Training your assistant

Training means making the bot improve through use: adding what is missing, correcting mistakes, updating what changes. It is not a phase: it is a light routine that lasts as long as the service.

How it works

The logic is simple: critical points are covered before go-live, then you go online, and the bot is refined through real use. No endless questionnaires, no mandatory routines: your customers' real questions tell you better than anyone what is missing, the system detects them, and the gaps are filled over time.

The 5 types of training

Type	When	Who does it
Initial	Before go-live	You and glacom®, together
Automatic	Always	The system
Continuous	Always, in real use	The system and glacom®, you when you notice something
From feedback	At every report	You or glacom®
On cases	When something changes	You

What you need to do

Two things, when they happen: report a wrong answer when you notice it (it takes thirty seconds from the app), and update us when something changes in the company: a new product, a promo, a closure. The rest runs on its own.

What glacom® does

The system automatically detects questions left unanswered and keeps automatic sources updated. glacom® experts intervene on reports regarding functionality: see Reporting to glacom®.

Limits

• Training improves answers, it does not turn the bot into an employee: what to expect remains valid even after months.
• The bot does not train itself on conversations without human control: this is a choice, see Past conversations.

From the app

Much of the training is done conveniently from the app: reading conversations, evaluating answers, adding FAQ.