NeuroAI NEUROAINEUROAI.SITE
China AI, reported clearly
1,300+ verified stories on models, robotics, BCI and the business of AI.
Browse stories
ESC

From weeks to 15 seconds: how ZhongAn (众安) automated nearly half its claims with a large model (大模型)

ZhongAn's (众安保险) 2025 claims report shows AI closing 45% of health claims in seconds — a real deployment with caveats buyers should know.

2026-11-13 · 903 words · NeuroAI
From weeks to 15 seconds: how ZhongAn (众安) automated nearly half its claims with a large model (大模型)

A claimant in a hospital bed should not have to become a paperwork project manager. Yet for years, that is exactly what filing an insurance claim demanded: scans, forms, waits, and a phone call that never comes.

In its January 2026 report, one of China's largest online-only insurers, ZhongAn (众安保险, HK: 6060), published a claims report showing a large model (大模型) now closes a meaningful slice of cases without a human touch.

The headline number: the fastest health claim settled in 15 seconds.

What "automated" actually means here

Automation in ZhongAn's (众安) report is not a single robot. It is a chain stretched across the claim:

  • A smart assistant guides the user so that more than 90% of health-insurance claimants upload their documents correctly the first time.
  • An AI reviewer reads the documents, checks them against policy terms, and either approves or routes the hard cases to a person.
  • In health insurance, over 45% of claims were closed by automated review in 2025.
  • The fastest case finished in 15 seconds; the company says its AI claim expert can translate a professional conclusion into plain language within 3 seconds.

Car insurance followed a similar path. Online self-reporting reached 88.2%, and for sub-100,000-yuan (about USD 14,000) cases the fastest settlement was 13.3 minutes, with damage assessment as quick as 116 seconds using AI imaging.

The numbers behind the report

ZhongAn (众安) is China's first online-only insurer, established in 2013 and listed in Hong Kong (6060). Its 2025 claims report, released in January 2026, is unusually specific:

  • Full-year claims paid: 17.9 billion RMB (about USD 2.5 billion).
  • Pet insurance: more than 5.6 million animals insured and over 1 million pet claims processed.
  • Pet and accident lines: automated processing above 10,000 cases per month, with an anomaly-detection precision the company put at 98%.
  • Small and micro businesses: 17.88 million claims paid, totaling 421 million RMB (about USD 59 million).
  • Major disasters: the company activated emergency response more than 100 times, prepaying 36.98 million RMB (about USD 5.2 million) through AI damage assessment.
  • The company reported handling an average of roughly 2.9 million claims per day across its full portfolio — a volume that includes high-frequency micro-claims such as e-commerce returns and screen repairs, not only large insurance lines.

How the pipeline works

The logic is familiar from other AI claims systems: OCR turns paper and images into structured data, a model matches the claim against policy and medical knowledge, and straightforward cases auto-close while edge cases get a human. ZhongAn (众安) says it runs an internal AI platform, "Zhong You Ling Xi" (众有灵犀), that logged more than 2 billion large-model calls.

For high-frequency, low-value policies — phone-screen cover, e-commerce return shipping — the volume is enormous and the decisions are repetitive, which is precisely where a model beats a queue.

Pet and accident insurance is where the template is easiest to see. Claims there are image-heavy — a vet photo, a receipt, a policy page — and the model extracts the relevant fields, checks them against coverage, and either pays or flags. ZhongAn (众安) reported monthly automated volume above 10,000 cases in these lines with 98% anomaly precision. For a small-business owner who breaks a tool or a pet owner with a vet bill, the difference between a claim resolved in seconds and one lost in a mailbox is the difference between insurance that feels real and insurance that feels like a slogan.

Why an online insurer had the head start

A digital-native insurer collects claims as data from day one. There is no paper to digitize after the fact, no branch to fax from. That made ZhongAn (众安) a natural place for a large model (大模型) to sit directly on the claim flow rather than bolt onto it. ZhongAn's (众安) edge is also cultural: it was born from e-commerce, where a claim is just another API call, so treating the claim as a product moment came naturally rather than as a grudging digitization project.

The result is a template other carriers now chase: treat the claim not as a back-office cost but as the moment the product is actually delivered.

Honest limitations

  • The numbers are the company's own. They come from ZhongAn's (众安) annual claims report, not an independent audit. Strong, consistent, but self-reported.
  • "Nearly half" is the health line, not the whole book. The 45% applies to health-insurance claims; auto-closed share varies by product, and complex or large cases still need people.
  • Speed reflects volume of simple cases. Much of the gain sits in micro-policies where decisions are mechanical. A complex medical claim is not the same as a screen-replacement claim.
  • Appeals are the missing metric. Auto-approval is only as fair as the route to challenge an auto-denial. The report does not spell that out.
  • Privacy of medical data. Health documents flow through models; the safeguards deserve scrutiny as adoption grows.

What readers can do now

  1. When you buy insurance, ask the claims-AI question: what share of claims are auto-closed, what is the fastest time, and — critically — how do you appeal an automated decision?
  2. Small-business owners: check whether your insurer publishes automation metrics. Faster claims mean steadier cash flow after a loss.
  3. Watch the regulator, not just the marketing: if supervisors issue transparency rules for automated claims, that framework — not a company press release — will decide whether speed also means fairness.

Related coverage

Cite this article

NeuroAI (脑机网), "From weeks to 15 seconds: how ZhongAn (众安) automated nearly half its claims with a large model (大模型)," 13 November 2026, https://neuroai.site/a/na-app-zhongan-ai-claims

Machine-readable: BibTeX · RIS · JSON · Markdown

More in “AI in Action” → · Back to home · Markdown version

China AI, reported clearly
1,300+ verified stories on models, robotics, BCI and the business of AI.
Browse stories