← Sticker Wall AGENT · DETAIL

Corti-Style Emergency Dispatch Drill System: Phone triage simulation and assessment for emergency centers, starting at 20,000 RMB annual fee per institution

Workflow: Build an emergency call simulator using large language models: the AI plays the role of the caller, simulating various a

AGENT

Key Fields

FIELD STAMPS

IndustryEducation / Knowledge

RegionEurope

ScaleSME

ChannelOnline

🔧 Workflow

Build an emergency call simulator using large language models: the AI plays the role of the caller, simulating various accents, emotional breakdowns, and medical history descriptions. Student dispatchers answer and triage in real time. The system automatically scores them according to structured triage protocols and generates a review report. The daily routine includes: updating the case library and scripts, running trial sessions with prospective clients, and iterating prompt engineering and scoring rubrics based on evaluation reports. Inputs are public triage protocols, desensitized case materials, and client feedback; outputs are simulated call recordings, scorecards, and exportable training records.

🛠 Setup Requirements

Requires real-time speech recognition and text-to-speech capabilities, which can be assembled using commercial voice APIs and GPT-grade large language models without the need for in-house model development. Public materials that require deep study include medical emergency triage workflows and cardiac arrest phone recognition standards, translating scoring logic into explainable structured rules. Positioned as a training tool rather than a clinical decision system, it does not require medical device registration qualifications. A 1-to-2 person collaboration can produce a demonstrable version in about 2 months. The initial focus is not technology, but infiltrating the emergency training circle to obtain real student trial feedback.

🧰 Toolchain

🔧 OpenAI Realtime Voice API or similar speech recognition and synthesis services
🔧 GPT-4 class large language model to act as caller and scoring judge
🔧 Public documents on emergency triage protocols and desensitized case libraries
🔧 Web-based call interface and scorecard generation scripts

💰 Revenue

① Emergency command training centers and hospital emergency departments (main revenue): institutions subscribe annually and pay tiered fees per student seat, ranging from 20,000 to 80,000 RMB per institution per year. Signing 5 clients yields an annual revenue of 100,000 to 400,000 RMB, which breaks down to about 15,000 to 30,000 RMB per month, accounting for roughly 100% of monthly revenue (estimated based on case profiles, independent verification pending); ② Case library and triage script customization packs: existing clients purchase add-on services per project, with pricing, number of purchasing clients, and revenue share not yet quantified; ③ Student seat expansion and renewal: clients pay additional fees to renew annual subscriptions for newly added seats, with pricing standards, expansion volumes, and revenue shares unspecified; ④ Opportunity item—annual subscription of teaching edition for medical school emergency teaching and research groups: charged annually per seat, with revenue share currently unclear.

💸 Cost

Speech recognition, speech synthesis, and LLM APIs are pay-as-you-go, resulting in monthly expenditures of about 800 to 3,000 RMB under centralized scheduling; servers, domain names, and storage cost about 200 RMB per month; the biggest early-stage cost is actually the time spent sorting and converting triage protocols and writing case scripts.

⏱ Time Investment

During the setup phase, 4 to 6 hours per day are spent mainly on writing case scripts and debugging scoring rules; during the operation phase, 1 to 2 hours per day are spent maintaining the case library, answering client trial questions, and following up on annual renewals, with weekends available for recording new simulated scenarios.

🚀 Getting Started

Step 1: Thoroughly read Corti's public product documentation and research reports on Copenhagen's cardiac arrest phone recognition to understand the evaluation metrics of real dispatch scenarios. Step 2: Use ready-made voice APIs to build a simulated call demo featuring only a single disease type. Step 3: Take the demo to local emergency training centers or medical school emergency teaching groups in exchange for a one-month free trial and written feedback. Once signed feedback is secured, negotiate pricing.

🔑 Keys to Success

✅ Focus solely on training and assessment without touching real-time clinical decisions, explicitly avoiding regulatory red lines for medical devices and writing this into contracts
✅ Ensure the case library strictly adheres to real triage protocols with explainable and verifiable scoring standards that convince supervising instructors
✅ Leverage market education already completed by industry benchmarks like Corti to lower customer acquisition and persuasion costs
✅ Allow assessment scorecards to be exported and archived, embedding them into clients' existing training record systems to build high-frequency dependence

⚠️ 风险

⚠️ If deemed to be used for clinical decision support, it faces regulatory compliance risks for medical devices, requiring legal review of contract wording
⚠️ Large institutions have long procurement cycles, and cash flow depends on a few clients, making the loss of a single client highly impactful
⚠️ Vendors like Corti may launch their own official training modules, squeezing the space for third-party drill tools

📌 Real Cases

📌 Corti's AI achieves an identification rate of about 95% for cardiac arrest in emergency calls, higher than the 73% of human dispatchers in Copenhagen, demonstrating that machine scoring in this scenario has a quantifiable benchmark and the scoring rationality of the drill system is endorsed by a benchmark
📌 Corti's Symphony model stack, trained on over 1.5 million hours of clinical audio, covers clinical decision support, automatic documentation, and medical coding, serving emergency centers in multiple European countries and proving the genuine institutional willingness to pay for emergency speech AI
📌 Third-party evaluation site LaunchTools AI rated Corti 4.4 stars, and AI Suggests aggregated over 2,100 user reviews to give it 4.5 stars, showing that AI tools for emergency and clinical dispatch have formed a searchable and discussable product category, opening up the accompanying training market

Sources