In mid-2014 the Associated Press announced that the majority of its US corporate earnings stories would be produced using automation technology rather than written by hand. The AP partnered with the North Carolina company Automated Insights, whose Wordsmith software ingests structured financial data and generates short, human-readable prose. The wire service had previously published roughly 300 earnings stories each quarter; automation let it produce up to about 4,400, a more than tenfold increase in coverage.
According to the AP’s own announcement, the goal was not to cut staff but to free business journalists from rote data processing so they could focus on analysis, earnings-call reporting, and finding exclusive stories. Wordsmith generated stories of a few hundred words within moments of a company’s earnings release crossing the wire. The AP framed it as a way to expand coverage of smaller public companies that human reporters had never had time to write about.
This was one of the earliest large-scale uses of automated text generation in a major newsroom, predating the large language model era by years. It relied on template-and-rules natural language generation rather than neural models, but it established the template that later, much larger publisher-AI arrangements would follow: machines handle high-volume, formulaic copy while humans concentrate on judgment and original reporting. For business readers, it marked the moment automated writing moved from novelty to routine production infrastructure at a flagship news organization.