The Model Is Not the Product: What a Breast Cancer Lab Workflow Taught Me About AI Architecture
Friday Dec 4
11:30 –
12:15
Green Room
We often design AI systems assuming the model itself is the main differentiator. But what happens when your model reaches 62% accuracy in a workflow that requires over 99%?
And more importantly, what happens when the solution has nothing to do with improving the model?
This session explores lessons learned from real-world diagnostic and clinical software systems, where performance, reliability, and clinical constraints redefine what “good enough” means.
It challenges the model-centric view of AI systems and shifts the focus toward architecture, workflows, and system design as the true drivers of value.
A practical reflection on building AI systems in high-stakes environments where accuracy is not optional.
