Artificial intelligence is moving down the recruitment track at breakneck speed, but what happens when these systems completely lose touch with the human experience?
In a recent episode of the Rec Tech Podcast, host Chris sat down with HR industry veteran Torin Ellis to discuss his new venture, Ngoma—a service designed to bring authentic human oversight back to automated tech.
The full conversation can be explored in detail within the TorinEllis.srt transcript file. Here is a breakdown of the critical takeaways from their discussion.
The Problem: Designing for the “Normative” Case
Most AI platforms, mobile solutions, and wearable tech are developed with a “normative” user in mind. Because developers rarely factor in the edge cases of individuals with physical, auditory, or visual disabilities, automated systems often create unintentional barriers.
Torin highlighted a glaring, real-world example: a recent legal battle involving video interviewing platform HireVue and Intuit. An internal deaf employee seeking a promotion was flagged by the automated system as needing to “exhibit better listening skills” because the tech lacked the adequate context to evaluate an audible disability properly. Downstream mistakes like this hurt qualified candidates and land organizations in costly litigation.
The Solution: Ngoma’s Human “Strike Teams”
While existing HR tech vendors attempt to test for algorithmic bias, they almost exclusively use software to test software. Torin argues that this reliance on “synthetic data”—where algorithms essentially guess what authentic human life looks like—is severely underserving the disability community.
Ngoma provides a crucial “trust layer” by employing a remote workforce from the disability community to manually audit AI systems.
-
The Process: Over a six-to-eight-week discovery assessment, Ngoma’s team takes an AI system through a rigorous 177-to-300-step sequential evaluation.
-
The Metrics: Systems are mapped directly to the Ngoma Trust Index, which measures equity, accessibility, reliability, transparency, and overall human impact.
-
The Deliverables: Organizations receive a plain-English “trust risk memo,” an interactive dashboard, and hard evidence (video captures, text, and data artifacts) detailing exactly where the system fails or succeeds.
The Danger of Multi-Tenant Contamination
For employers who think automated bias is contained strictly to their own candidate pools, Torin dropped a sobering piece of data. He pointed to a Stanford study revealing that on many popular multi-tenant HR platforms, the bias originating in one company’s hiring process can actually cross-contaminate and negatively influence the algorithmic data used by other businesses on that same network.
Because data is shared on steroids across entire platforms, talent acquisition leaders must ask vendors tougher, more discriminating questions before buying into the “shiny” promise of a tech solution.
Humanity is the ROI
Derived from the Swahili word for drumbeat, Ngoma was named after Torin’s core philosophy: people are the true rhythm and pulse of an organization. Designing more equitable tech isn’t a charity project; it’s a massive market opportunity to properly engage the estimated two to three billion people globally living with a disability.
As Torin beautifully put it during the wrap-up:
“The ROI of D&I is greater humanity.”
To learn more about how to audit your automated hiring tools, visit Ngoma.io.
