## Two articles: pilots that do not stay, and edits that cannot be undone

Two memos from the lead on the same day, each a hypothesis to check before it was written up.
Registered as [D14 and D15](briefs.html).

- **[The pilot worked. Then somebody asked what else it could do.](article-pilots-do-not-stay-in-production.html)**
  The hypothesis: many generative AI and agent pilots never reach production, or are switched
  off after they do, because the business compares what the agent was meant to do with what it
  can do, at machine speed and on somebody's authority, and declines to sign for the difference.
  The article gives the published figures with the causes each source names, including the ones
  that point elsewhere: MIT's study blames learning, RAND blames problem definition and data, IDC
  blames data and integration. It shows seven cases where what a system said or did became the
  company's, and two often cited alongside them that do not fit. And it says plainly that no
  survey we found tests the mechanism. The fix it argues for is limits written in business units
  at design time, each beside whatever enforces it; the example numbers are labelled invented.
- **[A deleted meeting comes back. An edited one does not.](article-calendar-edits-cannot-be-undone.html)**
  Google Calendar keeps a deleted event in a trash for 30 days and documents no way for a user to
  restore an edited one, so edit, not delete, is the permission to worry about. Every recovery
  fact is quoted from Google's own pages. Checking the memo turned up two traps it did not have:
  deleting "this and following" events skips the trash, and an API change to the guest list
  discards the old list. It also turned up a qualification: Workspace editions with Vault keep
  earlier versions for an admin to export, though not to restore. Three places where Google's
  pages contradict each other are published unresolved, and nothing was tested on any account.

What was not done: the credit approval at two in the morning has no public source and is told as
an anecdote. The Calendar behaviour policies are in draft; the article quotes their rows and says
so. The claim that early users value their calendar over their inbox comes from a handful of
conversations, and is reported as that.