Understanding attributes

Attributes are the columns of the table. If objects answer what kinds of things do we track?, attributes answer what do we know about each one? A person has a name, an email, a phone number; a deal has a value and a stage; an event has a date and a headcount. Each of those is an attribute.

Attributes you define

Every object comes with sensible attributes out of the box, and you can add your own to any object — standard or custom. Adding an attribute is the equivalent of adding a column to a spreadsheet: decide what you want to know about every record of that type, give it a name, pick what kind of value it holds (text, a number, a date, a choice from a set, a link to another record), and it appears on every record of that object.

Two habits keep attributes useful. First, add an attribute only when you will filter, sort, or report on it. Second, prefer a choice from a fixed set over free text for anything you want to count later: Source: Website / Phone / Walk-in can drive a report; a free-text column with eleven spellings of "website" cannot.

AI attributes

Some columns should not need a human to fill them in. The guest told you their party size in the second email; the budget came up in a reply to your quote; the preferred date is in an older thread. In a spreadsheet, someone reads the conversation and types the value into the cell. In Winston, that column can be an AI attribute: you define the field, and Winston fills it in by reading the record's conversation history.

• You define what to capture — party size, budget, dietary requirements, preferred dates — as a normal attribute on the object.
• Winston reads the timeline — every email thread — and fills the value in, keeping it current as the conversation continues.
• You can always see and edit the value. An AI attribute is a regular field on the record, not a black box — if Winston got it wrong, correct it like any other cell.

AI attributes are the reason structure pays off: once the answer lives in a real column instead of buried in a thread, it can drive lists, filters, workflows, and reports. A list of wedding inquiries over 100 guests only takes one filter — because the guest count is a column, filled in by Winston, not a fact someone has to remember reading.

Where do the values live? On records — the next article. And when you are choosing which attributes your objects need, define your data model has a rule of thumb: start from the questions you ask every week, and work backwards to the columns that answer them.