ScopeAT

Roz Francuz-Harris, vice president of talent acquisition at Zillow, has a simple reason for using AI so much in her work: “I love my family,” she told the audience during her recent Talent Connect talk. “AI is one of the tools I leverage to get time back to be with them.”

But Roz — who has built and led talent teams for IBM, Uber, Slack, and WarnerMedia — also finds AI valuable for another reason. Namely, it gives recruiters more time for the human aspects of their work. 

“AI has the capability to make our process more objective, effective, and efficient,” she said, “which can free us up as human recruiters to focus on things that matter, such as relationships and strategic decision-making.”  

While AI is often the subject of hype, we’ve noticed that the nitty-gritty details of how it can help talent teams is often missing from the conversation. That’s why we found Roz’s talk so refreshing. She offered step-by-step, practical details of how to use AI to increase diversity, complete mundane tasks, and even deal with an insensitive hiring manager. 

Let’s check out the practical tips from her session on Revolutionizing Talent Acquisition: AI Meets Inclusivity and Authenticity. 

1. Make a game of learning AI 

Roz uses generative AI five to ten hours a week, so it seemed like a no-brainer to her that her team would do the same. When she suggested this to her director of talent acquisition experience and enablement, Shahbaz Ali, he told her,“We’re going to gamify this so the team can actually digest what you want them to learn.” 

Shahbaz created “Prompt-a-thon” and then crowdsourced ideas from recruiters, coordinators, sourcers, and talent brand managers, asking them to submit use cases in which Zillow’s internal generative AI, ZGChat, could help. 

In one example, the team prompted the AI to “Act as if you’re giving a presentation on key data findings and theme takeaways from survey responses around our representation recruiting survey.” They shared a hyperlink to Google Sheets with the survey. The tool quickly generated a well-written presentation that touched upon key points, including that respondents wanted more transparent and frequent communication. 

The Prompt-a-thon lasted from April to June, and judges from Roz’s leadership team and the greater People team then gave prizes to the winning teams and individuals who performed strongest during the competition. But most importantly, it got Roz’s team really excited about AI and how it could make their jobs more enjoyable.

2.  Use AI to write peer or performance reviews 

“So, I’m going to give you my secret sauce for how I got 13 peer reviews done in three hours,” Roz told her Talent Connect audience. 

Late this past summer, Roz had a midyear review that had already passed and she was behind in writing 13 peer reviews. 

So she opened up Microsoft Word, hit “dictate,” and told Word what she wanted to say about each employee, what their strengths were, as well as where she saw opportunities for development. She corrected any mistakes or misspellings in the text and dropped it into ZGChat. Her prompt: “Please summarize this feedback for a peer review for a peer that I trust. I want to make sure they get this feedback to help enable their performance going forward.” Within seconds, the technology created the summary. She repeated the process 12 more times. 

“I just saved myself, easily, 40 or 50 hours of work,” she said. “So, again, it’s not about replacing you, it’s about making you more efficient while still being genuine and accurate.”

3. Increase the diversity of your applicants and new hires 

“I believe AI has immense potential to improve diversity and inclusion by discovering talent from underrepresented communities based on skill,” Roz said. She called out two AI tools particularly useful for this. 

The first was LinkedIn. “OK, so for broad-based sourcing, where we can just scan numerous sources for diverse candidate pools, LinkedIn already does that,” she told the audience. Linkedin can hide names and photos for bias-free screening and can also offer diversity recommendations based on an organization’s DEI commitments tab on its company page. “It’s data-driven decision-making,” she said, “based on your company’s values.” 

Roz also mentioned Pymetrics (now Harver), citing Unilever as an example of how this tool can help. Unilever recruits more than 30,000 people a year and processes around 1.8 million job applications. Because they wanted to increase their diversity, they partnered with Pymetrics to build an AI-driven platform that included resume screening, video interviewing, and predictive analysis. “They added AI to their recruitment process,” she said, “and increased the diversity of their new hires by 16%.” 

Roz cautioned that part of the reason Unilever was so successful was because Pymetrics already had safeguards built in against bias. Generally, however, recruiters need to be careful about what we teach AI. “It learns your tone, your voice, and your biases,” Roz said, “and it emulates those and even evolves them in some instances.”  So, stay aware of this as you interact with generative AI.

4. Consider how AI might help in screening applicants

During Zillow’s Prompt-a-thon, recruiters also presented a case study in which ZGChat might help screen applicants (this is not a practice that the real estate company is currently implementing, by the way). They used the prompt, “Based on the following rubric and the following detailed interview notes, would you recommend we advance this candidate to an onsite interview for a Senior Backend Engineer role?” The Zillow team included the rubric and a detailed candidate code from HackerRank. 

ZGChat evaluated the information and generated a report in which it found the candidate had “sufficient competence” but that he or she may not have enough “knowledge of binary data transfer techniques such as Protobuf or Avro.” Overall, though, the AI recommended “advancing the candidate to a virtual onsite interview.” In real life, however, it’s recruiters at Zillow — not AI — who make the decision to move a candidate forward.

5. Ask for help in dealing with a difficult or insensitive hiring manager

If there’s one thing recruiters hate, it’s dealing with difficult hiring managers or having to offer them guidance on how to behave during interviews. AI can provide a lift with that.

In one of her Prompt-a-thon examples, Roz included a theoretical prompt in which a recruiter asked ZGChat to “write an email to a hiring manager who gave biased feedback about a candidate being a mother and not being as much of a fit as other candidates they interviewed.”  Roz made it clear that this was not a real example of a manager at Zillow, but an interesting case study. 

The recruiter asked the AI to be thoughtful, provide research, and give a firm perspective on why the hiring manager’s evaluation was problematic, while still emphasizing that the recruiter and hiring manager were “on the same team striving to get the best hire.” 

The results speak for themselves:

Final thoughts: Remember, humans have a remarkable ability to adapt

Finally, Roz shared an anecdote that put this new era of AI into perspective. She said that she was recently speaking with an older colleague, who shared that when he first started recruiting, he used the White Pages (the personal listings in the paper phone book) to find candidates. Upon the introduction of the internet, he had to learn a new way of working if he wanted to survive. 

“We had a good laugh about that, and then it dawned on me: He’s still working in our industry,” Roz said. “The internet did not eliminate him or his role. Nor will generative AI.”

Uncategorised