BOSS Zhipin (BZ) Now Runs Over 10,000 AI Interviews a Day

BOSS Zhipin (BZ) said on its 2026-08-25 earnings call that employer-side AI interview functions now run more than 10,000 interviews a day, and a deputy CFO for the first time named the expanded AI features as a driver of the quarter's revenue growth per paying customer. On the 2024-03-12 call this AI was only a newly filed in-house model plus one example of posting a job quickly for a start-up boss.

BOSS Zhipin (US ticker BZ) is a Chinese online recruitment platform where employers post roles, find people and chat with candidates. Its money comes from the paid value-added services recruiters buy; average revenue per paying user (ARPPU) measures what one paying customer spends. The employer-side AI is a toolset for the hiring side: it writes and optimises job posts, learns and remembers the hiring requirement in conversation, searches the job-seeker database for a shortlist with reasons, screens resumes, talks to candidates for the recruiter until contact details are exchanged or the candidate declines, and interviews them with a customisable digital-human interviewer that asks follow-ups and issues a report. Its users are start-up bosses, junior HR staff, blue-collar and manufacturing recruiting teams, campus-recruitment teams at large state-owned enterprises (SOEs), contract-recruitment customers, and in-house and third-party headhunters. These tools are packaged into the paid products sold to recruiters, so they are core business, not back-office support.

How the employer-side recruiting AI evolved

The capability and its users widened while the yardstick kept changing. On 2024-03-12 there was only a model: the self-developed recruitment large model had completed generative-AI filing and was in use on the platform, the example being a start-up boss posting a job quickly, with only a public benchmark standing as its result and no usage, revenue or cost outcome [7]. On 2025-03-11 the AI began communicating for the recruiter: once a recruiter said they would try a candidate, the AI handled all later communication until contact details were exchanged or the candidate declined, and it was open only to selected blue-collar enterprise users [6]. On 2025-05-22 two numbers appeared in one call: job seekers answered by the AI showed a 15% increase in achievement rate, and enterprise users of the recruitment assistant a 25% increase in achievement efficiency at the same conversation volume. The two percentages do not share a denominator [5].

On 2025-08-20 the AI interview product for large SOEs had just launched, with a customisable digital-human interviewer, follow-up questions and an interview report; it had already supported more than 20 campus AI interview events at more than 10 large SOEs, drawing nearly 30,000 participants, the first countable usage figure here [4]. On 2025-11-18 the AI communication capability was folded into paid products already on sale, and the company said their average mutual achievement conversion ratio rose 7% as a result, changing the yardstick again [3]. On 2026-03-18 AI-assisted interviewing went fully live on the enterprise and job-seeker sides of the platform's own interview rooms, with usage and repeat usage said to improve noticeably, though management still called it an experiment and gave no number [2].

What the latest quarter showed

The 2026-08-25 call brought two new things. AI interview functions run more than 10,000 interviews a day for white-collar and blue-collar enterprise customers, inside a loop from sourcing to onboarding [1]. And a deputy CFO said ARPPU rose 7% year on year, driven by more efficient and more valuable services, naming the expanded AI-powered features as encouraging higher customer spending [1].

The path is narrow: the AI is not billed separately but folded into products recruiters already buy, so customers pay more for the same package. It reaches ARPPU first and revenue second, and it explains price per customer, not the number of paying customers.

The evidence boundary is clear. AI is one of several drivers named; the share of the 7% belonging to AI was not broken out, and the surfaces earning it were not listed. The company has never separately disclosed AI-related revenue, cost, gross margin or headcount effects, and the latest quarter is not yet in the structured financial database, so absolute figures cannot be independently checked. The same call applied AI candidate search to in-house and third-party headhunter workstreams, reaching senior candidates outside the monthly active user base, but gave a capability and an expectation only, with no realised output [1].

What to watch next

Employer-side recruiting AI at BOSS Zhipin now runs more than 10,000 AI interviews a day, yet five successive efficiency yardsticks do not connect, AI is only one driver of the 7% year-on-year rise in ARPPU, and its revenue, cost and headcount effects have never been disclosed on their own.

Only one follow-up would change that: the company breaking out, on a future earnings call, the share of ARPPU growth coming from AI, or giving an AI-related revenue figure on a clear basis. Until then, more than 10,000 interviews a day is usage that connects to no money anyone can check.

Application assessment

• AI on the employer side of BOSS Zhipin, the block of business that helps a company fill a role, covers every recruiter-facing AI surface named across the window: AI-assisted job posting and posting optimisation for start-up bosses and junior HR (2024-03 onward), the beta AI agent that turns a stated hiring need into a justified candidate shortlist, AI Quick Hiring and the high-end 'AIS2CHAO' sourcing tool over a 250 million job-seeker database, AI search/AI sourcing for in-house and third-party headhunters, the AI communication assistant that talks to candidates on the recruiter's behalf, the HAMR recruitment agent, AI resume screening, and the AI interviewer (customisable digital-human interviewer, follow-up questions, interview report) sold to large SOEs and contract-recruitment customers and running inside the platform's own interview rooms. First reported in this window on 2024-03-12. - Business position: core business - Deployment stage: limited production - Scope: single business unit - Value type: revenue growth

Sources

[1] Drillr - BZ - 2026-08-25 - earnings call

"Our AI interview functions are now working on more than 10,000 interviews every day."

[2] Drillr - BZ - 2026-03-18 - earnings call

"From the JobSeeker side, our AI-assisted interview feature experiment has been fully rolled out to users within both GP's interview rooms."

[3] Drillr - BZ - 2025-11-18 - earnings call

"As a result, the average mutual achievement conversion ratio of these products has increased by 7%."

[4] Drillr - BZ - 2025-08-20 - earnings call

"Using this technology, we have supported over 20 AI-powered interview events for recent graduates at over 10 large state-owned enterprises, attracting nearly 30,000 participants, and the response has been positive so far."

[5] Drillr - BZ - 2025-05-22 - earnings call

"For the recruiters who have great scale used our AI recruitment function, efficiency of achievement increased by 25% under the same amount of conversations."

[6] Drillr - BZ - 2025-03-11 - earnings call

"The process unfolds as follows: once the recruiter selects a candidate, they deem suitable and explicitly express their willingness to proceed by saying, I'm willing to give this candidate a try. The AI handles all subsequent communication on behalf of the recruiter."

[7] Drillr - BZ - 2024-03-12 - earnings call

"The effect of this model has reached the industry leading level on some public benchmarks and has gradually been applied in some recruiting and job seeking scenarios."