Review volume outgrew the team's capacity to reply

Because the house's reputation is at stake, the owner replies herself. She has no time for it, so it happens irregularly — and only where it hurts.

Replying becomes a role with fixed time, not a person with good intentions. The owner sets tone and limits and signs off critical cases. Someone with a template and a tool handles the rest.

Every reply is written from scratch. Some are warm, some irritated, some boilerplate. A reader who reads three replies meets three different houses.

A one-page tone guide: salutation, what we always say, what we never say, how we handle unfair criticism. With it, a tool can deliver drafts that sound like the house. Without it, every AI text is interchangeable.

Either the portal's automatic reply is switched on and every answer sounds the same — or, out of fear of exactly that, nothing is switched on and everything stays undone.

AI writes the first draft per review, based on the tone guide and the specific review. A person checks and sends. Four eyes on anything below three stars. No text goes out that nobody has read.

The reply closes the case. That the same point — breakfast hours, parking, Wi-Fi — has come up for the fourth time escapes everyone, because nobody counts.

A monthly analysis of themes — by hand if need be, better with a tool that sorts reviews by topic. The pattern becomes an item for the department meeting. Replying is communication; analysing is operations.