical expertise and a firm grasp of business ethics.
Addressing this challenge requires a combination of deep technical expertise and a firm grasp of business ethics. Adithya Sekhar Gummadi, a Senior Workday Consultant and PhD scholar, is working at this precise intersection. His research proposes a practical, ethical blueprint for AI in HR, aiming to ensure that automated systems are designed to be traceable, explainable, and governable, which is a fundamental challenge for HR executives.
The motivation for Gummadi’s research stems from direct observations in his consulting work with Fortune 500 companies, where he noted a recurring pattern: organizations experimented with AI-driven strategies while governance lagged significantly behind the technology. This gap created ethical vulnerabilities, as Gummadi identified opaque data paths and a lack of standard fairness checks.
“I saw sensitive decisions, like performance calibration or merit recommendations, influenced by models or heuristics that weren’t documented, weren’t explainable, and weren’t auditable,” Gummadi states. This ambiguity meant organizations could not reconstruct the process with confidence if challenged, framing the problem as an ethical and structural gap rather than merely a tooling issue. Consequently, he focused on an “ethical blueprint” to make AI in HR traceable and governable by design.
A central concept in Gummadi’s work is ‘lineage-awareness,’ which creates a transparent and auditable trail for every AI-driven decision—described to non-technical leaders simply as “the ability to show your work”. This involves tracing a decision backward through every step, aligning with Federated Computational Governance principles by establishing shared data responsibility.
For Gummadi, this traceability transforms AI from a “black box” into a governed process, providing the foundation of trust for multiple stakeholders. He notes that for HR and Legal teams, it provides evidence of what was used in a decision, while for employees, it translates abstract system outputs into understandable factors.
One of the most critical applications of Gummadi’s framework is in identifying and mitigating fairness issues. In a pilot study regarding merit recommendations, his bias and lineage audit rubric uncovered a subtle issue where a combination of location and legacy pay bands indirectly disadvantaged a specific group.
“The model itself wasn’t ‘intentionally biased,’ but it was learning from historical patterns that baked in old inequities,” Gummadi explains. By removing location as a direct signal and adjusting benchmarking inputs, the team took a proactive approach to improve outcomes—a vital step given that strong employee sentiment correlates with 40% higher retention rates. Gummadi emphasizes that without this structured rubric, the issue would have remained hidden; instead, they identified a concrete fairness issue they could “define, explain, and fix”.
A common concern is that the diligence required for ethical AI will hinder the speed that automation promises. However, Gummadi argues that the true problem lies with unstructured automation. His work focuses on designing flows where fairness checks are integral, “first-class steps” rather than afterthoughts.
In his designs, orchestrations run fairness checks before proceeding to manager review, building a human-in-the-loop dynamic directly into the workflow. “The same automation that reduces manual effort also forces discipline,” Gummadi states, ensuring that speed and ethics move in tandem because the process is designed for both from the ground up.
The reliability of any AI system is fundamentally dependent on the quality of its underlying data. Gummadi’s research addresses ‘cross-module friction’—data inconsistencies across different HR systems—warning that without intervention, “Your ‘AI’ will just learn and scale those inconsistencies”.
His work implements systematic patterns like standard staging layers to create cleaner data, likening this preparatory work to tuning a car engine before installing a self-driving assistant. “It doesn’t make the AI glamorous,” Gummadi notes, “but it makes it reliable and defensible”.
Even with advanced automation, Gummadi positions AI as a “senior analyst, not a hidden boss,” where the system proposes, but the human disposes. A key feature of this design is structured override with learning: if a manager disagrees with a recommendation, they can override it but must provide a brief reason.
These logged overrides empower managers while providing the organization with valuable feedback to refine its models. The objective is clear: “AI as a transparent advisor, with humans clearly responsible, equipped with both insight and the right to challenge the system”.
Implementing a transparent AI framework involves more than technical architecture; it requires a significant cultural shift. Gummadi sees major hurdles, including discomfort with full transparency and an attachment to the unquantified “art of HR.” He observes that leaders may resist when AI becomes “a mirror showing patterns in their decisions” rather than a tool they can blame.
Many leaders are accustomed to relying on intuition, a practice that can be challenged when AI surfaces patterns in their decisions. Gummadi says, “They may accept AI if it’s a black box they can blame or ignore, but resist when it becomes a mirror showing patterns in their decisions.” This shift from intuition to evidence-based decision-making requires new skills and a culture that values learning over blame.
Successfully navigating this transition requires more than a top-down mandate, a common critique of frameworks like Kotter’s 8-step model. Gummadi has found success by starting small with concrete processes and involving HR, Legal, and business partners early. He concludes, “I show that transparency reduces personal risk—because you can prove you followed policy and relied on governed tools, not ad-hoc spreadsheets.”
As AI becomes further embedded in the workplace, frameworks that prioritize transparency, fairness, and human oversight are no longer optional—they are essential for building sustainable trust. The work of consultants and researchers like Gummadi provides a crucial playbook for organizations seeking to balance technological innovation with ethical responsibility. By ensuring that automated systems augment human judgment rather than replace it, organizations can move beyond the fear of the “black box.” Instead, they can embrace a future where AI serves as a reliable, governable partner, driving operational efficiency while strictly adhering to the values of fairness and accountability.
