Context
In retail, store staff see what goes wrong first, but their suggestions and complaints tend to get lost for lack of a channel. EscutaAI is a Bázico B2B SaaS to collect them and turn them into action. I have been on it since the initial idea.
How it works
A company opens check-ins by store, area and period. Answers land in a feed. An AI layer reads each complaint, points out the pain behind it and suggests a fix. The manager approves or not, and what is approved becomes a task with an owner and a deadline.
The AI suggests; a person decides. No task is created without approval.
The decision: isolate tenants in the database
Several companies share the same database. Filtering by company only in application code lets one forgotten query leak another company’s data.
With in Supabase , the rule of who sees what lives in the database. Even a query without a filter only returns the rows of the signed-in user’s company.
What this does not solve
Rules in the database are harder to test and debug than rules in code: a wrong policy fails silently, returning nothing.
Outcome
The product is live at escutaai.ia.br.
