Set up guardrails to influence alby's responses in different situations.
In addition to the guardrails built into alby's foundational LLM models and prompt structure, retailers can add a layer of custom guardrails to influence alby's responses.
There are 3 types of guardrails brands can set up directly in the alby UI:
| Type | Description | Scope |
| Rules | Tell alby how to behave in certain situations | Product, Category, or Full Catalog |
| Verified Q&As | Script a response to certain questions | Product, Category, or Full Catalog |
| Global Guidelines | Tell alby how to behave in every response | Global |
Rules & Verified Q&A
Rules & Verified Q&A can be created:
- From a conversation in the Inbox > response details > thumbs down
- From the Widget Playground > response details >thumbs down
- From the Knowledge Section directly
Both Rules & Verified Q&A require 3 fields:
- 'When a customer says something like': Define when alby should follow the guidance. alby will then be able to reference the Rule or Verified Q&A when a shopper asks a similar question.
- Response / Action: Tell alby what to say (Verified Q&A) or what action to take (Rules).
- Scope: Determine if the guidance is relevant for a specific product, a category, or to the entire catalog. This adds a layer into when alby will reference the Rule / Verified Q&A.
 |
 |
Global Guidelines
Create Global Guidelines in Settings
- Global Guidelines provide the highest level of influence over alby responses. alby will be able reference Global Guidelines in every response to a shopper
- Global Guidelines can include things like overall tone, specific policies, or things you never want alby to say or respond to.
- There is a limit of 10 Global Guidelines per brand
