How Ariel Chen built trust before she built automation

Ariel Chen coordinates People Operations at Figma, where a two-person team tracks over 100 background checks each month across 12 countries. In a role where a missed screen can delay someone's start date or create compliance issues, Ariel built a system her team can rely on.

When every screen says “pending”

Figma hires globally, and international background checks often take longer than the two-week lead time between onboarding cohorts. 

But every background check showed the same generic "pending" status. The escalation path depends on which specific screen is pending. Stakes range from a routine nudge to FCRA adverse action. Catching the one check that actually mattered meant manually watching all of them.

"The challenge was that an overall 'pending' status doesn't tell you what to do: the escalation path depends on which screen is pending, and the stakes run from a routine nudge to FCRA adverse action."

Before the system, Ariel’s team opened every background check one by one, multiple times a week, clicking into each report to catch those at risk.

Ariel's manager mentioned that Zapier might help automate the tedious part of the process. Ariel had no prior experience building automations, so she started with the basics: could she connect Checkr and Zapier? Learning how API keys and webhooks worked gave her the first breakthrough. Once that connection worked, she became curious about what else the tools could do.

She learned the jargon as she went, including APIs, webhooks, JavaScript, and how information moved between systems. Each small build made the next one easier to understand. The work became accessible because she did not need to know everything before starting.

How she built a system the team could trust

Ariel didn't automate judgment away. She built a system that makes the right human judgment easier to make.

Four Zapier workflows orchestrate the entire process. Live status sync pulls background check events continuously into a Google Sheet that serves as the team's onboarding view. Every Checkr event gets logged to an append-only status log—any report can be reconstructed.

The key insight: daily AI by Zapier triage that reads each pending screen and assigns real urgency. Not just an overall "pending" flag, but which specific screen is pending and how close the candidate is to their start date. Outputs identify who's at risk before it's too late to escalate.

When a check hits the danger zone, the system flags the urgency so that Ariel’s team can route escalations to senior approvers. But all reports receive human judgment through specific workflow steps; automation never replaces final judgment.

The discipline that made it work: Ariel ran a daily accuracy test, comparing the results against Checkr screen by screen for five clean days before the team relied on the automation. 

A pattern worth repeating: In high-stakes workflows, earn trust with testing and logging before you lean on automation.

Testing before relying on the system

Ariel didn't treat the first working version as finished. She compared the data and AI summaries against Checkr, the team's source of truth, on a screen-by-screen basis. She paid particular attention to complex checks, including international reports and checks with several pending screens.

The testing revealed bugs, including a row-alignment issue that could cause information to be written to the wrong place. It also showed Ariel that the daily AI summaries needed clearer prompts to be concise and actionable. She ran bug-bashing sessions, fixed the issues, and tested again.

Ariel continued verifying each result until every check matched the source of truth for five consecutive days. Only then did the team begin relying on the workflow.

That discipline matters in People Operations. The system can make the work easier to review, but it can't remove the need for human judgment. Ariel's team still checks the output, works with audit and compliance experts, and uses a manual fallback when the automation can't be trusted.

What three hours a week looks like compressed into minutes

Manual monitoring dropped from more than three hours per week to a daily scan of under 15 minutes. Status entry became fully automated. The two-person People Ops Onboarding team can track more than 100 checks without repeatedly opening every report.

The time savings are useful. Figma's People Operations team supports many cross-functional processes while running lean. Background checks were one of the team's biggest time drains, and repeatedly reviewing the same pending items required sustained focus. The new workflow gives the team more mental bandwidth for other detail-oriented work and more time to focus on priorities on its roadmap.

The larger lesson is about how people build. Ariel started with a problem she understood, learned the technical concepts as she needed them, and kept questioning the output when it did not match the process. When a simpler version of the workflow could not surface the detail her team needed, she chose quality over simplicity and rebuilt the more complex part herself.

The wider signal

This is what People Ops automation looks like when it is built right. Not replacing judgment, but creating the conditions where good judgment can happen faster and at scale. Not trusting the system immediately, but building trust through testing.

“I don’t know how to code, but with Zapier I turned hours of manual work into a system that runs itself and scales with our hiring,” she says. “So now I look at almost any repetitive task and ask, how can I build something for this?”