Getting a disease from a million cases down to a thousand is difficult. Getting it from a thousand to zero is harder, slower, and more expensive than everything that came before, because the last cases are the hardest to find. They sit in the places with the least data, the weakest connectivity, and the fewest laboratories.
That is precisely the kind of problem artificial intelligence is good at. Finding a faint signal inside an enormous amount of noise. Forecasting where the next case will surface. Reading a diagnostic where there is no laboratory to read it. The technology is no longer the constraint.
The constraint is people.
The men and women who decide whether an AI system helps or harms a health system are rarely the people who built it, and most have had no structured preparation for that decision. Meanwhile, the professionals who will use these tools every day have had almost no accessible way to learn to use them.
