Anima Cuts Clinic Call Waits by 75% with GPT-6 and GPT-Live

Hi everyone, Shii-chan here! Today we're looking at an OpenAI customer story where AI is helping out in clinics.

OpenAI News openai.com

What was announced?

In OpenAI News, the healthcare platform Anima was featured. Anima brings patient requests, clinical documents, and consultation notes into one platform, with AI teammates that gather patient histories, process documents, and draft clinical notes.

It began in the UK, where it serves over 14 million patients across more than 1,000 clinics and processes more than 800,000 cases a month. It's expanding into the US, Australia, and beyond.

Why it matters

When work piles up, patients wait longer and important information sits in queues. CEO Shun Pang says he started Anima after seeing, during clinical training, a patient whose abnormal test results sat in a queue for six months.

His view is that the problem is less about medicine and more about how information moves, so Anima aims to give care teams more capacity to respond.

What changes

Here are the results Anima reports:

• Phone waits: from 40-50 minutes down to 10 minutes (at least a 75% reduction)
• 86% of surveyed patients said they would use the voice agent again
• Documents: clinics that processed about 15 per hour now handle 30-45 (top clinics 45-60)
• Some clinics cleared backlogs of over 100,000 documents within days
• Consultation notes and referral letters are drafted in seconds (referral letters used to take 5-15 minutes)
• Most cases are resolved within 24 hours, with a median of about 3.5 hours

Dive Deep

Anima uses different models for different jobs.

• Andi (AI receptionist): built on GPT-Live for phone calls. A proprietary policy engine guides deterministic, clinically approved intake workflows. GPT-5.4 mini selects the right questionnaire, and the completed history appears as a structured case in the dashboard
• Documents: GPT-6 models extract sender, recipient, document type, and dates. A multi-stage pipeline matches each document to the right patient, identifies clinical codes, and suggests the next action
• Scribe: GPT 5.6 Luna turns consultation transcripts into structured notes, then reuses that context to draft referral letters and follow-up messages

Clinical coding was the hardest part: there are around 800,000 possible SNOMED codes, and practices choose differently. Instead of per-customer configuration or fine-tuning, Anima uses in-context learning to adapt to each practice, which needs models that follow instructions consistently.

These figures are all reported by Anima, and clinicians review the AI's output.

Wrap-up

• Anima supports phone intake, document processing, and note drafting with AI
• Phone waits dropped from 40-50 minutes to about 10, and document throughput roughly doubled or tripled
• Different models are used per workflow: GPT-Live, GPT-6, GPT-5.4 mini, and GPT 5.6 Luna
• Practice-specific differences are handled with in-context learning

A useful reference if you want to bring LLMs into healthcare or other operational workflows!