Copilot vs Classic

Creator Studio experiences are significantly upgraded in Copilot compared to Classic (legacy). View differences for each workspace:

• Paths plugins
• Queries plugins
• Events plugins work the same in Copilot.

Paths

Copilot's generative framework brings reasoning to your Paths use cases. Your use cases receive the following benefits:

1. Implicit slot detection

   Classic

   Classic bot needs to ask each question explicitly.

   

   Copilot

   Copilot auto can detect slots from previous utterances & ask explicitly when needed.

   

2. Handle multiple actions

   Classic

   Classic bot presents multiple actions, user can only pick 1 at a time.

   

   Copilot

   Copilot can conduct multiple logical actions at once.

   

3. Easy context switching

   Classic

   Classic bot requires stepping outside the use case to switch context.

   

   Copilot

   Copilot allows context switching between use cases & native skills. Users can pick up where they left of.

   

4. Improved dialogs

   Classic

   Classic bot solely relies on the use case configuration to generate dialogs.

   

   Copilot

   Copilot generates dialogs on-the-fly. It uses in-chat context & knowledge of your organization to be more specific.

   

5. Conversation reversibility

   Classic

   Once you answer a question, you need to restart the flow to modify it.

   

   Copilot

   Copilot allows switching responses to previous questions.

   

Queries

1. Better Query filtering

   Classic

   Classic bot always presents the entire Query response verbatim.

   

   Copilot

   Copilot is able to apply basic filters to the Query response on-the-fly.

   
2. Better Query summarization

   Classic

   Classic bot always presents the entire Query response.

   

   Copilot

   Copilot summarizes the relevant sections of a Query.

   

Copilot Implications

As you move to a more conversational, generative AI experience, you'll need to prepare for how plugins will behave differently:

1. Plugin design: Copilot reasons using the names, descriptions & configuration of your plugin. This means high-quality & unique names for your plugin, slots, actions are important. Please review the best practices guide for building with Copilot.
2. Limited context window: LLMs are subject to context window limitations, the maximum query response size that our platform can support is 8k tokens. For reference, 1 token = 4 characters, so for APIs, that’s typically going to be ~32k bytes (or 32kB) with ± 5% accuracy. If your query response exceeds 8k tokens, it will not work in Copilot.

Adaptive Responses

Sometimes the bot is confident a single plugin (Query, Article, Path, Form, etc.) will solve a problem. But often, questions have complex answers where multiple plugins may be relevant.

Copilot

Copilot default tries to clarify intent & find the most relevant plugin.

Classic

Classic shows list of relevant plugins for user to pick from.