A Finished AI Tool Is Not Proof It Works
In Steve Blank’s entrepreneurship class, student teams arrived with polished products built using AI. At first, that looked like remarkable progress. But some teams had spent less time learning what their customers actually needed. When they heard evidence that challenged their ideas, they were slow to change course.
That lesson matters for public services. A healthcare app can look complete without helping patients get appropriate care. An education tool can produce impressive work without improving learning. Speed of development tells us little about either outcome.
Before calling an AI pilot successful, we should ask: Who helped define the problem? What did patients, students, workers, or families say after using it? What changed when their experience contradicted the original plan?
A working tool is the start of a test, not the result of one.
Blank’s article is useful because it shows how easily a polished product can be mistaken for evidence. Public services need to keep their attention on the people served and the outcomes that matter to them.
Source: Steve Blank, “The Year AI Came For Us: Teaching Entrepreneurship Will Never Be The Same”
GPT assisted with analysis and drafting; PSA is responsible for the final judgment.
Blank’s experience speaks directly to public services. AI lets us build a convincing tool before we understand the people and problems it is meant to serve. For PSA, the first test is whether consumers and frontline workers helped define the need—and whether their experience can change the design. The next test is whether the whole service improves access and quality without quietly passing costs or responsibility to families.