EY Built an AI Hiring Avatar

 Most AI hiring tools are optimization problems dressed up as equity initiatives. EY's Coach Eve is something more considered than that, and the reason why is stranger and more instructive than the product itself. 

Key Points

• EY receives nearly 400,000 job applications annually and hires 5,000 to 8,000 candidates, creating a scale problem no human recruiting team can solve fairly.
• Coach Eve, an AI avatar built by EY's Intelligent Realities Lab, gives candidates access to practice interviews, case study prep, and cultural questions without their conversation history being shared with recruiters.
• Artists-in-residence working alongside engineers shaped key design decisions, including a head-tilt animation that cuts perceived response latency nearly in half.
• EY deliberately kept Coach Eve visually stylized rather than photorealistic, drawing on research into the uncanny valley and work by artists exploring identity and representation.
• The first artist-in-residence to transition into a full-time AI engineering role at EY describes the core tension in responsible AI: "A lot of people will say 'Keep humans at the center' but then they'll make something that automates the very thing a person does."

What EY Built and Why the Scale Problem Matters

Forbes reported that EY developed Coach Eve through its Intelligent Realities Lab as part of a broader virtual platform called WeVerse. The avatar is grounded in EY's internal knowledge base and leads candidates through practice interview questions, simulated case studies, and practical questions about culture, dress code, and compensation — the kind of questions a nervous applicant would never ask a human recruiter.

The equity argument behind it is straightforward. Applicants from Ivy League schools arrive at EY interviews knowing how to run case studies and carrying the institutional knowledge that alumni networks provide. Applicants from community colleges and state schools are equally capable and substantially less prepared for the specific format. "We really needed to get away from the perfection of a resume and start looking at potential," said Ginnie Carlier, EY's Americas Chief Talent and Culture Officer.

Coach Eve gives every applicant the same preparation access, with one meaningful privacy guarantee: candidate chat histories are never saved or shared with recruiters. The interaction is purely for the candidate's benefit. That design choice is not incidental. It's what creates the psychological safety that makes the tool useful. In one early test, an intern spent ten minutes asking Coach Eve about the difference between EY's pension and 401(k) — the kind of granular question that shapes whether a talented person actually accepts an offer, and that no applicant would raise in a formal interview.

Why Engineers Alone Didn't Build This

The more unusual part of the Coach Eve story is how it was designed. EY's Intelligent Realities Lab, led by Domhnaill Hernon, runs a formal partnership with NEW INC, the New Museum's talent incubator for artists working at the intersection of technology, design, and culture. A rotating cohort of artists-in-residence works directly alongside the lab's engineers, and their input shaped Coach Eve in ways that are measurable, not decorative.

The clearest example is the latency problem. Even after substantial engineering work, Coach Eve took about three seconds to respond to a question — long enough that users found it uncomfortable and unnatural. The engineering team had optimized as far as the technology allowed. The artists-in-residence, through conversations about how human beings perceive the duration of events, offered a different approach. The team added an animation: when a candidate asks a question, Coach Eve tilts its head and enters a brief pause, as if considering the response. That animation consumes about 1.5 seconds and cuts the perceived latency nearly in half. The actual response time didn't change. The experience of waiting did.

"Time is relative," said Hernon. "You have little tricks that you can play to change the perception of time. And that insight came from working with our artists."

The avatar's appearance reflects the same thinking. Coach Eve is deliberately not photorealistic, even though the technology exists to produce a convincing human likeness. The decision draws on work by artist Josie Williams, whose project Ancestral Archives explored how avatars modeled on West African masks avoided the uncanny valley — the discomfort people feel when something almost-human reveals itself not to be. A stylized avatar lets candidates make mistakes, ask naive questions, and practice without the social pressure that a near-human face would create.

What This Model Says About How AI Tools Get Built

The standard model for enterprise AI tool development runs through product managers, UX researchers, engineers, and legal review. The EY model adds a third language to the room, one that asks different questions. Who is this impacting? Are we talking to them? What does it feel like to interact with this from the inside?

Josie Williams, who completed an artist residency with the Intelligent Realities Lab and was subsequently hired as an AI engineer, describes the tension directly. "A lot of people will say 'Keep humans at the center' but then they'll make something that automates the very thing a person does."

That observation lands differently when the tool in question is screening job applicants. The stakes in hiring AI are not abstract. Bias encoded at the design stage compounds at scale. A tool that processes 400,000 applications and systematically disadvantages any group is a discrimination engine, regardless of intent. EY's approach doesn't eliminate that risk, but the presence of artists asking whose experience is centered — and who is missing from the room — creates a friction that pure engineering optimization does not.

Hernon is candid about what the translation function costs when it's absent. "The language artists use is very different. You might be the most gifted person who could profoundly change a business, and no one will know because they literally can't understand what you're saying."

What Marketers Should Take From the Coach Eve Model

The instinct in marketing AI adoption is to evaluate tools on output quality and workflow efficiency. Coach Eve is a useful case for a different question: what does the design process behind this tool reveal about whose experience it was built to serve?

That question matters for AI marketing tools in the same way it matters for hiring tools. Personalization engines, content generation systems, and customer-facing chatbots all encode assumptions about who the audience is and what they need. Those assumptions get baked in at the design stage and are rarely visible in a product demo.

EY's Chief Innovation Officer Joe Depa put it plainly: "Human decisions — especially the ones that require ethical judgment — can't be delegated to a machine. We need curiosity and empathy to craft good prompts, interpret outputs thoughtfully, and spot issues a machine might miss."

That framing applies directly to how marketing teams use AI. Prompting well, reading outputs critically, and catching what the model misses are human skills that don't get less important as the tools get more capable. They get more important, because the outputs reach more people faster. If you're thinking through how to build AI into your marketing operations in a way that holds up under scrutiny, our growth strategy team works through those design questions with organizations before they become problems at scale.