Sanaf, the AI employee

AI reputation management

Asking for reviews is a job nobody remembers to do, and answering a bad one is a job everybody puts off. This does the asking on schedule and drafts the answering, and it does not post anything under your name until you have read it.

What it does

Concretely

  • Requests customer reviews on the schedule and wording you set.
  • Monitors reviews as they arrive rather than when someone remembers to look.
  • Drafts a response to each one, in your voice, for you to approve.
  • Identifies unhappy customers early, from what they wrote rather than the star count alone.
  • Sends anything alleging something serious straight to you, unanswered.

The asking is the part that gets dropped

Most businesses with a reputation problem do not have unhappy customers. They have happy customers who were never asked, and a review page made up entirely of the small number of people motivated enough to write without prompting. Asking consistently is dull, which is precisely why it does not survive a busy month.

Nothing is posted in your name without you

This is the rule the job exists under: every public response is drafted and waits for approval before anything appears under your name. A reply is a public statement by your business, and an automated one that misjudges the tone of a complaint does more damage than the complaint did. Drafting is the time-consuming part, and that is the part handed over.

A serious allegation is not a drafting problem

A review that alleges something serious goes to you immediately and unanswered. There is a category of complaint where the correct first move is a person reading it carefully, possibly with advice, and where any drafted reply is the wrong artifact to have produced. Recognising that category and stopping is part of the job.

Questions

Frequently asked

Does it write fake reviews or incentivise them?
No. It asks real customers for honest reviews and drafts honest replies. Fabricated reviews are against the terms of every platform worth appearing on, and we do not build tools to produce them. The same rule applies to anything we publish about ourselves.
How does it know a customer is unhappy before they leave a review?
From what they write rather than from a score. Identifying unhappy customers is part of the job, and the signal is usually in the wording of an ordinary exchange well before it becomes a public rating. What it does with that is bring it to a person.
Can I approve replies in bulk?
You set how much room it has, per job. Some businesses read every draft, others promote it over time for the straightforward positive replies and keep approval on anything negative. The approval rules are yours to set and to tighten again.