The line that changed Athena Karp’s career was not outside a nightclub or an airport lounge. It was for the men’s bathroom at an investment bank. During a break in her first job after college, she saw men waiting and no queue at all for the women’s room. In ordinary life, the pattern often runs the other way. On Wall Street, the reversal made the workforce visible in a way an org chart never could. Who had entered this building? Who had not? And what, exactly, had decided the difference?
The observation followed her. Karp had graduated from Georgetown’s School of Foreign Service in 2008 with a degree in international politics, then entered technology and media investment banking at Bank of America Merrill Lynch. She later moved to Altaris Capital, where she worked on healthcare and hospital-technology investments. The route was a compact education in systems with consequences: public companies, regulated data, expensive decisions, institutional habits.
In 2013, she turned the question from that hallway into HiredScore. The New York company would help large employers find qualified people in sprawling pools of applicants and existing workers. It would use data science and machine learning, but Karp’s real product brief was more demanding than faster sorting. The system needed to be explainable. It needed to be tested for bias. A person still had to own the decision.
“We never focused on ‘exit’, we focused on impact.”Athena Karp, reflecting on HiredScore’s acquisition
A founder trained to notice systems
Karp is sometimes described as an AI founder, which is accurate but incomplete. She arrived at software through finance and policy, not a computer-science lab. That background shows in the questions she asks. What information may enter a model? Which old decisions are too compromised to teach a new system? Can an employer inspect why a recommendation appeared? The concerns sound philosophical until a résumé is accepted, buried, or rejected. Then philosophy becomes workflow.
HiredScore developed into what the company called talent orchestration. The phrase is ungainly, as enterprise-software phrases tend to be, but the job is practical: connect data scattered across recruiting systems, surface candidates who match stated requirements, rediscover people already known to an employer, and help workers find internal roles. Karp offered a better metaphor - a “digital exoskeleton.” The machine provides leverage while the human remains inside it, moving, judging, and accountable.
The company’s growth followed a widening map of work. Candidate screening led to rediscovery, then internal mobility, total talent management, and workforce planning. In a 2022 conversation, Karp said HiredScore had seen more than 600 million hiring decisions. At that scale, the data could reveal which skills were rising, which requirements persisted, and where employers were asking for credentials they did not truly need.
The art of leaving data out
AI companies usually brag about how much data they consume. HiredScore’s more interesting choices concerned what its systems should not eat. Its responsible-AI framework excluded non-job-related signals such as hobbies and extracurriculars. Exact addresses were kept out because geography can proxy for wealth or ethnicity. Gender and race were separated from scoring. Historical recruiter choices could be removed when they reproduced unequal selection patterns. Learning sets were balanced, and recommendations were logged so they could be inspected.
The distinction matters because hiring data is historical by definition. A model trained carelessly on yesterday’s preferences can automate yesterday with impressive speed. Karp’s proposition was that responsible AI had to begin in the architecture, not arrive later as a committee and a PDF. This was also a sales argument. Global employers do not merely need a ranking. They need something that legal teams, recruiters, hiring managers, and candidates can live with.
She carried that position into public policy. In November 2020, Karp testified before the New York City Council in support of proposed oversight for automated employment decision tools. Opaque and biased systems, she argued, could harm workers and employers alike. Her crispest line was only four words: “These problems are avoidable.” The claim contains both optimism and an invoice. If the harms can be avoided, builders are responsible for doing the work.
Two rooms that needed an introduction
Karp’s interest in opportunity did not stop at the applicant-tracking system. She served with Community Education Alliance’s network of charter schools in West Philadelphia, working on public education and job readiness. The experience put her in two rooms that rarely shared notes. In one, corporate talent leaders explained the skills they could not find. In the other, school leaders asked what students should learn to become ready for work.
The absurdity was structural. Employers possessed a live feed of demand but struggled to convert it into usable guidance. Educators wanted the guidance but lacked a clean view into changing jobs. Karp saw workforce intelligence as a possible bridge - not simply a tool for choosing among people after they apply, but a map that could help people prepare before opportunity arrived.
The work also complicates the familiar picture of an AI entrepreneur. Karp is a Georgetown-trained international-politics graduate, a former investor, a 2018 Henry Crown Fellow, and a member of the World Economic Forum’s Global Shapers community. In 2024, Georgetown named her Entrepreneur of the Year. The common thread is less technology than institutions: how they allocate access, how their rules harden, and where a well-designed system might make them less arbitrary.
The cleverest feature in consequential AI may be a boundary: what the system refuses to notice.
The acquisition without an ending
Workday announced its agreement to acquire HiredScore in February 2024 and completed the deal that April. Karp marked exactly 11 years since she had started the company. Her public note thanked clients, colleagues, family, and the women founders who had shared the journey. She also resisted the tidy language of arrival. An exit is an investor’s noun. Impact, in her telling, is a daily verb.
She began the next chapter as general manager of HiredScore at Workday. The fit was already practical: HiredScore had a certified, bi-directional Workday integration, and the two companies shared large enterprise customers. Under one roof, the product could reach further into recruiting, employee mobility, skills, learning, and contingent work. The original machine had acquired a much larger engine room.
Karp did not remain inside the acquired-company box. By the autumn of 2025, her role had expanded to Senior Vice President of AI Strategy, spanning Workday’s broader AI agenda. In 2026, she was discussing autonomous agents across HR and finance, and how customers might scale their use as needs changed. The founder’s problem had grown from matching people with jobs to redesigning how people and machines divide work.
Her language has shifted with the canvas. She now talks about a “chief work officer” mindset - not necessarily another executive title, but a shared responsibility across HR, technology, and operations. Agentic AI can reorder entire processes rather than trim a few minutes from them. That possibility makes readiness and accountability more important, not less. Someone must decide which work changes, how people are prepared, and what good performance looks like when the team includes software agents.
The question inside the machine
There is a pleasantly human footnote to all this gravity. A recruiting podcast once opened with teasing about a Minions hat. Karp confessed she had never seen the film. When the host offered to send a DVD, she replied that she would need to buy a DVD player. The exchange is useful precisely because it reveals nothing strategic. Even people who spend their days talking about the future of work can be defeated by obsolete home entertainment.
Karp’s serious gift is similar: she notices the ordinary object that gives away the system. A bathroom queue. A résumé field. An exact street address. Two meetings discussing the same skills gap in different languages. None looks like a grand theory until she asks what produced it and who bears the cost.
The hiring machine will keep getting faster. It will read more applications, coordinate more actions, and increasingly act on behalf of people. Karp’s work offers a harder standard than velocity. The machine should know its limits. Its builders should be able to explain its choices. And the human at the controls should never be allowed the luxury of pretending nobody is there.