# How suggestions work

What Axelize reads, how it decides what to suggest, and how to change what you get.

## The three inputs

**The live captions.** Axelize reads the last stretch of the conversation, not
just the final sentence, so it can tell a pricing objection from a passing
mention of budget.

**The account context.** Cached before the call: the company's site, recent news,
your notes and files, and previous meetings with them.

**Your product records.** What you sell, its limits, and how you position it.

## Why suggestions are cached in advance

Retrieval during a call is too slow to be useful. Axelize caches the account's
context when the meeting is created, so at the moment an objection lands the model
is composing rather than searching.

## Changing what you get

If suggestions feel generic, it is almost always the product record. Add the
limits, the competitive positioning, and the objections you actually hear. Those
are what turn a plausible sentence into a specific one.