The AP Is Using Robots To Write Earnings Reports

The Associated Press is going to start using “automation technology” AKA robots to deliver stories recapping the quarterly earnings reports of thousands of companies starting in July.

According to a post on the AP official blog, this will allow the AP to cover the earnings reports of approximately 4,400 companies each quarter, as opposed to their usual volume of around 300. This will not eliminate any jobs, but rather allow for the journalists at the AP to write more analysis around what the numbers mean. The earnings reports are mostly canned text anyway, consisting of a few important numbers and culminating in a brief analysis.

The company providing the technology, Automated Insights, recently closed a $5.5 million Series B round, with participation from Steve Case, Samsung, and the Associated Press.

From the perspective of a journalist, I must say that this is pretty exciting. Earnings season is a stressful time for all media publications, as it involves an often unclear schedule of multiple companies releasing numbers each day for weeks at a time. And there’s no standard for those earnings reports across these companies — each releases their quarterly financial summaries in a different format.

In fact, for the rest of us journalists, the AP has long been a more standardized place to ingest the numbers than the earnings reports themselves.

This means that it usually takes quite a bit of time and effort to pull together these numbers in a comprehensive and contextualized fashion. Plus, analytical stories around earnings reports can be published much faster, now that writers aren’t focusing on crunching and recapping the numbers the second they’re released.

The AP already uses automation technology for data feeds around sports, but this is the first time the media organization is using robots to write actual stories. According to the post, each story will be between 150 and 300 words.

Of course, box scores and earnings reports are one thing, but analytical reporting that requires source work and investigative digging or cultural insight will be more difficult (and even impossible?) to automate. Here’s hoping…