Hardest Part

Balancing sensitivity and specificity in acute symptom routing to avoid the defensive medicine trap where the model uselessly sends every patient to the emergency room. This requires rigorous grounding in clinical protocols and exhaustive adversarial testing against false negatives.

Min Viable Scope

Build exclusively for adult primary care, handling only asynchronous symptom intake and routing to standard care pathways. Deliberately exclude pediatric cases, chronic disease management, and automated prescription generation to strictly cap clinical liability in the initial version.

Cold Start Problem

The model needs tens of thousands of validated symptom-to-outcome pairings to tune its routing thresholds before clinical deployment. Break this by partnering with a mid-sized telehealth provider to run the system in shadow mode on historical chat transcripts, benchmarking AI decisions against actual physician routing.

Time To First Value

2–4 weeks for EHR integration and shadow-mode validation before going live with patient traffic.

Data Moat Available

true

Technical Difficulty

High

Wedge

The beachhead targets high-volume independent urgent care chains. These organizations experience severe triage bottlenecks but lack the multi-year IT procurement cycles of large hospitals. After proving safety and wait-time reductions here, the product expands into mid-sized regional hospital emergency departments, and finally into inpatient specialist routing.

Timing

Large multimodal models process unstructured clinical text and standard lab values with accuracy matching human triage nurses. Acute nursing shortages force health systems to adopt automated clinical workflows they previously resisted.

Why This ICP

Emergency departments and high-volume urgent care centers face immediate financial penalties and liability for delayed critical care. They require immediate wait-time reductions, making them highly motivated buyers compared to standard primary care.

Size Of Prize

Approximately 14,500 combined emergency departments and urgent care centers in the US spend an average of $60,000 annually on dedicated triage nursing labor and routing overhead. This yields a total addressable prize of $870 million per year.

Gap Narrative

Clinical triage relies on overburdened mid-level providers who manually review intake symptoms and preliminary labs to determine patient acuity. This manual process causes severe delays in emergency departments and misallocates specialist time. No current system autonomously reads raw patient histories to assign an immediate acuity score and routing directive.

Defensibility

Defensibility stems from deep workflow lock-in and localized data compounding. As the system ingests a facility's specific triage outcomes and specialist availability patterns, it customizes its routing logic to their operational quirks. Replacing the system requires retraining a new engine on months of facility-specific routing preferences, establishing high switching costs.

Why This Thesis

A Service-as-Software approach directly replaces the labor bottleneck. These facilities need a completed task—a routed, prioritized patient—rather than another dashboard for nurses to manage.

SAM

~$400M-600M representing US outpatient medical imaging centers and mid-sized independent radiology practices

SOM

~$15M-30M achievable within 3 years targeting early-adopter US imaging networks

TAM

~30k global medical imaging centers and hospital radiology departments × ~$40k-60k/yr software allocation ≈ ~$1.2B-1.8B

Growth Rate

~12-18%/yr, driven by growing aging-population scan volumes outstripping radiologist workforce supply

Paid Comparable Spend

~$80k-150k/yr per center currently spent on outsourced locum tenens radiologists, off-hours teleradiology contracts, and manual PACS administrators

Kill Thresholds

• Average sales cycle exceeds 120 days for the first 10 pilot conversions
• PACS and EHR integration requires > 30 days per facility
• Radiologist manual override rate > 15% after week two of deployment
• Customer Acquisition Cost exceeds $20,000 while ACV remains below $40,000

Leading Metrics

• Time-to-first-triage-flag
• Radiologist manual override rate
• Average scan-to-read turnaround time
• Percentage of off-hours scan volume routed to AI
• Days to complete PACS integration

What Proves Right

Imaging centers route at least 40% of their off-hours scan volume through the triage system within the first 60 days of deployment. Early adopter cohorts maintain a net revenue retention over 110% as they expand the deployment across satellite clinic locations. Buyers sign $45,000 annual contracts with sales cycles under 90 days, successfully reallocating budgets from outsourced teleradiology.

What Proves Wrong

Radiologists manually dismiss or override the AI prioritization flags on more than 20% of scans, indicating a fatal lack of clinical trust. Integration bottlenecks with legacy PACS prolong deployment timelines beyond three months, destroying implementation margins. Hospital compliance and risk committees outright block procurement due to uninsurable liability concerns surrounding automated clinical decision support.