The back end of a software system is seldom as tidy as the front. Any database that's been in use for a few years is liable to accumulate some records that, on their face, don't really make sense. This usually has to do with changes in business process and/or streams of data from other automated systems
which might arrive directly at the back end without any meaningful human review.
But a good pareidolic squint can transform these dull by-products of soulless mechanical processes into evidence of exciting secrets, whose stories are crying out to be told.
Such data glitches could be reported to a sort of unserious version of Wikileaks, whence they could provide a source of authorial inspiration.
It's the opposite of AI authorship; the machines unwittingly provide prompts to the humans, not vice versa.
Examples:
"On [date in 2019], ten road trains left [a mine site in the Pilbara] carrying over a thousand tonnes of cargo, and heading for [port]. There is no record that they ever arrived. This is their story."
On a lighter note,
"On [date in ?2003], [name of hotel] in [name of city] was block-booked by a wedding party. It was also block-booked by a convention of endoscopists. This is their story."
"On [date in the late 1990s], a group of Gulf-Arab princes opened accounts for holding Eurobonds with [a trading house] in London. But the 'know your customer' rules that we now have were not yet in place, and they protected their privacy by holding their accounts in the names of Old Master painters. This is their story."
For a small fee, the hosting organisation would periodically check in with contributors, and if any of those leaked a little too much reality, and become suddenly not-OK, the organisation would award a prize, perhaps posthumously, for services to creativity.
The best stories would be those which wove together many such prompts into a manic, Tom-Sharpe-like farce.